DRAFT SITE CONCEPTUAL MODEL FOR THE MIGRATION AND FATE OF CONTAMINANTS IN GROUNDWATER AT THE SANTA SUSANA FIELD LABORATORY, SIMI, CALIFORNIA

Transcription

1 DRAFT SITE CONCEPTUAL MODEL FOR THE MIGRATION AND FATE OF CONTAMINANTS IN GROUNDWATER AT THE SANTA SUSANA FIELD LABORATORY, SIMI, CALIFORNIA OVERVIEW AND 20 SITE CONCEPTUAL MODEL ELEMENTS VOLUME 1 OF 4 Prepared for: THE BOEING COMPANY NATIONAL AERONAUTICS AND SPACE ADMINISTRATION UNITED STATED DEPARTMENT OF ENERGY December 2009 Prepared by: THE SSFL GROUNDWATER ADVISORY PANEL, in association with THE UNIVERSITY OF GUELPH MONTGOMERY WATSON HARZA HALEY & ALDRICH AQUARESOURCE INC.

2 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL DRAFT December 2009 Site Conceptual Model Definition from U.S. EPA, 1993: The site conceptual model [SCM] synthesizes data acquired from historical research, site characterization, and remediation system operation. The site conceptual model typically is presented as a summary or specific component of a site investigation report. The model is based on, and should be supported by, interpretive graphics, reduced and analyzed data, subsurface investigation logs, and other pertinent characterization information. The site conceptual model is not a mathematical or computer model, although these may be used to assist in developing and testing the validity of a conceptual model or evaluating the restoration potential of the site. The conceptual model, like any theory or hypothesis, is a dynamic tool that should be tested and refined throughout the life of the project. The model should evolve in stages as information is gathered during the various phases of site remediation. This iterative process allows data collection efforts to be designed so that key model hypotheses may be tested and revised to reflect new information. The conceptual model serves as a foundation for evaluating the restoration potential of the site and, thereby, technical impracticability as well. the clarity of the conceptual model (and supporting information) is critical to the decision-making process.

3 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL DRAFT December 2009 Summary: Site Conceptual Model of Contaminant Transport in Groundwater for SSFL Santa Susana Field Laboratory Located on top of a sandstone mountain (2850 acres) Annual rainfall 18.6 inches About half of the groundwater originating on the SSFL discharges along slopes at seeps, springs and phreatophytes. no contaminants found Water table near mountain top Contaminants degrade or decay Strongly retarded, stationary plume fronts Vadose Zone Essentially all contaminant mass is in rock matrix TCE is deepest due to DNAPL flow initially: no DNAPL remains now Annual Recharge ~ 0.4 to 1.3 Inches Ubiquitous interconnected, systematic fractures with small hydraulic apertures Strongly attenuated plumes Sandstone has reactive components, low matrix K and low to moderate bulk hydraulic conductivity Shallower perchlorate and tritium plumes Faults do not drain the mountain Fresh water marine salt flushed away over millions of years Perched groundwater occurs locally, flows into deeper groundwater Shale zones generally lower bulk K About half of the groundwater from SSFL follows deep regional paths Nearly immobile brackish water relic marine salt Schematic cross section with vertical exaggeration (Not To Scale) Prepared by SSFL Groundwater Panel December 2009 Contaminants, primarily TCE, originating from industrial activities at SSFL decades ago are present in the fractured sandstone and shale beneath the SSFL mountain. Plumes formed as a result of transport by active groundwater flow in networks of ubiquitous but small interconnected fractures. The groundwater system beneath the SSFL is continually replenished by rainfall infiltrating through the vadose zone to the water table causing a groundwater mound hundreds of feet above the adjacent population centers of Simi and San Fernando Valleys and Chatsworth. This mounded groundwater results in a groundwater flow system causing the transport of contaminants downward and outward; a portion moves towards seeps and springs along the mountain slopes near the SSFL and a second component moves deeper into the regional groundwater system. The rock matrix blocks between the fractures have large porosity but very low permeability; these properties provide a large storage space for contaminants and limit their mobility. Groundwater flow in the fractures has caused plumes to develop to a size and shape sufficient to characterize and monitor. Other processes have caused the rates of plume expansion and measured maximum concentrations in both the source zones and plumes to decline over time (decades) such that the plumes have now or nearly reached stationary positions. This decline of maximum concentrations and near immobility of plume fronts has for all contaminants been caused by combined effects: volumetric groundwater flow is relatively small, molecular diffusion, transverse dispersion, and in some cases sorption, degradation, or decay occurs at SSFL. The molecular diffusion into the rock matrix is a dominant mechanism that tends to bind the contaminant mass near source zones for long periods of time; meanwhile other degradation mechanisms continue to act on that bound mass. Therefore, transport of contaminants is limited and groundwater discharging to seeps and springs or sampled at offsite wells is uncontaminated. The hydrogeologic system has complexity due to historical pumping, topography, the nature of fracture networks, and sloping beds and faults; however, the evidence shows no indication of potential longdistance contaminant migration on large, extensive preferential pathways. Some contaminant mass remaining in the vadose zone has the potential to contribute contaminant flux to the groundwater plumes; however, conditions at these sites have reached maturity and vadose zone contaminant fluxes have minimal or no influence on the location or concentrations observed at plume fronts. Although contaminants of the type occurring at the SSFL have been and likely will continue to be found in valley floor wells located in the population centers adjacent to the SSFL, these occurrences cannot be attributed with reasonable scientific plausibility to the SSFL. Such contaminant sources are common in a wide variety of commercial and light industrial activities in communities across the United States. Although all of the scientific investigations (which began in 1983 and continue to date) of contaminant distribution and behavior concerning SSFL groundwater are not yet complete, the current understanding of the contaminant distribution, transport, and fate is now well advanced and strongly supports the conclusions reported at this time (late 2009). This SSFL site conceptual model (SCM) fits within a more general SCM for contaminant behavior in fractured sediment rock tested intensely at other sites in North America.

4 DRAFT SITE CONCEPTUAL MODEL FOR THE MIGRATION AND FATE OF CONTAMINANTS IN GROUNDWATER AT THE SANTA SUSANA FIELD LABORATORY, SIMI, CALIFORNIA SITE CONCEPTUAL MODEL DOCUMENT 0-1: SYNOPSIS Prepared for: The Boeing Company National Aeronautics and Space Administration United States Department of Energy Prepared by: The SSFL Groundwater Advisory Panel

5 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-1: Synopsis DRAFT December 2009 This document presents the results of intensive investigations concerning the development and assessment of the site conceptual model (SCM) for subsurface contaminant behaviour at the Santa Susana Field Laboratory (SSFL). Field investigations of groundwater contamination commenced in 1984 at the SSFL, which encompasses 2850 acres located on the top of a bedrock mountain formed of fractured porous sandstone with shale and siltstone interbeds (the Chatsworth Formation, Cretaceous age). The complexity of the site hydrogeology is enhanced by dipping strata and several faults. The SSFL was an active industrial facility serving the space and nuclear industries beginning in the late 1940 s, with greatly diminished activity beginning in the 1980 s and decommissioning now in progress. When the facility was active, various types of contaminants entered the subsurface at different locations on the property; chlorinated solvents (TCE and its daughter products) are now the most distributed in the groundwater. The water table in the bedrock is shallow, generally within feet of ground surface, except where lowered by pumping for operational needs or as an interim remedial measure. The water table elevation beneath the site is high, rising feet above the populated lowlands nearest the SSFL (Simi, Chatsworth, and San Fernando Valley). In the bedrock, groundwater flow from the SSFL goes downward and outward towards these communities. This condition highlights the need for determining the subsurface transport and fate of SSFL contaminants. When the earliest phase of groundwater investigations at SSFL began, a preliminary SCM was proposed in which rapid long-distance contaminant transport with minimal attenuation occurred along some major discrete geologic feature pathways such as faults (the rapid transport model). In 1997, an alternative SCM for transport and fate of bedrock contamination at the SSFL was proposed (the matrix diffusion model) by the Groundwater Advisory Panel comprised of Drs. Cherry, McWhorter, and Parker when the Panel was initiated. In this model, active groundwater flow was conceptualized to occur almost exclusively in the bedrock fractures, with little physical groundwater flow in the adjacent unfractured sandstone blocks. The process of molecular diffusion moved contaminants in the fractures out of the active groundwater system into the nearly immobile groundwater in porous but low permeability sandstone. These conditions were conceptualized to result in essentially all contaminant mass at the site residing in the low-permeability rock blocks between the fractures. The matrix diffusion effect combined with other processes causes strong attenuation of both source zone and plume maximum concentrations as well as strong plume-front retardation relative to the active groundwater flow in the fracture networks. 1

6 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-1: Synopsis DRAFT December 2009 The scientific literature shows the nature of the transport and fate of groundwater contaminants in fractured porous sedimentary rocks, such as the Chatsworth Formation, is considerably different than in granular aquifers such as sand and gravel. However, when the most intensive phase of investigations regarding transport and fate in the Chatsworth Formation began in 1997, the literature contained only minimal insights concerning behaviour of plumes in fractured sedimentary rock. As intensive site investigation work has progressed at the SSFL, mounting evidence supporting the matrix diffusion model has been obtained, making the rapid transport SCM increasingly improbable. Over the last decade or more, site investigations have focused on testing the matrix diffusion conceptual model by more thoroughly sampling source and plume areas, resolving the complex geological stratigraphy and structure, deciphering the 3-D groundwater flow system, and characterizing the role of the fracture network in the transport of contaminants-of-concern (COC s). Examples of specific methods used to investigate the site include geologic mapping, rock core contaminant analyses, hydraulic tests at several spatial and temporal scales, application of hydrochemical and environmental isotope techniques, collection and analyses of borehole geophysical logs, and seeps sampling and characterization. In addition, a number of simple and more complex mathematical models have been applied using standard and new innovative approaches to quantify components of the water balance and simulate historical and future site conditions. The first comprehensive report on the status of the SSFL site conceptual model was issued in 2000 (Conceptual Site Model, Movement of TCE in the Chatsworth Formation, Montgomery Watson, 2000). Since then, many additional reports providing information about specific components have been produced. This present comprehensive report provides analysis and synthesis of information concerning groundwater contamination, transport, and fate in the bedrock acquired since studies began at SSFL. It includes interpretations based on the most scientifically advanced understanding of contaminant behaviour in a fractured sedimentary rock setting. Preparing a thorough analysis of the groundwater conditions and fate of SSFL contaminants required the cooperative efforts of the site owners and a large team of scientists and engineers. Those responsible for the studies reviewed in this SCM report include the Panel and more than 40 additional professionals. The main conclusion of studies reported in this document is that essentially all (>99.9%) of the contaminant mass resides in the low-permeability rock matrix in close proximity to where the contaminants entered the subsurface decades ago. This is the case even though contaminants occur at many locations in the subsurface where active groundwater flow occurs in bedrock fractures. A few areas 2

7 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-1: Synopsis DRAFT December 2009 exist where SSFL plume fronts have been identified or projected to have migrated beyond the property boundaries. However, these plume fronts remain beneath the mountain slopes adjacent to SSFL and have not reached off-site receptors. Although some bedrock contamination is present beyond the SSFL property boundary, the contaminants have not been found in any off-site seeps and springs along the mountain slopes or in off-site wells. Natural processes over past decades have caused strong attenuation of the maximum plume concentrations and retardation of plume front migration, and are responsible for the lack of reported impacts to off-site receptors. These processes will continue to govern the bedrock contaminants in the future. The matrix-diffusion SCM can be used to reliably forecast the expectation of no off-site impacts in the future. The combined natural effects of matrix diffusion, sorption, decay, and degradation form the basis of this forecast. This work provides the scientific framework required to design plans for further work to fill identified data gaps and to formulate a long-term monitoring plan for the SSFL to confirm the forecasts concerning contaminant behaviour. This SCM report is comprised of an introductory section followed by twenty sections. The first of these twenty sections describes the strategy and methods used for the site characterization. The other nineteen sections address specific elements of the SCM, where each element represents a hypothesis or question. Except for three, each element is comprised of between two and ten documents for a total of 65 documents pertaining to the many particular data sets and/or modeling tasks. Some of these documents have been published in refereed (peer reviewed) scientific journals and many others represent manuscripts in progress that will eventually be submitted for publication. 3

8 DRAFT Overview of the Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the Santa Susana Field Laboratory, Simi, California Prepared for: The Boeing Company, NASA, and U.S. DOE by John Cherry, David McWhorter, and Beth Parker SSFL Groundwater Advisory Panel With the Assistance of: MWH, Haley & Aldrich, AquaResource, and Schlumberger Water Services DRAFT December 2009

9 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 SCM DEFINITION FROM U.S. EPA (2003): The site conceptual model [SCM] synthesizes data acquired from historical research, site characterization, and remediation system operation. The site conceptual model typically is presented as a summary or specific component of a site investigation report. The model is based on, and should be supported by, interpretive graphics, reduced and analyzed data, subsurface investigation logs, and other pertinent characterization information. The site conceptual model is not a mathematical or computer model, although these may be used to assist in developing and testing the validity of a conceptual model or evaluating the restoration potential of the site. The conceptual model, like any theory or hypothesis, is a dynamic tool that should be tested and refined throughout the life of the project. The model should evolve in stages as information is gathered during the various phases of site remediation. This iterative process allows data collection efforts to be designed so that key model hypotheses may be tested and revised to reflect new information. The conceptual model serves as a foundation for evaluating the restoration potential of the site and, thereby, technical impracticability as well. the clarity of the conceptual model (and supporting information) is critical to the decision-making process. i

10 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 ABBREVIATIONS amsl bgs Boeing COC DFN DNAPL DOE DTSC EPM FSDF gpm HSC K b K f K m LNAPL mm NASA NDMA RI SCM SD SSFL TCE VOC above mean sea level below ground surface The Boeing Company contaminant of concern discrete-fracture network dense non-aqueous phase liquid U.S. Department of Energy Department of Toxic Substances Control equivalent porous media former sodium disposal facility gallons per minute Health and Safety Code bulk hydraulic conductivity hydraulic conductivity of fracture network rock matrix hydraulic conductivity light non-aqueous phase liquid millimeter National Aeronautics and Space Administration n-nitrosodimethylamine Remedial Investigation site conceptual model site derived Santa Susana Field Laboratory trichloroethylene volatile organic compound ii

11 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 PANEL MEMBER BIOGRAPHIES John A. Cherry holds geological engineering degrees from the University of Saskatchewan and University of California Berkley, and earned a Ph.D. in hydrogeology from the University of Illinois in He joined the faculty at the University of Waterloo in 1971 for field research on the migration and fate of contaminants in groundwater and their remediation. He retired from Waterloo in July 2006, but he continues research as a Distinguished Professor Emeritus. He co-authored the textbook Groundwater with R.A. Freeze (1979) and co-edited and co-authored several chapters in the book Dense Chlorinated Solvents and Other DNAPLs in Groundwater (1996). He has participated in the development of technologies for groundwater monitoring and remediation, co-holds several patents, is a Fellow of the Royal Society of Canada, and has received awards from scientific and engineering societies in Canada, the United States, and the U.K. He held the Research Chair in Contaminant Hydrogeology at the University of Waterloo from 1996 to 2006 and is currently the Director of the University Consortium for Field-Focused Groundwater Contamination Research, established in 1988 and now based at the University of Guelph, Guelph, ON. David B. McWhorter received a Bachelor s degree in petroleum engineering from Colorado School of Mines and a Ph.D. in groundwater hydrology from Colorado State University (CSU). He was a Professor of Agricultural and Chemical Engineering at CSU from 1970 to 1999 and is currently a Distinguished Professor Emeritus at CSU and an independent consultant. He has extensive experience in teaching and research in multi-phase flow in porous media, specializing in the combined use of mathematical models and laboratory experiments to develop practical methods for more effective site analysis and remediation. He co-authored with D.K. Sunada the textbook Groundwater Hydrology and Hydraulics (1977). He has won university and national awards, most recently the M.K. Hubbert Award of the National Ground Water Association. He recently led a research team focused on various aspects of DNAPL mass removal and related issues of exposure reduction. For more than 30 years, he has been a frequent consultant to government and industry on problems involving DNAPLs and LNAPLs in the subsurface. Beth L. Parker has degrees in environmental science and economics (B.S., Allegheny College), environmental engineering (M.S., Duke University, 1983) and hydrogeology (Ph.D., University of Waterloo, 1996), and is now a Professor in the School of Engineering at the University of Guelph. She was a research faculty member at the University of Waterloo from 1996 to 2007 and has more than 25 years of experience as a groundwater professional investigating subsurface contamination issues at industrial sites. Prior to obtaining her Ph.D., she worked for more than 5 years for a large corporation in New York State managing DNAPL site investigations and remediation in fractured bedrock. Her current research and consulting activities emphasize field and laboratory studies of DNAPLs in sedimentary rocks, clayey deposits, and heterogeneous sandy aquifers, and focus on the effects of diffusion into and out of low permeability zones and on DNAPL fate, plume attenuation, and controls on remediation. She is currently involved in research and technology demonstration projects at Superfund and RCRA facilities in the United States and similar sites in Canada. In July 2007, she was awarded an NSERC Canada Industrial Research Chair in Fractured Rock Contaminant Hydrology. In December 2009, she received the John Hem Award from the Association of Groundwater Scientists and Engineers of the United States National Groundwater Association. iii

12 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 TABLE OF CONTENTS SCM Definition from U.S. EPA (2003):... i Abbreviations... ii Panel Member Biographies... iii Table of Figures... v 1.0 Problem Statement and Scope of This Document Historical Context and Research Framework for the SSFL SCM Investigation Introduction to the General Conceptual Model for Contaminant Behavior in Fractured Sedimentary Rock Use of Mathematical Models at the SSFL Elements of the Site Conceptual Model for the SSFL Main Conclusions of the SSFL SCM Implications for Monitoring Implications for Remediation Current Status of Investigations and Reporting Concerning the SSFL SCM References iv

13 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 TABLE OF FIGURES Figure 1-1 Regional topography and location of the Santa Susana Field Laboratory (SSFL) with sub-areas... 3 Figure 1-2 Schematic of SSFL Site Conceptual Model (SCM)... 4 Figure 1-3 Map Showing Areas of Persistent Bedrock Contamination... 5 Figure 3-1 Measured Chlorinated Solvents in the Rock Matrix Figure 3-2 Evolution of Contaminant Source Zone and Plume in Sedimentary Rock Figure 3-3 Schematic of Average Linear Groundwater Velocity in Fracture Rock Figure 3-4 Schematic of Fracture Porosity versus Matrix Porosity Figure 3-5 Conceptualization of Bulk Fracture Porosity Figure 3-6 Schematic of Retardation in a Single Fracture Figure 5-1 The Discrete Fracture Network Approach (DFN) for Characterization of Bedrock Groundwaters Figure 5-2 Matrix Porosity Image Figure 5-3 Minerals and Organic Carbon in SSFL Rock Samples Figure 5-4 Conceptual Representation of Lithologic Beds, Bedding Planes, and Joints Figure 5-5 Randomly Generated Fracture Network Using FRACTRAN Figure 5-6 Diagrams of Water Table Shape Figure 5-7 Plan View Conceptualization of Falut/Shear Zone Features Figure 5-8 Fracture Aperture Concepts and Definitions Figure 5-9 Schematic of the Subsurface Hydrologic Continuum Figure 5-10 Conceptual Representation of Groundwater Recharge Figure 5-11 Geologic and Hydrogeologic Variability as Represented in the FEFLOW Model Figure 5-12 Chloride, Sulphate, and Bicarbonate Versus Depth Figure 5-13 Springs, Seeps, and Phreatophytes at and Around the SSFL Figure 5-14 Conceptual Diagram of TCE DNAPL Source Zone Evolution in Fractured Porous Rock Figure 5-15 TCE Source Mass Distribution v

14 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-16 Conceptualization of Transport Distances in a Single Fracture for TCE and Various Radionuclides Relative to Water Figure 5-17 Conceptual Model for Contaminant Transport from the Vadose Zone Figure 5-18 Example of Plumes Generated Using FRACTRAN Figure 5-19 FRACTRAN Simulations of Contaminant Transport in a Single Fracture Figure 5-20 FRACTRAN Simulations of Attenuation Processes on TCE Plumes Figure 6-1 Summary of Degradation Pathways Figure 6-2 Recent TCE Concentrations (2009 sampling round) compared to Historical Maximum TCE Concentrations vi

15 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December PROBLEM STATEMENT AND SCOPE OF THIS DOCUMENT The Santa Susana Field Laboratory (SSFL) occupies 2,850 acres on top of a fractured sandstone bedrock mountain (elevation range 1,175-2,265 ft above mean sea level [amsl]) near Simi, California (elevation 1,000 ft amsl). The SSFL is located in the southeast corner of Ventura County, 29 miles northwest of downtown Los Angeles, California. The location of the SSFL and its surrounding vicinity are shown in Figure 1-1. Between the mid 1940 s and 1990 s, various research and industrial activities on this land caused contamination in parts of the thin overburden deposits and more generally in the underlying fractured bedrock of the Cretaceous age Chatsworth Formation which is composed of sandstones with interbedded siltstones and shales. This document presents a summary of the site conceptual model (SCM), shown schematically in Figure 1-2, describes the distribution, migration, and fate of contaminants in the bedrock at the SSFL and the investigations conducted to develop and assess this model and alternative models. The SSFL is jointly owned by The Boeing Company (Boeing) and the federal government (administered by the National Aeronautics and Space Administration [NASA]) and is operated by Boeing. The U.S. Department of Energy (DOE) used a portion of the SSFL; however, active DOE operations are no longer occurring and the facilities are undergoing decommissioning and demolition. Numerous investigations have shown the Chatsworth Formation beneath the SSFL has been impacted by historic releases of various chemicals from operational activities. Trichloroethene (TCE) is the compound detected in groundwater at the highest concentration and is the most widespread (lateral extent and greatest depth) (Figure 1-3). TCE is the most common organic contaminant of industrial origin found in groundwater worldwide and occurs at many thousands of sites in North America, Europe, and Asia in fractured sedimentary bedrock. The contaminants of primary concern include chlorinated solvents (mostly TCE), perchlorate, 1,4-dioxane and n- nitrosodimethylamine (NDMA), tritium, and some radionuclides, all of which initially entered the subsurface decades ago. The water table beneath the SSFL is relatively shallow (generally ft below ground surface [bgs]) creating a mound of groundwater at the site that plateaus at feet above the lowlands that contain the population centers of Simi, the City of Chatsworth, and the San Fernando Valley. During periods of site operation SSFL pumping 1

16 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 temporally lowered portions of the water table hundreds of feet at some locations. Groundwater originating at the SSFL flows generally downward and outward towards the surrounding valley floors providing potential pathways for contaminant migration. Therefore, the current distribution, transport, and long-term fate of the contaminants originating at the SSFL are being rigorously investigated using state-of-the-science methods. Subsurface investigations began in 1983 when the first monitoring wells were installed. The understanding of the contaminant distribution, transport, and fate at the SSFL was the subject of a previous comprehensive report compiled in 2000 based on all data available at that time (Montgomery Watson, 2000). The results of this work provided the basis for launching the next phase of characterization from 2000 to The results of all groundwater investigations are documented in the Site-Wide Groundwater Remedial Investigation Report submitted to the Department of Toxic Substances Control (DTSC) on December 15, 2009, pursuant to the Consent Order for Coreective Action signed by the California Environmental Protection Agency, Department of Toxic Substances Control (DTSC), The Boeing Company (Boeing), the National Aeronautic and Space Administration (NASA), and the Department of Energy (DOE) in August In addition to the comprehensive report of investigation results in the remedial investigation (RI) document, supporting detailed data sets and analyses are organized using the U.S. EPA Site Conceptual Model approach (see page i) and presented in this SCM report. The documents making up this report are founded on extensive, comprehensive, and innovative scientific investigations conducted by a large team of scientists and engineers that applied standard and new investigative methods and state-of-the-science mathematical models. Document 0-3 identifies all of the contributors to this report. This comprehensive SCM update is being provided at this time because the SCM has been greatly refined based on information collected since 2000, and there is interest in communicating this advanced state of the SCM to project stakeholders. In parallel with the advancement of the SSFL site investigations, there have been major advances in the general understanding of contaminant behavior in fractured sedimentary rock based on studies completed at other sites, as well as theoretical studies. These advancements provide an improved framework within which to position the site specific SSFL results. 2

17 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 1-1 Regional topography and location of the Santa Susana Field Laboratory (SSFL) with subareas 3

18 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 1-2 Schematic of SSFL Site Conceptual Model (SCM) Contaminant plumes occur in the fractured bedrock (Chatsworth Formation) at many locations due to industrial activities decades ago. These plumes are characterizable and monitorable and have now attained nearly stationary positions due to the combined effects of diffusion into the low permeability rock matrix block Between fractures, dispersion sorption and degradation/decay. The water table is high in the mountain because the recharge is small and the bulk hydraulic conductivity (K b ) of the mountain is small to moderate. Shale strata have fractured but have low K b. Faults have variable K b but no evidence of faults with lateral extensive high K b causing mountain drainage. 4

19 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 1-3 Map Showing Areas of Persistent Bedrock Contamination Map showing areas where analytical results of groundwater samples from many monitoring wells show the existence of persistent contamination by TCE and other chemicals of concern. 5

20 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December HISTORICAL CONTEXT AND RESEARCH FRAMEWORK FOR THE SSFL SCM INVESTIGATION The production of this document was directed by the SSFL Groundwater Advisory Panel comprised of Drs. John Cherry, Distinguished Professor Emeritus, University of Waterloo, and Adjunct Professor, University of Guelph, Guelph, ON; David McWhorter, Distinguished Professor Emeritus, Department of Chemical and Agricultural Engineering, Colorado State University, Fort Collins, CO; and Beth Parker, Professor and Canadian NSERC Industrial Research Chair in Fractured Rock Contamination Hydrology, School of Engineering, University of Guelph, Guelph, ON, and formerly a research professor at the University of Waterloo ( ). The Panel members first visited the SSFL in 1996 for familiarization with the subsurface conditions. The Panel was formed by Rocketdyne at that time to provide advice and guidance concerning the design and execution of investigations to characterize the distribution, transport, and fate of contaminants in the bedrock beneath the SSFL. The panel was also tasked to develop a bedrock site conceptual model (SCM) for contaminant behavior. It was recognized that the conventional approach for investigating contaminated bedrock sites was unsuitable because it was based primarily on equivalent porous medium concepts, approaches deemed inadequate for assessment of contaminant transport in fracture networks. Based on a decade of previous research, the Panel members developed a preliminary diffusion based general conceptual model for the behavior of chlorinated solvents released as the immiscible liquids or dense non-aqueous phase liquids (DNAPLs), into fractured porous geologic media, such as fractured sandstone (Parker et al. 1994; Parker et al. 1996; Parker et al. 1997). This general conceptual model formed the basis for the initial site-specific SCM for the SSFL proposed by the Panel in In this conceptual model, contaminant source zones and plumes are strongly attenuated and plume front movement is retarded due to diffusion effects. An alternative conceptual model for contaminant behavior at the SSFL had been proposed prior to involvement of the Panel. In this prior model, plumes underwent minimal attenuation and moved rapidly away from their source areas by flow channeled in extraordinary flow zones in geologic features such as faults. Therefore, the Panel proposed investigations directed at assessing these two models and other conceptual models, involving combinations of the two as working hypotheses on which to found future work. A conceptual model characterizing plume 6

21 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 scale contaminant behavior that depends on flow in fractures had not been subjected to comprehensive assessment at any contaminated fractured rock sites. Standard methodologies were not adequate to obtain data sets needed to test the model in Therefore, concurrent with the investigation conducted at the SSFL, major new field investigation methods were identified, developed, applied, and improved as needed at the SSFL. In the development information on the overall site condition including the geology and contaminant behavior at the SSFL, two consulting firms (since 1987 Haley & Aldrich, previously represented by Groundwater Resources Inc, and Montgomery Watson Harza (MWH) since 1999) have had major responsibilities for project management, geologic mapping, drilling, well installation, groundwater sampling, hydraulic testing, and data compilation and display. Numerous reports have been prepared regarding data collected at the SSFL. Although specific references are not provided for these reports, additional information concerning the SSFL is included in the Draft Site-Wide Groundwater Remedial Investigation Report (MWH 2009) and its associated appendices. Since 1999 three senior professionals from these companies (Mr. Richard Andrachek, P.E. of MWH; Dr. Ross Wagner, P.G., formerly of MWH; and Mr. Larry Smith, P.G. of Haley & Aldrich) have been active participants in meetings involving the Panel. In 2002, The Boeing Company retained AquaResource, Inc. to develop a three-dimensional numerical model to describe the bedrock groundwater flow at the SSFL and Mr. Paul Martin, P.Eng., and Mr. Daron Abbey, P.Geo., from this firm, have participated in many meetings concerning the SCM. In 2005, The Boeing Company retained Dr. William Woessner, Regents Professor and Department Chair, Department of Geosciences, The University of Montana, Missoula, MT, to provide oversight and advice to AquaResouce and the Panel concerning the modeling and to participate in SCM discussions with the Panel. In mid 2009, The Boeing Company retained Schlumberger Water Services to use state-of-the-science methods to analyze site geophysical and geological borehole records. This work was specifically focused on charactering the subsurface fracture frequency and network. At the University of Guelph the Panel has been assisted frequently since 1997 by Mr. Steven Chapman, P.Eng., M.Sc., with both the field work and modeling and, since 2004, by Ms. Amanda Pierce, M.Sc., with hydrogeochemical and project management aspects of the research. 7

22 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 The development of the SSFL SCM and consequently this document are based on SSFL site data and also, importantly, on Panel members and their collaborators academic research and field experience characterizing the behavior and fate of dense immiscible organic chemicals (e.g. chlorinated solvents) and dissolved contaminants in fractured porous rocks (including sandstone and shale) at several industrial sites. Since 2003, the suite of advanced methods used to investigate the distribution, transport, and fate of industrial contaminants at the SSFL have also been applied intensely by the Panel and collaborators at other fractured sedimentary rock sites in the United States and Canada. The strong commonalities between sites add confidence to the reliability of each of the site-specific SCMs, including the one developed for the SSFL. Therefore, this SSFL SCM is based on a more general conceptual model for contaminant behavior and fate in sedimentary rocks recently developed and tested at several sites. A comprehensive investigative methodology specific to fractured sedimentary rock, known as the discrete fracture network approach (DFN approach), was developed to support the data required to test the general and site specific conceptual models. The development of this DFN approach has involved broad research programs supported financially by government agencies in Canada and the United States and several corporations including Boeing. Diffusion-driven contaminant mass transfer is a strong process affecting the transport and fate of COCs at each of the bedrock sites included in the research program. The work at SSFL is a component of this research effort. As a result, a number of graduate student thesis studies (Sterling 1999; Hurley 2003; Pierce 2005; Darlington 2008; Zimmerman in prep) and also work by collaborating faculty members, research associates, and others have been produced. Therefore, while elucidating subsurface contaminant distributions and behavior at this site, the investigations conducted at the SSFL are also serving to advance the science of contaminant behavior in fractured sedimentary rock. Some of the research has been published in peer-reviewed scientific journals and preparation of many more scientific papers based on results of investigations at the SSFL is in progress. As part of this effort to make the knowledge produced by the SSFL investigations readily available to the global scientific community and the public, the site data are being assembled into a readily accessible relational data management system. 8

23 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December INTRODUCTION TO THE GENERAL CONCEPTUAL MODEL FOR CONTAMINANT BEHAVIOR IN FRACTURED SEDIMENTARY ROCK This description of the general conceptual model is extracted in part from Parker (2007). The behavior of contaminants in fractured rock is now one of the few remaining scientific frontiers in physical hydrogeology. The status of knowledge concerning groundwater flow and contaminant migration in fractured rock has been reviewed by the U.S. National Research Council (1996), Lapcevic et al. (1999), Berkowitz (2002) and Neuman (2005). These reviews detail considerable published literature concerning the conceptual nature of fractures and hydraulic conditions in fracture networks based on borehole investigations in uncontaminated fractured rock (primarily based on work by the petroleum industry, USGS studies of the Mirror Lake granitic system in New Hampshire, and investigations of prospective radioactive waste repositories). Furthermore, many publications concern mathematical models representing hypothetical or idealized fracture networks for contaminant behaviour in fractured rock systems (e.g., Smith and Schwartz 1984; Sudicky and McLaren 1992; Smith and Schwartz 1993; Therrien and Sudicky 1996, and many others). However, these modelling endeavours generally do not represent actual field sites or any particular type of rock, and field data of actual contaminant distributions and contaminant behavior in fractured rock, particularly sedimentary rock, are addressed in relatively few publications. Unlike behavior in igneous rock, contaminants in sedimentary rock can reside predominately in the porous rock matrix while downgradient transport occurs in the fractures (e.g., Foster 1975; Lipson et al. 2005). Therefore, determination of the contaminant distribution in sedimentary rock requires measurement of contaminant concentrations in both the fracture network and the rock matrix. Lawrence et al. (1990; 2006) and Sterling et al. (2005) provide examples of measurement of chlorinated solvent concentrations in the rock matrix (Figure 3-1). Most literature pertaining to groundwater flow and solute behaviour in fractured rock concerns igneous rock such as granite. Several countries have proposed the creation of deep repositories for radioactive waste in granitic rock, and the search for and assessment of prospective sites has involved intensive field studies. However, these studies have not involved evaluating existing contaminant plumes as such plumes do not (yet) exist in these environments because no radioactive waste has been disposed of in this type of rock. Fractured rock research has included field tracer experiments but their spatial scales are small in relation to the relevant plume scale 9

24 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 anticipated in fractured sedimentary rock environments. The literature contains no welldocumented cases of substantial industrial contaminant plumes in any type of fractured rock, except for the sites included in the academic research program that includes the SSFL. The challenge posed in the quest to delineate and understand plumes in fractured rock is much greater than that posed by granular media because the scale of variability and complexity imposed by fracture networks is much greater and costs per borehole in rock are generally much larger. Abundant literature exists concerning karst; however, the nature of karst channel networks is much different than fracture networks in rock without dissolution channelling. Core analyses in sedimentary rock provide contaminant mass and phase distributions more relevant to contaminant behavior than those obtained from monitoring wells or other types of borehole water sampling alone. The determination of the nature and extent of the contamination, with emphasis on elucidating the internal anatomy of contaminant plumes (including contaminant distribution in the rock matrix where groundwater is nearly immobile due to low permeability) is the foundation for understanding the processes governing the contaminant distribution. The rock-core based approach has several advantages over conventional methods for contaminant investigations in fractured sedimentary rock. For example, it provides a timeintegrated fingerprint of plume behaviour. In the rock matrix block, the extent of the halo evolving outward from each fracture can increase over several decades, depending on the duration of the DNAPL source. Thus, the halo extent can be used as an indicator of the age of contamination (time since contaminant arrival) on a fracture by fracture basis. In contrast to analyses pertaining to the rock matrix, which generally has low permeability, groundwater sampling in the borehole using depth-discrete multi-level groundwater monitoring systems allows the current chemical concentrations in the hydraulically active factures to be determined and permits evaluation of plume variability over time. However, drilling and related borehole cross connection effects can influence the results. The rock core analysis method avoids this problem because the low permeability matrix is not easily cross connected during drilling and core retrieval prior to sample collection (Sterling et al. 2005). In addition, the rock core contaminant analyses provide a direct measure of contaminant mass storage because the pore 10

25 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 space in the rock matrix constitutes nearly the entire contaminant mass storage volume; the exception is the potentially large contaminant mass percentage stored in the fractures if DNAPL persists. However for the rock core analyses to show the actual mass distribution with useful accuracy, the samples must be collected from the core at closely spaced intervals (Lawrence et al. 2006). Parker et al. (1994; 1997) proposed a new conceptual model for chlorinated solvent DNAPL source zones, supported using analytical models for DNAPL behavior in water-saturated fractured porous media such as clay and sedimentary rock. In this model, the immobile DNAPL film in the fracture dissolves into the contiguous water film in the fracture, establishing an aqueous concentration gradient driving mass into the porous matrix by diffusion. This mass transfer can cause complete dissolution of the DNAPL phase after some period of time depending on the thickness of the DNAPL film (i.e., fracture aperture and initial fracture DNAPL saturation) and the diffusion driven mass transfer rate into the matrix. However, this time is short relative to the time elapsed since contamination of these sites (decades ago). Building on the work of Parker et al. (1994; 1997), VanderKwaak and Sudicky (1996) developed a numerical model to show the dissolution time is dramatically shortened when active groundwater flow is present in a fracture containing the DNAPL. In this model, the source zone evolves relatively rapidly (i.e., DNAPL dissolution followed by continued changes in concentration distribution and contaminant flux from the source to the plume) and has a strong influence on plume development and internal concentration behavior. The lack of DNAPL persistence in all or major parts of the source zone represents a major difference between typical source zones in fractured porous sedimentary rock and those of granular aquifers where DNAPL as free product and / or residual can persist for an extremely long time (Pankow and Cherry 1996). The conceptual model for complete loss of the DNAPL phase from chlorinated solvent source zones has been assessed using closely spaced sampling of continuous rock core at each of the four field sites. These results support the conclusion that the DNAPL phase has completely dissolved away (Hurley and Parker 2002; Sterling et al. 2005) and all or nearly all of the mass is stored in the matrix (Goldstein et al. 2004). Complete dissolution of the DNAPL phase may not occur when the DNAPL is of low effective solubility as a complex mixture of compounds, such as at the Wisonsin site and sites with creosote, coal tars, and PCBs. 11

26 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Another major conclusion drawn from applications of the rock core VOC method in zones where DNAPL contamination had occurred is that the concentration profiles (concentration vs. depth) indicate the occurrence of numerous pathways for contaminant migration in each hole, consistent with observations of fracture occurrence in the cores. However, the existence of numerous active fractures is not consistent with results of conventional borehole fluid resistivity and temperature logging and borehole flow metering that typically indicate only two or three active fractures are present in each hole (Sterling et al. 2005; Pehme et al. 2007). Therefore, the rock core VOC results support the conceptual model for fractured sedimentary rock in which the DNAPL initially occupied many, mostly small to intermediate aperture fractures, and then dissolved away allowing the mass to be transferred by diffusion into the nearby porous matrix. Groundwater flow through the DNAPL zone in the fractured rock causes a down-gradient dissolved-phase plume to form. In this conceptual model, the plume forms in a network of many interconnected fractures of variable aperture and length without the dominance of any large-aperture fractures over long distances. The evolution of a chlorinated solvent source zone and plume in fractured sedimentary rock is illustrated in cross-section at three stages in Figure 3-2. The strong influence of diffusive mass transfer into the low permeability matrix blocks both in the source zone and plume is illustrated. In the past few years, the conceptual model for chlorinated solvent DNAPL behavior in fractured porous media outlined above has been combined with a conceptual model for the formation and evolution of contaminant plumes from a source zone, referred to here as the general conceptual model for source zones and plumes. This model, which includes DNAPL disappearance after several years or a couple of decades and plume formation in networks of many interconnected fractures within a porous medium, has been represented stylistically with simulations using 2-D discrete fracture models (e.g., FRACTRAN bysudicky and McLaren 1992). In these systems the assignment of fracture and matrix parameters are consistent with borehole measurements in the field and laboratory measurements on core samples. In studies of contaminant migration in granular media (i.e., non-indurated geologic deposits), the plug flow advance of plume fronts is estimated using the average linear groundwater velocity ( ) which is the Darcy flux divided by the effective porosity relevant to transport. This porosity is 12

27 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 commonly between (Freeze and Cherry 1979). The estimating concept is also applicable for values in fractured rock when many interconnected fractures exist (Figure 3-3). In these cases the is the Darcy flux divided by the bulk fracture porosity shown conceptually in Figure 3-4 and Figure 3-5 rather than the matrix porosity. For intact fractured rock, typical values of are 10-3 to The overall magnitude of Darcy flux variations in fractured rock are in the same general range as in granular media; and therefore, the calculated range typical of fractured rock is orders of magnitude larger than that of granular media. For example, kilometers per year. in fractured sedimentary rock is generally on the order of a kilometer to tens of Evidence from field studies indicate that in fractured rock aquifers is very large (in the calculated range indicated above); however, results of the plume investigations conducted to date at several sites including the SSFL indicate that the actual plume fronts have advanced over past decades at rates that are orders of magnitude smaller than the respective. The ratio between calculated and observed plume front migration rates is referred to as plume front retardation (Figure 3-6). Plume front retardation is a concept initially established by Foster (1975), further elucidated by Freeze and Cherry (1979), and represented in several DFN modelling papers (e.g., Grisak and Pickens 1980; Lipson et al. 2005) but which has not previously been demonstrated using field data. Therefore, research is in progress at several sites that is directed towards quantifying the processes responsible for the observed strong plume front retardation relative to groundwater advection. These studies have two thrusts: i) improved estimates of based on better measurement of Darcy flux and bulk fracture porosity; and ii) more detailed examination of plume front migration. 13

28 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 3-1 Measured Chlorinated Solvents in the Rock Matrix Example of rock core analysis results for TCE in sandstone at a location near TCE DNAPL source zone at the California site. The detection limit for all analyses are well below TCE solubility levels, indicating the lack of DNAPL presence. (Modified from Sterling et al., 2005) 14

29 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 the plume continues to grow, stabilizes or shrinks Figure 3-2 Evolution of Contaminant Source Zone and Plume in Sedimentary Rock The discrete fracture network (DFN) approach for investigating contaminated sites on fractured sedimentary rock includes intensive data acquisition from contaminated cores and from the corehole. Open hole conditions are minimized. Illustration of conceptual stages in the time evolution of source zone and plume at chlorinated solvent DNAPL sites on fractured porous sedimentary rock: a) DNAPL flows in the fracture network and begins to dissolve and diffuse into the rock matrix. DNAPL flow ceases soon after DNAPL input to the rock ceases; b) All DNAPL mass has dissolved completely and the contaminant mass now exists almost entirely in the rock matrix as dissolved and sorbed mass due to diffusion driven mass transfer. Therefore, the source zone no longer has DNAPL and no distinct difference in contaminant state persists between the zone initially referred to as the source zone and the plume; c) Groundwater flow through the initial DNAPL source zone has caused complete mass translocation from much of the initial source zone into the downgradient plume; the plume front is migrating only slowly or is stable or shrinking due to the combined effects of matrix diffusion and degradation. 15

30 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 3-3 Schematic of Average Linear Groundwater Velocity in Fracture Rock a) Tortuous travel path of a water molecule in a fracture network; b) Average linear groundwater velocity in fractured media: V f represents water velocity along straight line path from A to B (Waterloo DNAPL Short Course 2006) 16

31 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 3-4 Schematic of Fracture Porosity versus Matrix Porosity Fracture porosity is typically several orders of magnitude less than the matrix porosity of sedimentary rocks. Figure 3-5 Conceptualization of Bulk Fracture Porosity Conceptualization and estimation of bulk fracture porosity for fracture networks that are: (a) cubic and (b) tabular matrix 17

32 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 3-6 Schematic of Retardation in a Single Fracture Conceptual representation of retardation due to matrix diffusion of a dissolved contaminant transported by groundwater flow in a single planar fracture (adapted from Freeze and Cherry (1979)): a) no retardation because the rock matrix has minimal porosity (e.g. granite), b) strong retardation relative to case a) due to matrix diffusion and c) retardation further enhanced by sorption 18

33 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December USE OF MATHEMATICAL MODELS AT THE SSFL The conclusions arrived at in this SCM are based on the large and diverse data sets from the site and interpretation of these data based on the experience of the project team, comprehensive review of the scientific literature, and intensive application of mathematical models. Some of the models are simple, such as the steady state method (Theim equations) based on Darcy s Law to derive formation transmissivity values from borehole test data. Other models are numerical and much more complex, such as the finite element model (FEFLOW) used to simulate the 3-D groundwater flow system (based on the equivalent porous medium premise) and the FRACTRAN model used to simulate transport and fate in networks of ubiquitous interconnected fractures. These numerical models have been used to generalize the results of location-specific data and apply them across the SSFL domain. The modeling efforts include a particularly comprehensive use of the groundwater model FEFLOW, a widely used commercially available model. FRACTRAN is also a well established, commercially available model, but it and other numerical DFN models are not commonly calibrated with site field data sets as done in this work. Therefore the application of FRACTRAN in the development of the SSFL SCM is new and the approach is outlined below. The mechanisms by which solutes are transferred from water flowing in fractures to relatively more stagnant waters in adjacent matrix blocks, mostly via diffusion processes, are well known and understood. The effects of such mass transfer on the rate of solute transport in a single uniform fracture sandwiched between infinite matrix blocks can be calculated from analytical models. One important insight provided by this calculation is that the progression of the solute front in the fracture is slowed by the diffusion of solute from the fracture to the matrix. The degree to which the solute front is retarded compared to the rate of groundwater flow in the fracture depends on the magnitude of water flux in the fracture (which is a function of the hydraulic fracture aperture and hydraulic gradient), rock matrix porosity, the effective diffusion coefficient, solute adsorption on solid surfaces (fracture surface and/or within the matrix blocks), biotic and abiotic reactions, and radioactive decay (when applicable). 19

34 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 At SSFL, solute transport occurs within a complex, three dimensional, network of interconnected fractures and matrix blocks. The above simple model suggests the progression of the solute front will be slowed, even in this much more complex system where the magnitude of water flux within the network is expected to be highly variable owing to spatial variability of fracture length, orientation, aperture, and spacing. The model FRACTRAN provides a means of quantifying solute transport as affected by the most important of the complicating factors that occur in the real system at SSFL. While FRACTRAN calculates transport in only a twodimensional plane and within an orthogonal fracture network, the calculation is made with realistic spatial variability of fracture spacing, aperture, and length. The FRACTRAN simulated network is based on the SSFL fracture network by using the observed statistics for fracture spacing, aperture, and length observed or calculated from field data. In addition, groundwater fluxes and hydraulic gradients along particle traces from SSFL source areas obtained using the site-wide 3D equivalent porous media (EPM) groundwater flow model are used to constrain and condition the FRACTRAN simulations. However, the fracture network used in the numerical calculations in FRACTRAN is not expected to match any specific part of the actual fracture network at SSFL, other than statistically. For this reason and because FRACTRAN computes only two-dimensional transport, calculated concentrations are not expected to exactly match observed concentrations. A representative FRACTRAN model is expected to generate concentration distributions that are stylistically similar to observed distributions. FRACTRAN does not simulate actual plumes; rather, it is a tool that reliably informs and bounds our judgment concerning how far and how rapidly solute fronts are likely to move, using input parameters tailored to conditions at SSFL to the extent possible, under a wide variety of circumstances. 20

35 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December ELEMENTS OF THE SITE CONCEPTUAL MODEL FOR THE SSFL In this document, the SCM is presented in a framework with 20 elements. Except for Element No.1 which pertains to the study strategy and study methods, each element represents a specific hypothesis or question assessed by site data analysis, literature review, and modeling. The SSFL SCM is illustrated schematically in Figure 1-2, in which 19 of the 20 elements are included. A major challenge in the production of this document in a peer reviewable form was to divide the science into small enough pieces for the data, concepts, and analyses to be readily evident to each document reviewer well grounded in the science most relevant to the particular document. Except for three of the elements (14, 15, and 17) there are two or more documents (technical memoranda or paper manuscripts) focused on particular data sets or modeling tasks. These 20 elements are arranged in a logical sequence wherein information included in each element generally builds on some or all of the previous elements. All elements present data and data analysis that, in one way or another, are the foundation for the scientific basis of the last three Elements (18, 19, and 20) that are focused on the contaminant behavior and fate. These 20 SCM elements are supported by a total of 65 documents as listed in Document 0-4. The organization of the SCM in this format is intended to present pertinent scientific information in a manner that facilitates review following the normal procedures for peer review used in the scientific research community. Therefore, each document is organized in the format of a scientific manuscript. In a few cases, the document has already been published in a journal, in a few other cases the manuscript is nearly ready for submission to a journal, and in many cases the document needs additional data analysis, expansion, or editing before it attains submittal quality. An SCM overview was first presented to the California DTSC in 2007; however, in this 2007 document there were only 16 elements. Some of the statements have been modified from the 2007 version and the number of elements has expanded from 16 to

36 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The Discrete Fracture Network (DFN) Approach was required to characterize contaminant distributions and behavior in the Chatsworth Formation (Figure 5-1). The characterization of contaminant distribution and behavior in fractured sedimentary rock such as the Chatsworth Formation required a new approach for field investigations that emphasizes detailed data acquisition from both fractures and low-permeability blocks between fractures. This approach is required because diffusion-driven mass transfer in contaminated fractured rock typically causes much of the contaminant mass to reside in the low-permeability blocks of rock between fractures while nearly all groundwater flow occurs relatively quickly in the fractures. Conventional approaches for subsurface investigations of contaminated sites rely on open borehole testing and conventional monitoring wells and, as such, have been found suitable for granular aquifers but not for fractured rock. A comprehensive method known as the DFN approach, which includes a suite of different but complementary methods, has been developed and applied to the characterization of the Chatsworth Formation at SSFL. Development of the DFN approach began in 1996 and has not reached maturity. It has been applied at other sites providing data suitable for comparison to SSFL results. 22

37 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-1 The Discrete Fracture Network Approach (DFN) for Characterization of Bedrock Groundwaters 23

38 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The rock matrix porosity provided by interconnected pores is large and the bulk fracture porosity is extremely small (Figure 5-2). Hundreds of measurements of porosity on rock core samples with concurrent geophysical logging of many holes indicate that the matrix porosity ranges between 2 and 20 percent with an average of 13 percent. This matrix porosity is more than three orders of magnitude larger than the porosity provided by the fracture network, referred to as the fracture porosity (i.e. the total void space provided by interconnected fractures with hydraulic apertures that average at about 0.1 mm- 100 microns). However, the hydraulic conductivity of the rock matrix is generally small so that most all of the groundwater in the Chatsworth Formation is stored as nearly immobile water in the rock matrix. The relatively large rock matrix porosity facilitates the diffusion of dissolved contaminants. The surface area in the rock matrix is also very large and promotes sorption of contaminants onto organic matter or reaction with natural minerals in the rock. 24

39 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-2 Matrix Porosity Image The rock matrix porosity provided by interconnected pores is large. Images of matrix porosity of Chatsworth Formation sandstone using BSEM (Backscatter Electron Microscopy) on impregnated thin sections (adapted from Hurley (2003)): (a) C4 at ft bgs, (b) C6 at ft bgs, and (c) C6 at ft bgs. In the BSEM images (left), the pore space is represented by the black color and the rock grains as grey. BSEM images were subjected to a process called segmentation, using image analysis software, where the grayscale images were converted to binary images (right) allowing the pore space to be shown in black and the rock grains and cement as white. Shown in (d) are a porosity histogram and cumulative distribution based on 105 measurements on SSFL sandstone samples. 25

40 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The rock matrix composition includes abundant reactive minerals and appreciable organic matter (Figure 5-3). The presence of reactive minerals and organic matter in the sediments at SSFL enhances contaminant degradation and sorption. The strata of the Chatsworth Formation were deposited by turbidity currents in a deep water marine environment during the Cretaceous, and this particular sedimentary depositional environment results in the sandstones being relatively abundant in reactive minerals such as pyrite and biotite-mica, and also solid-phase organic matter. These reactive minerals and organic matter have a large surface area that facilitates geochemical reactions between the sediment and the contaminants that diffuse into the rock matrix. 26

41 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-3 Minerals and Organic Carbon in SSFL Rock Samples The rock matrix is composed of many different minerals and organic matter providing reactivity for various types of contaminants: (a) petrographic thin-section of Chatsworth Formation sandstone, major minerals are plagioclase feldspar, quartz, biotite, minor minerals include pyrite and calcite, cementation is carbonate (image from Pierce M.Sc. Thesis presentation); and histograms of fraction of organic carbon (foc) measured on SSFL samples (Hurley 2003) from (b) sandstone, and (c) siltstone/shale lithologic units. Jakobsen (1981) and Link (1984)provide further description of the mineralogy of the Chatsworth Formation. 27

42 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The fracture network is a systematic arrangement of bedding parallel fractures and steeply-dipping joints (Figure 5-4). The observed presence of a systematic fracture network in the Chatsworth Formation is represented in the discrete fracture network numerical models used to represent contaminant transport and fate. Outcrop mapping, rock core examination, and borehole imaging (e.g. borehole TV, acoustic televiewer, FMI) indicate the presence of ubiquitous rock fractures in the Chatsworth Formation. These fractures have two common orientations: generally parallel to the bedding planes (dips from 25 o to 30 o to the northwest with a strike of N70 o E) or steeply-dipping joints, only some of which cut across bedding planes. These intersecting bedding plane fractures and joints form a generally systematic fracture network, which is a common feature of many types of sedimentary rock. 28

43 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-4 Conceptual Representation of Lithologic Beds, Bedding Planes, and Joints Conceptual representation of the lithologic beds, bedding planes and joints in the Chatsworth Formation (provided by Dr. Ross Wagner). The fracture network is a systematic arrangement of bedding parallel fractures and steeply dipping joints. Two perspectives are shown: a) parallel to dip and b) parallel to strike. The bedding planes represent the surfaces demarking the bottoms and tops of the depositional units. These geologic fractures are not necessarily hydrologic fractures. Geologic fractures may be closed to allow no flow or may be open to allow flow (i.e. hydrologic fractures). 29

44 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The fracture networks have strong hydraulic connectivity both horizontally and vertically (Figure 5-5). The presence of fracture networks showing ubiquitous fractures with strong hydraulic connectivity has an important role in understanding contaminant behavior at the SSFL. Many of the site bedding parallel fractures and joints are hydraulically active (i.e. have groundwater flow) and they create a high degree of fracture interconnectivity along and essentially orthogonal to the bedding planes. Each fracture likely has small to moderate length (0.5 to 20 m), and existing hydrologic data suggest many fractures connect to other fractures creating a continuum of groundwater flow in the fracture network. Within the fracture network, the smaller fractures restrict the bulk hydraulic conductivity, preventing dominance of the larger fractures. Strong interconnectivity of fractures is indicated by several lines of evidence, including response of the groundwater system to pumping tests and long-term groundwater withdrawals, as well as distributions of site contaminants and environmental isotopes, including atmospheric tritium. 30

45 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-5 Randomly Generated Fracture Network Using FRACTRAN The fracture network is interconnected hydraulically. Example of (a) a randomly generated network using FRACTRAN consisting of sets of bedding parallel fractures and joints that have high degree of interconnectivity; the fractures have (b) variable apertures on a log-normal distribution, and (c) variable length. 31

46 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The bulk hydraulic conductivity (K b ) is low to moderate (Figure 5-6). Several lines of evidence indicate the bulk hydraulic conductivity (K b ) of the Chatsworth Formation is low to moderate. The evidence includes the shallow elevated position of the bedrock water table above the valley floor and the results of hydraulic (pumping and slug) tests. These field observed values are also supported by the calibrated 3-D site flow modeling effort. Although the K b is low to moderate, it is generally one to two orders of magnitude larger than the rock matrix hydraulic conductivity (K m ). This is because the K b incorporates the influence of the hydraulic conductivity provided by interconnected fractures (K f ). The shale strata are generally expected to have lower K b values than sandstone because the shale is finer grained and more ductile. No evidence indicates large scale extreme anisotropy of K b, and the relationship of K b versus depth shows much variability in the depth range where nearly all field data occur (< 1000 ft bgs). 32

47 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-6 Diagrams of Water Table Shape The bulk hydraulic conductivity is low to moderate: At many locations across the SSFL, groundwater is encountered at shallow depths, generally feet below ground surface. This observation was used in 1997 to obtain a general estimate of the bulk hydraulic conductivity of the geologic formation using the formula shown above. Two different shapes were considered a) the shape of a ground water mound that forms a section of a sphere and b) a shape that forms a long ridge of constant cross section. Many other measurements of hydraulic conductivity have been made using multiple methods at various spatial scales and the general values of hydraulic conductivity are similar and include local areas of both higher and lower values (from Montgomery Watson, 2000). 33

48 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December There is low to moderate large-scale transmissivity across and along the faults and shear zones (Figure 5-7). Fracture sets in sandstone and finer grained units and the large-scale structural features (faults and the shear zone) and together comprise the overall fracture network in the Chatsworth Formation. Many structural features (faults, deformation bands) exist, some of which exert a strong influence on groundwater flow. The bulk large-scale transmissivity of these structural features is generally small to moderate. Although faults and the shear zone may have local segments of higher bulk transmissivity, no data indicate any provide long-distance continuity of relatively large hydraulic transmissivity zones that would promote rapid contaminant migration. Figure 5-7 Plan View Conceptualization of Falut/Shear Zone Features The faults and shear zones have low to moderate large-scale transmissivity generally not larger than the Chatsworth Formation sandstone: Faults and shear zones are conceptualized as having two different zones (gouge zone and damaged zone) relative to the general fractured rock mass (protolith). The gouge zone is considered to have an appreciably lower hydraulic conductivity than the protolith and is generally thin (inches to a foot or so). Alternatively, the damaged zone is considered to have a moderately higher hydraulic conductivity than the protolith and may extend a distance of a foot or more from the gouge zone to the protolith. Since the thickness (b) at both the gouge zone and damaged zones are small relative to the protolith, these features are not very transmissive over their full length (adapted from (Forster et al. 1994). 34

49 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The effective fracture apertures that dominate groundwater flow and contaminant transport are small to moderate (Figure 5-8). Although many fracture apertures observed on rock outcrops and borehole images appear large (> 1 mm), hydraulic tests show the effective hydraulic apertures calculated using the Cubic Law are much smaller, generally between 50 and 250 microns (roughly equivalent to the diameter range of human hair). These effective hydraulic apertures, which govern groundwater flow, are much smaller than the outcrop and borehole imaging apertures because flow in each aperture is restricted due to the constrictions along its plane and intersections with other fractures. Overall, these small to moderate apertures, rather than the larger apertures, govern bulk groundwater fluxes and contaminant plume behavior as indicated by sensitivity analysis of contaminant and environmental isotopes distributions using DFN numerical models. 35

50 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-8 Fracture Aperture Concepts and Definitions Effective fracture apertures are small to moderate. This figure depicts the various concepts and definitions of hydraulic fractures (Waterloo DNAPL Short Course 2006) In a) fractures are conceived as a set of smooth parallel plates as an idealized case. In b) a fracture surface is conceived as having rough walls where the opening between fracture surfaces varies in thickness, resulting in a hydraulic aperture. In c) the effective aperture is defined as controlling fluid flow at the smallest distance between the fracture walls (i.e. throat). In a fracture network d), the flow is controlled by the smaller fractures; in this example a flow constriction occurs in the middle of the network that connects the left and right portions of the network. 36

51 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The groundwater in the overburden and highly fractured zone at the top of the bedrock forms a subsurface hydrologic continuum with deeper groundwater in the Chatsworth Formation (Figure 5-9). The depth-to-groundwater beneath the SSFL is shallow, generally ft bgs, except where large scale pumping from site water supply wells has depressed the water table to depths of ft bgs. The shallow groundwater zone in the overburden (alluvium and colluvium) and rock is replenished by infiltration (recharge) from rain, and this water eventually passes through the shallow groundwater zone to replenish the deeper bedrock. Therefore, the entire subsurface water regime is a hydrologic continuum. In parts of the SSFL a shallow groundwater zone is perched above the regional site groundwater system in the Chatsworth Formation. Therefore, in these areas water readily flows downward from the surface to the deeper regional site unconfined groundwater system. Evidence for this active hydraulic continuity between the two zones is provided by hydraulic head profiles and distributions of atmospheric tritium and contaminants. This flow continuity is important because it results in contaminants from the shallow zone ultimately ending up in the deeper zone. 37

52 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-9 Schematic of the Subsurface Hydrologic Continuum The shallow groundwater zone in the overburden (alluvium and colluvium) and rock is replenished by infiltration (recharge) from rain, and this water eventually passes through the shallow groundwater zone to replenish the deeper bedrock. Therefore, the entire subsurface water regime is a hydrologic continuum. (Figure and caption provided by Dave McWhorter) 38

53 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December Groundwater recharge is a small percent of the mean annual precipitation (Figure 5-10). The mean annual rainfall at the SSFL is 18 inches and using a variety of methodologies to estimate recharge, only a small percentage (2-7%) of this water infiltrates all of the way to the water table resulting in recharge to the groundwater flow system in the Chatsworth Formation. This recharge is the equivalent to 160 gallons per minute (gpm) when totaled over the entire land area of SSFL (2,850 acres). When this 160 gpm is distributed areas representing the contaminant plumes, each plume has only a few gallons per minute or less of recharge available to transport contaminants from the source areas toward the discharge areas. This relatively small total groundwater flux (i.e. discharge from the SSFL per plume) limits the magnitude spatial extent of the plumes. 39

54 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-10 Conceptual Representation of Groundwater Recharge Conceptual representation of groundwater recharge. Recharge is the water that infiltrates to the water table. Chloride has been used to estimate recharge by applying a mass-balance approach (Eriksson and Khunakasem 1969) and has been found to average about 6% of the average annual precipitation or 1.1 inches per year. 40

55 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The geologic and hydrogeologic variability in the Chatsworth Formation are appropriately represented in the 3-D groundwater flow model (FEFLOW) used to search for zones that could allow exceptionally large or rapid contaminant migration to offsite areas of concern (Figure 5-11). A mountain scale model for 3-D groundwater has served as an essential tool for assessment of the 3-D groundwater flow system. This model uses the FEFLOW finite element code wherein the site geological information from comprehensive surface mapping, and borehole geologic and geophysical records were incorporated along with field and literature derived hydrogeologic properties. The model uses an equivalent porous medium representation of geologic conditions. This approach was supported by many lines of evidence that general hydrogeologic conditions could be represented by using bulk hydraulic conductivities and other identified hydrogeologic properties. This model used lateral exterior boundary conditions, and a groundwater basal boundary derived from a preliminary regional large scale modeling effort. The groundwater model formulation represents individual sandstone units, mudstone/shale units, faults, and a shear zone. It allows for spatial variability in the hydrogeologic properties of and within these units. The model provides a tool to integrate multiple data sets to represent observed and conceptualized site conditions as well as to simulate groundwater flow through the subsurface using physical equations. The model parameters have been optimized by applying state-of-the-art tools, and the model is being utilized to evaluate the implications of alternative conceptual models on groundwater flow paths. 41

56 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-11 Geologic and Hydrogeologic Variability as Represented in the FEFLOW Model The figure above shows a FEFLOW finite element mesh designed to accommodate geologic and hydrogeologic complexities: topography, faults, bedding, and irregular geological units. 42

57 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The groundwater contains many dissolved inorganic constituents of natural origin that provide insight concerning groundwater flow and contaminant behavior (Figure 5-12). Groundwater samples from monitoring wells and multilevel systems were analyzed for many different types of inorganic constituents including: major ions (magnesium, sodium, calcium, chloride, sulphate, bicarbonate), minor constituents (strontium, lithium, barium, bromide, boron), redox indicators (nitrate, manganese, iron), dissolved gasses (methane, ammonium, carbon dioxide, hydrogen), stable isotopes ( 18 / 16 O, 2 H/ 1 H, 13 C/ 12 C, 37 Cl/ 35 Cl, 87 Sr/ 86 Sr, 34 S/ 32 S), and radioactive isotopes ( 14 C and 3 H), to discern patterns and trends indicative of flow system characteristics and hydrogeochemical processes relevant to contaminant behavior including chlorinated solvent degradation. The presence and distribution of observable constituents was used with other physical and contaminant distribution data sets to further test the SCM. Groundwater chloride concentration magnitudes and distributions support the low recharge rates presented in the SCM. The great depth to which groundwater with very low total major ion concentrations extends beneath this mountain suggests that groundwater flow involving water recharged on the mountain has been an active flushing fluid to great depth over a long period of time (Chatsworth Formation uplifted <6 million years ago (MYA)). Rock core porewater samples collected from a borehole terminated at 1399 ft bgs (317.1 ft amsl) showed no increase in chloride or other major ions with increasing depth. All soluble marine chloride and sulphate salts have been flushed out by groundwater flow during the several million years since this terrain was uplifted to its present state. 43

58 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-12 Chloride, Sulphate, and Bicarbonate Versus Depth There are no trends with depth in chloride, sulphate, or bicarbonate concentrations. The concentrations of these major anions measured in the near-surface groundwater are similar to those measured in the Chatsworth Formation groundwater. The conceptual model for groundwater recharge at the site is slow infiltration through the vadose zone. The hydrochemistry results presented in the previous figures and this figure indicate that the hydrochemical processes contributing to the composition of the Chatsworth Formation groundwater are occurring within the saturated matrix blocks in the vadose zone prior to groundwater reaching the water table. 44

59 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December A substantial portion of the groundwater originating in the SSFL property discharges at seeps, springs, and phreatophytes situated on the mountain slopes offsite from SSFL that show no detectable concentration of the SSFL contaminantsof-concern (Figure 5-13). Intensive searches for seeps, springs, and phreatophyte clusters downslope from SSFL have identified more than 150 of these groundwater discharge features. Hydrogeochemical studies and groundwater flow modeling indicate many of these features derive some or all of their water from SSFL. Therefore, groundwater flow paths for many of the SSFL contaminant plumes are expected to lead toward these groundwater discharge features. However, no contaminants have been found at any off-site seeps/springs. This is consistent with the SCM that requires strong plume front retardation and plume attenuation. The lack of SSFL contaminants found at offsite seeps/springs is consistent with the SCM for contaminant transport and fate. 45

60 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-13 Springs, Seeps, and Phreatophytes at and Around the SSFL Much of the groundwater originating in the SSFL property discharges at springs and phreatophytes situated on the mountain slopes: This figure shows (a) the locations within and around the perimeter of the SSFL where springs and phreatophytic vegetation have been identified based on field reconnaissance, and (b) groundwater flow represented by a two-dimensional flow model for an idealized case with flow converging at a spring. 46

61 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The chlorinated solvent contamination was initially caused by DNAPL penetration below the water table, but the DNAPL has since been converted to dissolved and sorbed mass now residing in the rock matrix, and therefore DNAPL flow no longer occurs (Figure 5-14). The depths directly beneath DNAPL release locations to which TCE and its degradation products have been found in the bedrock groundwater are consistant with the SCM. TCE at these locations penetrated to appreciable depths decades ago. DNAPL dissolution enhanced by groundwater advection and diffusion-driven mass transfer of dissolved mass from the groundwater in the fractures into the rock matrix water results in all or nearly all of the DNAPL to be converted to the dissolved and sorbed mass now found in the lower permeability rock matrix. These conditions would result t DNAPL flow to be arrested. Therefore all current plume related TCE is a result of groundwater flow and not downgradient migration of DNAPL. 47

62 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-14 Conceptual Diagram of TCE DNAPL Source Zone Evolution in Fractured Porous Rock Conceptual representation of DNAPL mass conversion to dissolved and sorbed mass in the rock matrix due to dissolution and matrix diffusion: a) at earliest time the DNAPL occurs as a continuous film in the fracture and with time the film disappears (Parker et al. 1994), b) as the DNAPL mass declines, the DNAPL film becomes discontinuous and evolves to immobile droplets or beads before complete disappearance (Waterloo DNAPL Short Course 2006). Once the DNAPL is disconnected it becomes immobile, held in place due to surface tension with the surrounding water in the fracture. The main period of TCE DNAPL release at SSFL was in the s and therefore there has been ample time for TCE to achieve immobility from nearly all fractures, as indicated by rock core VOC analyses. 48

63 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December Large TCE mass as dissolved and sorbed phases occurs at and in the local vicinity of TCE DNAPL input locations (Figure 5-15). Analyses of thousands of samples collected from rock core obtained at known or suspected historic TCE DNAPL input locations show abundant dissolved and sorbed mass of TCE and its degradation products in the rock matrix blocks between fractures. Summation estimates of this mass indicates much or all of the total DNAPL volume estimated to have entered the subsurface at SSFL can be reasonably accounted for by the contaminant mass found in the rock matrix at and near the DNAPL input locations. 49

64 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-15 TCE Source Mass Distribution Large TCE mass as dissolved and sorbed phases occurs in the vicinity of TCE DNAPL input locations: Example of source zone mass distribution at SSFL based on rock core VOC analyses: (a) estimated pore water TCE concentrations and (b) cumulative TCE mass at Corehole C6, located in the Delta test stand source area (modified from Hurley 2003). Using the estimated TCE subsurface input of 46,000 gal (CH2M Hill, 1993), the total footprint area would be about 110,000 m 2 (equivalent to a zone with a radius of 187 m) applying the total estimated C6 mass. 50

65 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December The behavior of all soluble chemicals and radionuclides is strongly influenced by diffusion into and out of the porous rock matrix (Figure 5-16). Not only do rock core contaminant analyses show substantial VOC contaminant mass in the rock matrix blocks between fractures, but rock core analyses of perchlorate and tritium also show these contaminants present in the pore water in the rock matrix. The porosity of the rock in all parts of the SSFL is substantial. The presence contaminants in the rock matrix along with measurements of the diffusion properties of the rock and the use of numerical models indicate molecular diffusion is occurring and has a strong influence on the transport and fate of all contaminants in the Chatsworth Formation. Figure 5-16 Conceptualization of Transport Distances in a Single Fracture for TCE and Various Radionuclides Relative to Water Conceptualization of contaminant transport distances in a single fracture for immiscible and soluble phases of TCE and various radionuclides relative to water. TCE would be transported the furthest due to its flow as a DNAPL through the fracture. Other solutes are transported shorter distances because of their individual characteristics associated with their physical properties (e.g. solubility, portioning and biological or radioactive decay). 51

66 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December Contaminant mass resides in the vadose zone but vadose zone mass now has minimal or no influence on plume fronts (Figure 5-17). The contaminant mass remaining in the vadose zone no longer contributes to plume front advances; formation of the contaminant plumes in the Chatsworth Formation was initiated decades ago due to contaminant migration through the unsaturated zone. The current contributions from the vadose zone are now too small to be influential. The contaminant mass that governs plume front behavior today is the accumulated mass now present in the groundwater zone. The current positions of the plume fronts, after decades of groundwater transport, are no longer appreciably influenced by contaminant mass remaining in the vadose zone because the annual contribution of mass from this zone is small relative to the mass already residing in the groundwater plumes. Field observations and mathematical modeling indicate the current plume fronts are stationary or nearly stationary due to the combined effects of matrix diffusion, sorption, transverse dispersion, and, in some cases, the contribution of degradation or decay. 52

67 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-17 Conceptual Model for Contaminant Transport from the Vadose Zone Liquid-phase TCE present in the matrix will be in contact with both water and air. The DNAPL will dissolve into both the aqueous and gaseous phases in accordance with the aqueous solubility and the vapor pressure, respectively. TCE vapor will diffuse through the largely stagnant gaseous phase, and simultaneously partition into the aqueous phase in accordance with Henry s law. The primary transport mechanism in the aqueous phase will be advection by the downward flowing recharge waters. Results of calculations indicate mass transfers to the groundwater from the vadose zone are on the order of tenths to a few kilograms per year (figure and caption provided by Dave McWhorter). 53

68 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December Contaminant plumes can be characterized and monitored (Figure 5-18). Three major lines of evidence indicate the plumes are orderly and monitorable: (i) many bedrock monitoring wells situated down the flow path from TCE source areas (i.e., areas where DNAPL entered the rock long ago) have detectable TCE; (ii) at nearly all locations, rock core VOC analyses at or near TCE source zones and at locations down the flow path from source zones show vertically- and laterally-distributed TCE, indicating transport in many fractures; and (iii) FRACTRAN simulations of plumes using boundary conditions and fracture and matrix parameters consistent with SSFL-specific data produce orderly and monitorable plumes. The monitorability of plumes in the Chatsworth Formation is also supported by delineation of a site-derived tritium plume. 54

69 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-18 Example of Plumes Generated Using FRACTRAN Contaminant plumes are orderly and monitorable. Examples of plumes generated using FRACTRAN for (a) wellinterconnected fracture network with no major preferential pathways (mean aperture=100 microns, range of bedding parallel fracture lengths= m); and (b) sparser fracture network with a similar overall bulk hydraulic conductivity but with larger fracture apertures (mean=250 microns) and longer fractures (range of bedding parallel fracture lengths=100 to 350 m). Plumes (relative concentration) from a constant source along the left boundary (z=200 to 290 m) for TCE in a sandstone matrix ( = 12%, R=2) are shown at (c) 20 years, (e) 50 years and (g) 100 years for the well-interconnected network, producing plumes that are orderly and monitorable; and at (d) 20 years, (f) 50 years and (h) 100 years for the sparser fracture network, producing plumes that would be much more difficult to monitor. 55

70 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December Matrix diffusion causes plume fronts to be strongly retarded relative to the mean groundwater velocity in the fracture network and along with dispersion, causes maximum plume concentrations to continually decline (be attenuated) (Figure 5-19). Based on the combined effects of diffusion-driven contaminant mass transfer from the fractures into the rock matrix, contaminant sorption and degradation, and hydrodynamic dispersion, the DFN modeling predicts: (i) strong plume front retardation increasing over time since plume formation was initiated decades ago ii) plume fronts are now stationary or nearly stationary and (iii) maximum plume concentrations have declined markedly over time and this decline will continue. The modeled plume front retardation and the stationary nature of the plume are supported by several types of field observations, including long-term trends of TCE concentrations versus time in many monitoring wells; the lack of detection of TCE, its degradation products, or any other contaminants in off-site seeps or springs; and the occurrence of TCE degradation indicated in studies in the field and in laboratory experiments. The strong plume retardation processes operating at SSFL are also supported by a relatively small total contaminant mass flux in each plume. 56

71 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-19 FRACTRAN Simulations of Contaminant Transport in a Single Fracture The contaminant plumes are strongly retarded relative to the rapid mean groundwater velocity in the fracture networks. Plots show results of single fracture simulations using FRACTRAN for scenarios with a 100 micron fracture and 1% hydraulic gradient: relative concentrations in the fracture and matrix at 50 years (left-side) and fracture concentration profiles at 10, 20, 50 and 100 years from a constant source (right side) for: (a) non-porous matrix, allowing transport at the same rate as groundwater velocity in the fracture (2580 m/yr) neglecting dispersion; (b) porous matrix ( =12%, τ=0.10) with diffusion; (c) porous matrix with diffusion and sorption (R=3); and porous matrix with diffusion, sorption and slow contaminant degradation (t 1/2 =10 yr). 57

72 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December Degradation is contributing to some degree in the attenuation of all contaminant plumes in addition to the attenuation caused by matrix diffusion, sorption, and dispersion (Figure 5-20). Strong natural attenuation is occurring in all contaminant plumes regardless of the chemical species (or radionuclide) due to the influence of matrix diffusion, sorption, and dispersion. For some contaminants, such as TCE, field and laboratory evidence indicates the attenuation is intensified by degradation. Radionuclides are attenuated by radioactive decay. A number of attenuating processes act in concert causing plumes to achieve stability and eventually, over a longer time period, to shrink and eventually disappear. 58

73 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 5-20 FRACTRAN Simulations of Attenuation Processes on TCE Plumes All of the contaminant plumes are being naturally attenuated due to a combination of processes. Examples of TCE plumes generated using FRACTRAN at 20, 50 and 100 years for a scenario with a well-interconnected fracture network in a sandstone matrix ( =12%,R=2). Plumes were generated from a finite-life (10 year) source representing the period of DNAPL presence, for cases with (a) no contaminant degradation, and (b) including slow first-order degradation with a half-life of 10 years. Both cases show strong attenuation effects. The average linear groundwater velocity in the fracture network exceeds 500m/year. In contrast to the rapid rates of groundwater flow, contaminant transport is much slower due to diffusion of contaminant mass into the porous rock matrix and sorption. For (a) with no degradation, the plume front just reaches the model boundary (300 m from the source) at 50 years and also shows strong reduction in internal plume concentrations; while in (b) with degradation, the plume completely degraded to below C/C 0 of 10-5 by 100 years in the simulated case. It should be noted that the simulations depicting degradation do not include production and transport of daughter products (e.g. cis-dce from reductive dechlorination of TCE). 59

74 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December MAIN CONCLUSIONS OF THE SSFL SCM Groundwater flow in the Chatsworth Formation is primarily through fractures. The void space provided by the fractures is very small compared to the interstitial void space in the porous rock matrix. Fracture porosity cannot be directly measured, but can be estimated using the Cubic Law for idealized fracture networks. Monitoring since the early to mid-1990 s at many locations on the SSFL property demonstrates persistent contamination by chlorinated solvents, primarily TCE, as depicted in Figure 1-3. Historically, TCE was used in the solvent phase (i.e., oily liquid), and field data indicate releases to the ground surface resulted in its movement as a separate liquid phase (i.e., DNAPL) into the subsurface where it entered the interconnected fracture network of the Chatsworth Formation. The DNAPL readily moved through the vadose zone and then into the groundwater zone. The zone in which the DNAPL achieved its distribution in an immobile state is referred to as the (subsurface) source zone. Today practically no TCE exists in the DNAPL phase in source zones or in downgradient plumes. The DNAPL initially present in the fractures of the Chatsworth Formation subsequently dissolved into the contiguous water also present in the fractures. Groundwater flowing through the fractures in the source zone transported dissolved TCE mass down-gradient within the fracture system spreading out the mass as the fracture network branched (transverse dispersion). The dissolved TCE in the fracture network also diffuses into the adjacent pore space in the matrix blocks between the fractures. The resulting contaminated groundwater zone comprised of dissolved and sorbed contaminant mass makes up the contaminant plume. Many measurements of chemicals in samples of rock and from wells indicate the plumes at the SSFL are generally stationary (see Figure 3-2) A second consequence of the matrix diffusion process is a remarkable slowing of the rate of dissolved TCE transport within the fractures. The TCE transport in the fractures is reduced by orders of magnitude relative to the computed rates of fracture groundwater flow. These factors 60

75 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 cause TCE mass to remain proximal to the locations where the DNAPL entered the subsurface. As a key factor in the conceptual model, we emphasize the strong effect that matrix diffusion has on the lack of DNAPL persistence in source zones and on retardation of plume front migration relative to the rate of groundwater flow in the fractures. The matrix diffusion retardation and attenuation model for solute transport in fractured sedimentary rock is sound science and consists only of processes that are well known, published in peer review journals, and understood within the scientific community. There is no question that matrix diffusion as described in the SSFL SCM occurs at this site. However, that is not to say that the degree to which matrix diffusion influences contaminant behavior and distribution is the same at all locations where contaminants entered the Chatsworth Formation. Certainly, site specific conditions determine the degree and magnitude contaminant attenuation. Although much of the field data that currently exists for SSFL was acquired during the course of TCE investigations, the transport and fate of the other SSFL contaminants of concern (COCs) (e.g., perchlorate, 1,4-dioxane, NDMA, metals, and radionuclides including tritium, strontium- 90, cobalt-60 cesium-127) are also strongly influenced by matrix diffusion. In addition, strontium and cesium and many other inorganic constituents are strongly influenced by sorption. The properties of the rock matrix that most strongly influence diffusion are matrix porosity and the ratio of pore-scale diffusion distance to macroscopic distance (tortuosity). These properties apply in a similar manner to all contaminants in aqueous form. Although matrix diffusion strongly influences the behavior of all COCs, the differences in contaminant input conditions (e.g., DNAPL versus aqueous-phases, volumes and concentrations, histories, and differences in chemical properties) necessitate that the transport and fate of each contaminant species be subjected to its own assessment. At each location where a COC entered the Chatsworth Formation, the existence of a plume is a reasonable expectation with migration of the plume in the direction of the predominant groundwater flow paths in the interconnected fracture network. Many of the plumes likely occur only at relatively shallow depths, particularly those formed from non-dnapl type inputs. Others that formed due to DNAPL inputs can be shallow, intermediate, or deep depending on the local conditions at the input locations. The lateral extent to which plumes have migrated also varies depending on local conditions. 61

76 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Two viewpoints are considered with respect to the migration of dissolved contaminants away from source zones: (1) migration zones that can be monitored and (2) chaotic migration zones that are difficult or impossible to locate and monitor. In viewpoint (1) contaminants migrate in a large number of interconnected fractures, with larger fractures connected to smaller fractures in a manner such that the larger fractures are not dominant over long distances. These migration zones have sufficient lateral spreading and internal mixing to be monitorable plumes. In contrast, viewpoint (2) consists of chaotic contaminant migration occurring in a small number of connected, larger fractures/pathways extending over large distances creating migration zones that are extremely difficult or impossible to locate and monitor. The various types of field data acquired to date from SSFL (e.g. rock core VOC analyses, COCs in monitoring wells, borehole hydraulic tests, pumping tests, geophysical logs, natural water chemistry) support or are consistent with the concept of monitorable plumes. The bulk hydraulic conductivity of the Chatsworth Formation beneath SSFL is not large, which causes the water table beneath SSFL to be positioned at elevations much higher than the adjacent lowlands of Simi and Chatsworth. The low rate of groundwater recharge into the Chatsworth Formation on the mountain top and the moderate bulk hydraulic conductivity of the mountain result in small volumetric rates of groundwater flow in the Chatsworth Formation. The occurrence of numerous shale and mudstone strata in the sandstone sequence contributes to the low to moderate bulk hydraulic conductivity of the mountain. A major shear zone runs through SSFL and several extensive faults have been identified. These features can influence the hydraulic conductivity and directions of groundwater flow at the local scale, but show no evidence of providing high transmissivity zones that drain the mountain or provide extensive rapid contaminant migration pathways either across or along the planes. It is well established in the scientific literature that the bulk hydraulic conductivity of fractured rock decreases with depth, however there is typically much spatial variability in the upper few thousands of feet such that the overall decreasing trend can be described in the noise. This appears to be the case at the SSFL however, more analysis of existing data is needed. The SCM for contaminant transport and fate applies to all parts of the SSFL property and adjacent areas even though substantial variability in the hydrologic nature of the geologic units 62

77 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 occurs across the site. The three areas used for detailed investigations (former sodium disposal facility [FSDF], Site-Derived [SD] Tritium plume, and northeast TCE plume) of hydrogeologic properties and contaminant migration and fate were selected to span the range in geologic variability. The NE TCE plume is located in the northeastern area where the bulk K is relatively large and the fractures are closely spaced. The tritium plume is in Sandstone 2 and the NE TCE plume is in Sandstone 1. These two sandstone units have similar lithology and turbidite origin but may have different structural influences on fracture occurrence. The tritium is much smaller than the NE TCE plume (Figure 1-3) likely because of the differences in hydrogeology such as the combined effects of low bulk K and small fracture apertures. A challenging aspect of the assessment of the transport and fate of subsurface contaminants at SSFL is the determination of the present-day frontal positions of the plumes as well as assessment of attenuation and ultimate fate of the contaminants. Such positions can be found, through current investigation methods, as has been demonstrated in the northeast area of the SSFL. In this context, the on-site groundwater recharge and the offsite discharge aspects of the groundwater budget are important. The best available estimate of the average annual groundwater recharge over the former operational areas of the SSFL (1,525 acres) is 90 gallons per minute (gpm). This indicates that the groundwater available for transporting contaminants from each source zone into each plume at the SSFL is very small, on the order of a few gallons per minute per plume. In the initial version of the SCM presented in the April 2000 Technical Memorandum, the potential for natural degradation to contribute to TCE attenuation was not considered. However, recent on-going, site-specific field and laboratory studies of the degradation of TCE and related volatile organic contaminants suggest natural degradation may contribute considerably to attenuation of TCE plumes (Figure 6-1 and Figure 6-2). Even small rates of degradation can be very influential on contaminant transport and fate under conditions where matrix diffusion plays a significant role such as at the SSFL. The main conclusion of this site conceptual model is that although contaminants occur at many locations in the subsurface at the SSFL and although active groundwater flow occurs in the 63

78 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 bedrock fractures, essentially all of the contaminant mass resides in the low permeability rock matrix near where the contaminants entered the subsurface decades ago. However, there is an area (northwest) where the front of a contaminant plume has migrated beyond the property boundary of the SSFL. This plume has not reached potential receptors and remains beneath the mountain slopes adjacent to the SSFL. The SSFL contaminants have not been found in any of the numerous seeps and springs that occur on the mountain slopes or in any of the few wells that exist, unused, beneath the mountain slopes. Over the past 3-6 decades, natural processes have caused strong attenuation of the maximum plume concentrations and strong retardation of plume front migration, and these effects will continue in the future. Therefore, no impacts on any receptors offsite have been observed and predictions founded on the SCM indicate no impacts will occur at anytime in the future due to contaminant retardation and attenuation caused by the combined effects of matrix diffusion, dispersion, sorption, and degradation. Not only is it predicted that SSFL plume will not reach offsite receptors, the site data suggest it is reasonable to expect that some or all SSFL plumes are stationary, or nearly so and that some or all plumes may shrink in the future decades. It is proposed on scientific grounds that long term monitoring be conducted to confirm these predictions over future decades. 64

79 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 6-1 Summary of Degradation Pathways Degradation processes have caused complete dechlorination of part of the TCE mass and degradation influences other contaminants. This figure summarizes the current conceptual model for degradation of TCE and 1, 1, 1-TCA in groundwater at the SSFL, developed based on results of a laboratory microcosm study conducted at Clemson University (Darlington 2008) and a field study conducted by the University of Waterloo (Pierce 2005). In this conceptual model, TCE transforms biotically to cis-dce, followed by minor biotic cis-dce transformation to vinyl chloride. There is a possibility that there is a small amount of complete biotic dechlorination of TCE and/or cis- DCE, however the major pathways for complete dechlorination of TCE, cis-dce and vinyl chloride are abiotic. The production of NSR (unidentified non-chlorinated compounds) in some of the microcosms and the finding of acetylene in many monitoring well samples is evidence for abiotic degradation of TCE and/or cis-dce. Abiotic degradation is likely enhanced by the common presence of biotite and pyrite in the Chatsworth formation strata. 65

80 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Figure 6-2 Recent TCE Concentrations (2009 sampling round) compared to Historical Maximum TCE Concentrations The recent results are lower in all wells, except two, and in some cases, recent results are an order-of-magnitude below the historical high. Additionally, TCE concentrations in some wells with higher maximum concentrations are below MDL for the recent sampling event. 66

81 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December IMPLICATIONS FOR MONITORING Prior to 1997, all efforts directed at groundwater monitoring in the Chatsworth Formation involved the use of conventional monitoring wells for delineating contaminant distributions. These wells have been used for long-term monitoring and many have been sampled quarterly since the mid-1990s. After 1997, a second approach, in addition to monitoring wells, was used to delineate contaminant distributions. This second approach involves measurement of contaminant concentrations in rock core. The experience gained at SSFL since 1997 indicates contaminant delineation and long-term monitoring are two distinctly different endeavors requiring very different approaches. Long-term monitoring has the primary goal of confirming plume behavior and detecting contaminant arrival at locations relevant to potential receptors. Design of an effective long-term monitoring system must be based on the results of effective site characterization. The results from the existing monitoring wells sampled over many years, the existence of a large contaminant mass diffused into the rock matrix, and the behavior of contaminant plumes indicated by the DFN modeling have important implications concerning long-term groundwater monitoring. The first implication is that because changes in contaminant distribution occur only slowly, monitoring wells need not be sampled frequently and some wells should be sampled much less frequently than others. For example, wells showing substantial concentrations at or near contaminant input areas (i.e., source areas) should be sampled very infrequently (e.g., at 5 year intervals) and those near actual or suspected plume fronts should be sampled more frequently (e.g. annually or twice-annually). The record of concentration versus time trends for each monitoring well should be assessed for selection of future monitoring frequency. 67

82 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December IMPLICATIONS FOR REMEDIATION In the context of source zone assessment and remediation, the National Research Council (2005) created five categories of hydrogeologic settings to draw attention to the particular difficulties and challenges imposed by site hydrogeology. Setting V in the NRC categorization includes those sites where DNAPL entered fractured sedimentary rock. This setting includes systems where fractures (secondary permeability) are the primary transmissive feature and there is large void space in the matrix... The challenges of managing contamination in this hydrogeolgoic setting include describing the extent of the source zone, characterizing the fracture network, delivering remedial solutions to the targeted areas in some cases, and understanding the potential for reverse diffusion to sustain contaminant concentrations in the transmissive fractures after depletion of DNAPL." (National Research Council 2005) DNAPL depletion may occur due to application of DNAPL removal, in situ destruction technologies, or, as is the case in the Chasworth Formation at the SSFL, by matrix diffusion where the original DNAPL mass stored in the fractures is transferred into dissolved and sorbed phase mass in the rock matrix (Parker et al. 1994; Parker et al. 1997). Subsurface characterization and long-term monitoring conducted at the SSFL has provided the information needed to frame the major issues concerning subsurface restoration potential for the Chatsworth Formation at this site. Two dominant features of the Chatsworth Formation are relevant with respect to remediation potential. First, it is now well established that nearly all (>99.99%) of the contaminant mass for all types of contaminants at SSFL occurs in the low permeability rock matrix blocks between fractures. Second, the bulk hydraulic conductivity (K b ) of the Chatsworth Formation sandstone is generally on the order of 10-5 m/s and in some areas smaller. In the broad context of contaminant hydrogeology, this order of magnitude for K b is more representative of aquitards than aquifers. The rock matrix hydraulic conductivity (K m ) is clearly in the typical range of aquitards hydraulic conductivities. Therefore, the assessment of the 68

83 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 restoration potential of the Chatsworth Formation is best framed in the context of aquitard restoration. To achieve restoration of groundwater in the Chatsworth Formation to the degree needed to meet regulatory limits for drinking water, essentially all (>99.99%) of the contaminant mass in the rock matrix must be removed either by extraction or in situ destruction. Without this high percentage mass removal, the dissolved concentrations in the fractures where active groundwater flow occurs will remain above the regulatory limits for drinking water. This drinking-water based requirement for restoration means that advective-based remediation technologies must be excluded from the list of potential technologies because they are incapable of removing substantial mass except over geologic timescales. The most common advection-based technology is pump-and-treat (mass removal by advection), which has already been used at SSFL and showed insignificant contaminant mass removal capability. This mass removal is insignificant because the rate at which contaminant mass diffuses out of the matrix blocks into the flowing groundwater was insufficient relative to what is needed for pump and treat methods to be effective. Another advection-based technology is steam flushing, but effective flushing would occur only in the fractures and the heating of the matrix blocks cannot extract enough contaminant mass from the low-k blocks unless the heating causes massive fracturing (e.g., opening up of new fractures). Another category of remedial technology for chlorinated solvents involves injection of materials, such as zero-valent iron or bacteria, into the fractures to cause destruction of contaminant mass in the fractures. As destruction occurs in the fractures, mass diffuses out of the mature blocks so that mass destruction in the fractures continues over time. Conceptually, this approach has advantage over advection-based mass removal; however, the time to achieve restoration using this approach cannot be reduced to reasonable periods because of the limitation imposed by reverse diffusion out of the rock matrix blocks into the fractures. Moreover, there are immense technical challenges that must be overcome to achieve effective spatial distribution of the treatment materials throughout the fracture network particularly when solid particles are used for treatment. 69

84 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Another remediation technology category involves creation of enhanced mass transfer using advection-based technologies for circulation of treatment fluids through the fracture network by injection/withdrawal to increase the contaminant removal rates by increasing DNAPL dissolution/solubility (i.e., surfactant flushing). However, this approach is only applicable to DNAPL mass, and the DNAPL remaining in the subsurface at SSFL at this time is zero or, at most, extremely small relative to the dissolved and sorbed mass. Therefore, technologies in the enhanced mass transfer category are unsuitable, Another factor limiting prospects for success using the enhanced mass transfer technologies is that hydraulic capture must be imposed everywhere continually. This would be very difficult to achieve at the SSFL due to the complexities of fractured rock. Therefore, the technologies worthy of consideration for application at SSFL are those in which the treatment medium can be delivered efficiently and effectively by advection throughout the fracture network thereby causing contaminant destruction in the fractures, and most importantly, in the low-permeability rock matrix blocks between the fractures. For mass destruction to occur in the rock matrix, diffusion of the treatment chemicals from fractures into the rock matrix blocks must occur. For these diffusion based technologies, preliminary modeling suggests that under the most ideal of conditions, where the treatment solutions are delivered perfectly throughout the fractures, the restoration time scales could be reduced from centuries or millennia to down to decades in the parts of the SSFL where matrix blocks are not very large. However, regardless of the technology selection, long time scales to achieve restoration are unavoidable due to the limitations imposed by the fact that nearly all contaminant mass resides in the low permeability rock matrix blocks. However, although the prospects for restoration of the Chatsworth Formation over practical timescales are minimal, there is mounting evidence that contaminant mass is being destroyed in situ by natural degradation. This applies to chlorinated ethenes, TCA, and likely perchlorate. Mass loss is certainly occurring for the radionuclides due to radioactive decay. The time scales for these natural processes to accomplish full mountain groundwater restoration are undoubtedly long, but they may be not much longer than the most favorable time scales for the engineered remediation options described conceptually above. 70

85 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December CURRENT STATUS OF INVESTIGATIONS AND REPORTING CONCERNING THE SSFL SCM An immense amount of data has been acquired and analyzed from the site since the previous comprehensive reporting on the SCM was completed in The SCM, as presented in this 2009 document, is now much more strongly supported by the data and modeling results and the prospects for any alternative conceptual model to be valid are minimal. However, a major phase of field data acquisition was completed only recently at the SSFL and not all data from this phase have been fully integrated with the previous data to achieve complete interpretation. The data analysis effort is on-going and many of the reports/draft manuscripts contained in this version of the SCM report are in the progress of being updated based on more advanced data analysis or edited as the data analyses displays and modeling are being fine tuned. However the 20 SCM Elements are firmly established and will continue to be the framework for the SCM. Based on the work completed to date, the SSFL Groundwater Advisory Panel attributes a high degree of reliability to the current understanding of contaminant plume characteristics and the transport and fate of contaminants in the Chatsworth Formation. However, it is recognized that on a plume specific basis, uncertainties exist and additional field data and further modeling will be required to answer questions concerning particular plumes. Degradation may be exerting a stronger influence on plume attenuation than has so far been recognized. This introduces the possibility that some plumes may be shrinking in size or start to shrink in the future. However, this type of plume behavior will require validation/assessment over the longer term through monitoring. 71

86 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December REFERENCES Berkowitz, B Characterizing flow and transport in fractured geological media: A review. Advances in Water Resources 25, no.8-12: Darlington, R Laboratory evaluation of chlorinated ethene transformation processes in fractured sandstone. Doctor of Philosophy in Environmental Engineering and Science, Clemson University. Eriksson, E. and V. Khunakasem Chloride concentration in groundwater, recharge rate and rate of deposition of chloride in the Israel coastal plain. Hydrology 7, no.2: Forster, C.D., J.V. Goddard, and J.P. Evans Permeability structure of a thrust fault, Open File Report Foster, S.S.D The Chalk groundwater tritium anomaly - A possible explanation. Journal of Hydrology 25, no.1-2: Freeze, R.A. and J.A. Cherry Groundwater, Englewood Cliffs, New Jersey: Prentice Hall. Goldstein, K.J., A.R. Vitolins, D. Navon, B.L. Parker, S.W. Chapman, and G.A. Anderson Characterization and pilot-scale studies for chemical oxidation remediation of fractured shale. Remediation 14, no.4: Grisak, G.E. and J.F. Pickens Solute transport through fractured media 1. The effect of matrix diffusion. Water Resources Research 16, no.4: Hurley, J.C Rock core investigation of DNAPL penetration and persistence in fractured sandstone. Master's thesis, Earth Sciences Department, University of Waterloo. Hurley, J.C. and B.L. Parker Rock core investigation of DNAPL penetration and TCE mobility in fractured sandstone. In Ground and Water: Theory to Practice Proceedings of the 55th CGS and 3rd Joint IAH-CNC Groundwater Specialty Conference, Niagara Falls, Ontario, October 20-23, Jakobsen, C Heavy minerals of the upper Cretaceous Chatsworth Formation, Simi Hills, California. In Simi Hills Cretaceous Turbidites, Southern California, The Pacific Section of the Society of Economic Paleontologists and Mineralogists, Los Angeles, USA. Lapcevic, P.A., K.S. Novakowski, and E.A. Sudicky Groundwater flow and solute transport in fractured media. 17 In The Handbook of Groundwater Engineering, ed. J.W. Delleur, Boca Raton, Florida: CRC Press. Lawrence, A.R., P.J. Chilton, R.J. Barron, and W.M. Thomas A method for determining volatile organic solvents in chalk pore waters (southern and eastern England) and its 72

87 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 relevance to the evaluation of groundwater contamination. Journal of Contaminant Hydrology 6, no.4: Lawrence, A.R., M. Stuart, C. Cheney, N. Jones, and R. Moss Investigating the scale of structural controls on chlorinated hydrocarbon distributions in the fractured-porous unsaturated zone of a sandstone aquifer in the UK. Hydrogeology Journal 14, no.8: Link, M.H., R.L. Squires, and I.P. Colpburn Slope and deep sea fan facies and paleogeography of the Upper Cretaceous Chatsworth Formation, Simi Hills, California. The American Association of Petroleum Geologists Bulletin 68: Lipson, D.S., B.H. Kueper, and M.J. Gefell Matrix diffusion-derived plume attenuation in fractured bedrock. Ground Water 43, no.1: MWH Draft Sitewide Groundwater Remedial Investigation Report, Santa Susana Field Laboratory, Ventura County, California National Research Council Rock Fractures and Fluid Flow: Contemporary understanding and applications, Washington D.C.: National Academy of Science. National Research Council Contaminants in the Subsurface: Source Zone Assessment and Remediation, Washington, D.C.: The National Academies Press. Neuman, S.P Trends, prospects and challenges in quantifying flow and transport through fractured rocks. Hydrogeology 13, no.1: Pankow, J.F. and J.A. Cherry Dense Chlorinated Solvents and Other DNAPLs in Groundwater: History, Behavior, and Remediation, Portland, Oregon: Waterloo Press. Parker, B.L Investigating contaminated sites on fractured rock using the DFN approach. In Proceedings of 2007 U.S. EPA/NGWA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, September 24-26, 2007, Portland, Maine, Westerville, Ohio: National Ground Water Association. Parker, B.L., J.A. Cherry, and R.W. Gillham The effects of molecular diffusion on DNAPL behavior in fractured porous media. 12 In Dense Chlorinated Solvents and other DNAPLs in Groundwater, ed. J.F. Pankow and J.A. Cherry, Portland, Oregon: Waterloo Press. Parker, B.L., R.W. Gillham, and J.A. Cherry Diffusive disappearance of immiscible-phase organic liquids in fractured geologic media. Ground Water 32, no.5: Parker, B.L., D.B. McWhorter, and J.A. Cherry Diffusive loss of non-aqueous phase organic solvents from idealized fracture networks in geologic media. Ground Water 35, no.6:

88 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-2: Overview DRAFT December 2009 Pehme, P.E., J.P. Greenhouse, and B.L. Parker The active line source temperature logging technique and its application in fractured rock hydrogeology. Journal of Environmental & Engineering Geophysics 12, no.4: Pierce, A.A Isotopic and hydrogeochemical investigation of major ion origin and trichloroethene degradation in fractured sandstone. Master's thesis, Earth Sciences Department, University of Waterloo. Smith, L. and F.W. Schwartz An analysis of the influence of fracture geometry on mass transport in fractured media. Water Resources Research 20, no.9: Smith, L. and F.W. Schwartz Solute transport through fracture networks. 3 In Flow and Contaminant Transport in Fractured Rock, ed. J. Bear, C.-F. Tsang, and G. de Marsily, San Diego, California: Academic Press. Sterling, S.N Comparison of Discrete Depth Sampling Using Rock Core and a Removable Multilevel System in a TCE Contaminated Fractured Sandstone. Master's thesis, Earth Sciences Department, University of Waterloo. Sterling, S.N., B.L. Parker, J.A. Cherry, J.H. Williams, J.W. Lane Jr., and F.P. Haeni Vertical cross contamination of trichloroethylene in a borehole in fractured sandstone. Ground Water 43, no.4: Sudicky, E.A. and R.G. McLaren The Laplace transform Galerkin technique for largescale simulation of mass transport in discretely fractured porous formations. Water Resources Research 28, no.2: Therrien, R. and E.A. Sudicky Three-dimensional analysis of variably-saturated flow and solute transport in discretely-fractured porous media. Journal of Contaminant Hydrology 23, no.1-2: US EPA Guidance for Evaluating the Technical Impracticability of Ground-Water Resoration, Interim Final, Directive , September VanderKwaak, J.E. and E.A. Sudicky Dissolution of non-aqueous-phase liquids and aqueous-phase contaminant transport in discretely-fractured porous media. Journal of Contaminant Hydrology 23, no.1-2: Waterloo DNAPL Short Course DNAPLs in Fractured Geologic Media: Monitoring, Remediation and Natural Attenuation. In San Francisco and Toronto, June Principal Lecturers: B.L. Parker, J.A. Cherry, D.B. McWhorter. Zimmerman, L. in prep. Contribution of multiple depth point sampling of unpurged well water columns to the understanding of TCE degradation in fractured sandstone. 74

89 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-3 DRAFT December 2009 Contributors to the Site Conceptual Model Documents The Groundwater Advisory Panel Dr. John Cherry Dr. David McWhorter Dr. Beth Parker Steven Chapman Maria Gorecka Jonathan Kennel Jessica Meyer Peeter Pehme Amanda Pierce Dr. Patryk Quinn Laura Zimmerman Dr. Ramon Aravena Dr. Tariq Cheema B.Sc. (Geol. Eng.), University of Saskatchewan; M.Sc. (Geol. Eng.), University of California, Berkeley; Ph.D. (Hydrogeology) University of Illinois; Distinguished Professor Emeritus, Dept. of Earth Sciences, University of Waterloo; specializes in contaminant hydrogeology. B.Sc. (Pet. Eng.), Colorado School of Mines; Ph.D. (Groundwater Hydrology), Colorado State University; Distinguished Professor Emeritus, College of Agricultural and Chemical Engineering, Colorado State University; specializes in multi-phase flow in porous media B.Sc. (Env. Sci. & Econ.), Allegheny College; M.Sc. (Env. Eng.), Duke University; Ph.D. (Hydrogeology), University of Waterloo; currently a Professor and NSERC Industrial Chair, School of Engineering, University of Guelph; specializes in contaminant hydrogeology Academic Institution University of Guelph B.Sc. Eng. (Civil), University of New Brunswick; M.Sc. (Earth Sciences), University of Waterloo; currently a Senior Research Engineer, School of Engineering, University of Guelph; specializes in contaminant hydrogeology and numerical modeling of contaminant transport and fate M.Sc. (Engineering) Gdansk University of Technology, Poland; currently the Laboratory Manager and Senior Analytical Chemist in Dr. Parker's research group at the University of Guelph; specializes in analytical chemistry B.Sc. (Non-Major Science), University of Waterloo; M.Sc. (Contaminant Hydrogeology), University of Waterloo; currently a Research Associate, School of Engineering, University of Guelph; specializes in contaminant hydrogeology B.Sc. (Env. Geol.) (Honors), University of Montana; M.Sc. (Hydrogeology), University of Waterloo; currently a Ph.D. Candidate, School of Environmental Sciences, University of Guelph; specializes in contaminant hydrogeology B.Sc., M.Sc.; currently a Ph.D. Candidate, Dept. of Earth Science, University of Waterloo and President of Waterloo Geophysics Inc; specializes in borehole geophysics B.Sc. (Geol.) (Honours), University of Toronto; M.Sc. (Hydrogeology), University of Waterloo; currently a Research Associate, School of Engineering, University of Guelph; specializes in hydrogeochemistry and contaminant hydrogeology Ph.D. (Earth Sciences), University of Waterloo; currently a Post Doctoral Fellow, School of Engineering, University of Guelph; specializes in packer testing in fractured porous media B.Sc. (Env.), University of Guelph; currently a Masters Candidate, School of Engineering, University of Guelph; specializes in contaminant hydrogeology University of Waterloo Licenciate Santiago, Chile (Chemistry), Universidad Catolica de Santiago; M.Sc. (Earth Sciences), University of Waterloo; Ph.D. (Earth Sciences), University of Waterloo; currently a Research Professor, Dept. Earth and Environmental Sciences, University of Waterloo; specializes in isotope hydrology and isotope geochemistry. B.Sc. (Geol.), Punjab University, Lahore, Pakistan; M.Sc. (Eng. Geol.) University of Leeds; Ph.D. (Geol. Eng.) South Dakota School of Mines; currently a Research Hydrogeologist with Syncrude Canada Ltd; specializes in groundwater assessment and remediation. 1 of 4

90 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-3 DRAFT December 2009 Dr. Tadeusz Gorecki Dr. Mario Ioannidis Dr. Will Robertson M.Sc. (Eng.), Gdansk University of Technology, Poland; Ph.D. (Chem.), Gdansk University of Technology, Poland; currently a Professor, Dept. of Chemistry, University of Waterloo; specializes in the development of analytical methods Diploma in Engineering, Patras, Greece; Ph.D. University of Waterloo; currently an Associate Professor and Director of Nanotechnology at the University of Waterloo; specializes in transport processes in porous materials Currently a Professor, Dept. of Earth and Environmental Sciences, University of Waterloo; specializes in physical hydrogeology and geochemistry of contaminated groundwater University of New Brunswick Dr. Tom Al B.Sc. & M.Sc. (Earth Science), Memorial University of Newfoundland; Ph.D. (Geol.), University of Waterloo; currently a Professor, Department of Geology, University of New Brunswick Diana Loomer B.Sc. (Env. Geochem.), University of New Brunswick; M.Sc. (Geol.), University of New Brunswick; currently a Research Scientist, Department of Geology, University of New Brunswick; specializes in geologic material characterization for determination of contaminant transport and fate University of Montana Dr. William Woessner Dr. Michael Perkins Dr. David Freedman Dr. Ramona Darlington B.A. (Geol.), College of Wooster; M.Sc. (Geol.), University of Florida, Gainesville; M.Sc. (Wat. Res. Mng.), University of Wisconsin, Madison; Ph.D. (Geol.), minor in Civil and Environmental Engineering, University of Wisconsin, Madison; currently the Chair of the Dept. of Geosciences, University of Montana University of Utah B.Sc. (Eng. Geoscience), University of California, Berkeley; M.Sc. (Geol.), University of California, Berkeley; Ph.D. (Geol.), University of Utah; currently an Adjunct Professor, Dept. of Geology and Geophysics, University of Utah Clemson University B.Sc. (Science and Environmental Change), University of Wisconsin, Green Bay; M.Sc. (Env. Eng.), University of Cincinnati; Ph.D. (Env. Eng.), Cornell University; currently a Professor, Dept. of Environmental Engineering and Earth Sciences, Clemson University Ph.D. Clemson University; currently with the Department of Environmental Engineering and Earth Sciences at Clemson University; specializes in microbiology and bio attenuation Richard Andrachek Dr. Garrett Hazelton Consulting Firms/Independent Consultants MWH B.Sc. (Mech. Eng.) West Virginia University; P.E.; currently the Principal Engineer at MWH; specializes in environmental characterization and restoration project management as well as California and Federal environmental regulations B.Sc. (Geol.); Ph.D. (Geol.), California State Polytechnic University, Ramona; currently a Senior Geologist with MWH; specializes in fault mechanics, geochronology, structural geology and the tectonics of Southern California Dr. Nicholas Johnson B.A. (Earth Sciences), University of California, Santa Cruz; M.S. (Hydrology), University of Arizona; Ph.D. (Earth Sciences), University of California, Santa Cruz; P.G., C.Hg.; currently a Principal Hydrogeologist at MWH; specializes in site and basin-wide conceptual and numerical models for watersupply and contaminant hydrogeology 2 of 4

91 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-3 DRAFT December 2009 Steve Reiners Dr. Geoffrey L. Upson Daron Abbey Paul Martin Kurt Blust Sheldon Clarke Laura Davis Larry Smith Edward Clayton Daniel Gomes Jennifer Hurley Robert Will Carl Keller Dr. John Greenhouse Sanford Britt Sean Sterling Seth Pitkin B.Sc. (Chem. Eng.), University of Colorado; CHMM; currently the Principal Engineer at MWH B.Sc. (Agronomy), Colorado State University; Ph.D. (Soil Science), Colorado State University, Fort Collins, Colorado; currently a Principal Geochemist at MWH; specializes in contaminant fate and transport modeling in the vadose zone AquaResource B.Sc. (Env. Sci.) (Honours), Carleton University; M.Sc. (Earth Sciences), Simon Fraser University; currently a Senior Hydrogeologist at AquaResource Inc; specializes in conceptual and numerical groundwater modeling B.Sc. (Civil Engineering), M.Sc. (Earth Sciences), University of Waterloo; currently President and Senior Hydrogeologist, AquaResource Inc.; specializes in physical hydrogeology conceptualization and numerical modeling Haley & Aldrich BS (Geology), University of Utah; PG, RG; currently a Senior Hydrogeologist and Vice President at Haley & Aldrich B.S. (Biological Science), University of Arizona; currently a Vice President at Haley & Aldrich B.S. (Pyschology), University of Notre Dame; M.S. (Hydrogeology), University of Arizona; PG, RG; currently a Senior Hydrogeologist at Haley & Aldrich B.A. (Geol.) SUNY Geneseo; M.Sc. (Geol.) University of Arizona; currently the Executive Vice President and Chief Operating Officer at Haley and Aldrich Schlumberger Water Services M.Sc.; currently a Senior Geophysical Logging Engineer, Schlumberger Water Services; specializes in hydrogeology and geophysical logging M.Sc.; currently the Operations Manager, Sacramento Schlumberger Water Services; specializes in hydrogeology and groundwater modeling M.Sc. (hydrogeology), University of Waterloo; P.Geo.; currently a Domain Expert for Schlumberger Water Services; specializes in hydrogeology and data management B.A. (Geol.); M.Sc. (Geophysics); Ph.D. (Pet. Eng.); currently a Principle Engineer and PETREL expert, Schlumberger Water Services FLUTe B.Sc. (Physics); M.Sc. (Physics), Rensselaer Polytechnic Institute; Founder and President of FLUTe; specializes in borehole hydraulic conductivity profiling Waterloo Geophysics Inc. B.Sc. (Physics), University of British Columbia; M.Sc. (Physics), University of British Columbia; Ph.D. (Geophysics), University of California, San Diego; currently works with Waterloo Geophysics Inc.; specialty is geophysical logging ProHydro Inc. B.A. (Geol.), University of Rochester; M.Sc. (Geol.), University of Southern California; currently the Principle Hydrogeologist, founder of ProHydro Inc., and inventor of the Snap Sampler Intera Engineering Ltd. B.Sc. Queens University; M.Sc. (Hydrogeology) University of Waterloo; currently a Senior Hydrogeologist at Intera Engineering Ltd. Stone Environmental Inc. B.Sc. (Geol.), The Evergreen State College; M.Sc. (Hydrogeology), University of Waterloo; currently the Vice President of Stone Environmental Ltd.; specializes in contaminant hydrogeology 3 of 4

92 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-3 DRAFT December 2009 Robert Gailey Dr. Ross Wagner Dr. Bill Arnold Dr. Scott James Independent B.Sc. (Geol. & Biol.), Brown University; M.Sc. (Applied Hydrogeology), Stanford University; MBA, University of California, Berkeley - Walter A. Haas School of Business; P.G., C.Hg.; founder and Principal of R. M. Gailey Consulting Hydrogeologist; specializes in contaminant hydrogeology and flow and contaminant transport modeling B.Sc. (Geol.), University of California, Berkeley; Ph.D. (Geol.), University of California, Berkeley; P.G.; currently a professional geologist working as an independent consultant Government Agency US DOE Sandia National Laboratory B.Sc. (Geol.), Kansas State University; M.Sc. (Geol.), Kansas State University; Ph.D. (Hydrogeology), University of Wisconsin, Madison; currently a Scientist in the Nuclear Waste Management Center, Sandia National Laboratories; specializes in modeling groundwater flow and radionuclide transport in porous media B.Sc./M.Sc. (Mechanical Engineering) University of California, San Diego; Ph.D. (Engineering) University of California, Irvine; currently Acting Manager of the Thermal/Fluid Science & Engineering Department at Sandia Labs in Livermore, California; specializes in a wide variety of reactive flow and transport modeling problems 4 of 4

93 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-4 DRAFT December 2009 Summary of Documents in Support of the Site Conceptual Model 0 SCM Element Section Introduction Synopsis Title Authors No. The Groundwater Advisory Panel 0-1 Overview of the Site Conceptual Model for the Fate and Migration of Contaminants in Groundwater at the Santa Susana Field Laboratory, Simi, California The Groundwater Advisory Panel Contributors to the Site Conceptual Model Documents Summary of Documents in Support of the Site Conceptual Model Element Description Title Authors No. The Discrete Fractured Network (DFN) Approach for Characterization of Contaminated Fractured Rock Sites Parker 1-1 Vertical Cross Contamination of Trichloroethylene in a Borehole in Sterling, Parker, Cherry, Fractured Sandstone (Ground Water, 2005, 43(4), ) Williams, Lane, Haeni 1-2 A New Depth-Discrete Multilevel Monitoring Approach for Fractured Rock (Ground Water Monitoring and Remediation, 2007, 27(2), 57-70) Cherry, Parker, Keller 1-3 The Active Line Source Temperature Logging Technique and its Application in Fractured Rock Hydrogeology (Journal of Environmental and Engineering Geophysics,2007, 12(4), ) 0-2 Pehme, Greenhouse, Parker DFN Approach "The Discrete Fracture Network (DFN) Approach was required to characterize contaminant distributions and behavior in the Chatsworth Formation." Improved Resolution of Ambient Flow through Fractured Rock with Temperature Logs (Ground Water, in press) Detailed hydraulic head profiles as essential data for defining hydrogeologic units in layered fractured sedimentary rock (Environmental Geology, 2008, 56, 27-44) A New Innovative Method for Continuous Hydraulic Conductivity Profiling in Fractured Rock Holes A New Method for Measurement of Volatile Organic Contaminants (VOCs) in Rock Core A New Downhole Passive Sampling System to Avoid Biases from Sample Handling A versatile packer testing system for high resolution hydraulic testing in fractued rock boreholes Development of a Storage System for Data from the Discrete Fracture Network (DFN) Approach Pehme, Parker, Cherry, Greenhouse 1-5 Meyer, Parker, Cherry 1-6 Keller, Cherry, Parker 1-7 Parker, Sterling 1-8 Britt, Parker, Cherry 1-9 Quinn, Cherry, Parker 1-10 Kennel, Meyer, Parker, Cherry Porosity 3 Reactive 4 Systematic "The rock matrix porosity provided by interconnected pores is large and the bulk fracture porosity is extremely small." "The rock matrix composition includes abundant reactive minerals and appreciable organic matter." "The fracture network is a systematic arrangement of bedding parallel fractures and steeply-dipping joints." Statistical Synthesis of Imaging and Porosimetry Data for the Characterization of Microstructure and Transport Properties of Sandstones (Transport in Porous Media, submitted) Amirtharaj, Ioannidis, Parker, Tsakiroglou The Chatsworth Formation underlies most of the SSFL and its matrix porosity provided by interconnected pores is large Hurley, Cherry, Parker 2-2 An Analysis of the Distribution of Matrix Vertical Hydraulic Conductivity within the Chatsworth Formation Hurley, Cherry, Parker 2-3 Mineralogical characterization of drill core samples from the Santa Susana Field Laboratory, Ventura County, California Loomer, Al 3-1 The rock matrix composition includes abundant reactive minerals Hurley, Cherry, Parker 3-2 The rock matrix composition includes appreciable amounts of solidphase organic carbon Hurley, Cherry, Parker 3-3 Characteristics of Joints at the Santa Susana Field Laboratory, Simi, California Wagner, Perkins 4-1 Ground Truthing and Interpretation of Discrete Feature Data from Core, Televiewer, and Geophysical Logs at the Santa Susana Field Laboratory (SSF), Ventura County, California Schlumberger Water Services (Will, Clayton, Gomes) 4-2 Determination of Fracture Density and Fracture Set Orientations from Geophysical and Geologic Logs for the Chatsworth Formation at the Santa Susana Field Laboratory, Simi, California 2-1 Kennel, Parker, Cherry 4-3 Lines of Evidence Indicating Strong Fracture Interconnectivity in the Chatsworth Formation, Santa Susana Field Laboratory, Simi, California Cherry Interconnected Evaluation of Fracture Network Characteristics Based on Detailed Hydraulic Head Profiles Insights Concerning Aquitards and Fracture Network Characteristics "The fracture networks have from Detailed Head Profiles in Fractured Sandstone strong hydraulic connectivity both horizontally and vertically." Evidence from temperature profiles for deep groundwater flow in an interconnected fracture network in sandstone Analysis of Pumping Test Results and Multilevel Interconnectivity Testing for Evidence of Interconnected Fracture Network Meyer, Parker, Cherry 5-1 Meyer, Chapman, Parker, Cherry Insights about Contaminant Migration Pathways in Fractured Sandstone Parker, Sterling Pehme, Parker, Cherry 5-4 Reiners, Johnson of 3

94 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-4 DRAFT December 2009 SCM Element 6 Bulk K 7 Faults 8 Apertures Element Description Title Authors No. Bulk Hydraulic Conductivity, Santa Susana Field Laboratory, Ventura County, California Reiners, Johnson 6-1 Martin, Abbey, Cherry, "The bulk hydraulic conductivity Decreased Hydraulic Conductivtiy with Depth at the SSFL Parker, Andrachek, 6-2 is low to moderate." McWhorter Mulitiple Lines of Evidence Concerning Anistropy in the Chatsworth Formation Martin, Abbey, Cherry, Parker, Andrachek, Johnson, McWhorter Evaluation of Fault Zone Permeability Structions Johnson 7-1 "There is low to moderate largescale transmissivity across and Indications of the Bulk Hydraulic Conductivity of Faults from the along the faults and shear zones." Three-Dimensional Mountain-Scale Groundwater Flow Model of the Santa Susana Field Laboratory "The effective fracture apertures that dominate groundwater flow and contaminant transport are small to moderate." Fracture Apertures, Chatsworth Formation, Santa Susana Field Laboratory Hydraulic Apertures Determined by Application of the Cubic Law to FLUTe Continuous Hydraulic Conductivity Profiling at the Santa Susana Field Laboratory, Simi, California Abbey, Martin, Zhang, James, Gabriel, Andrachek McWhorter, Reiners 8-1 Kennel, Parker, Cherry, Reiners Continuity "The groundwater in overburden and highly fractured zone in shallow bedrock forms a subsurface hydrologic continuum with deeper groundwater in the Chatsworth Formation." The Shallow-Deep Hydraulic Continuum at the Santa Susana Field Laboratory Insights from Atmospheric Tritium Concerning Groundwater Recharge and Solute Transport in Fractured Sandstone at the Santa Susana Field Laboratory, Simi, California Abbey, Martin, Bester, Andrachek, McWhorter Cherry, Pierce, Parker, Chapman, Abbey, Martin, McWhorter Recharge Variability Hydrochemistry "Groundwater recharge is a small percent of the mean annual precipitation." "The geologic and hydrogeologic variability in the Chatsworth Formation are appropriately represented in the 3-D groundwater flow model (FEFLOW) used to search for zones that could allow exceptionally large or rapid contaminant migration to offsite areas of concern." "The groundwater contains many dissolved inorganic constituents of natural origin that provide insight concerning groundwater flow and contaminant behavior." Concepts for Shallow Groundwater Behaviour and Recharge McWhorter 10-1 Groundwater Recharge, Santa Susana Field Laboratory, Ventura County, California Simulating Groundwater Flow in a Complex Setting: Development and Application of the Three-Dimensional Mountain-Scale Groundwater Flow Model of the Santa Susana Field Laboratory Application of Singular Value Decomposition and Regularization Techniques for Optimizing the Mountain-Scale Groundwater Flow Model for the Santa Susana Field Laboratory Application of the Three-Dimensional Groundwater Flow Model at the Santa Susana Field Laboratory to Evaluate Flow Path Uncertainties Hydrochemical Processes in Groundwater in a Late Cretaceous Fractured Sandstone Mountain, Simi Hills, California Use of Strontium Isotopes in Groundwater Flow Pathway Analysis in Fractured Sandstone Carbon-14 as an Environmental Tracer in Fractured Sandstone with Matrix Diffusion Origin of Minor Constituents in the Chatsworth Formation at the Santa Susana Field Laboratory, Ventura County, California McWhorter 10-2 Abbey, Martin, Andrachek, Woessner, Cherry, Parker, Wagner, Zhang, Gabriel Abbey, Martin, Zhang, James, Arnold, Andrachek, Gabriel, Woessner Abbey, Martin, Zhang, Woessner, James, Arnold, Andrachek, Johnson, Cherry, Parker Pierce, Cherry, Parker, Aravena, Loomer, Al Pierce, Aravena, Cherry, Parker Pierce, Aravena, Cherry, Parker, Chapman Pierce, Cherry, Parker, Aravena Seeps "A substantial portion of the groundwater originating in the SSFL property discharges at seeps, springs and phreatophytes situated on the mountain slopes offsite from SSFL that show no detectable concentration of the SSFL contaminants-of-concern." Origin and Hydrochemistry of Seeps and Springs Issuing from Fractured Bedrock in the Simi Hills (Transverse Ranges) of Southern California Geologic and Hydrogeologic Controls on Seeps Along Simi Hill Slopes Pierce, Wagner, Cherry, Parker, Aravena, Abbey, Martin Wagner, Pierce, Cherry, Parker, Abbey, Martin DNAPL "The chlorinated solvent contamination was initially caused by DNAPL penetration below the water table, but the DNAPL has since been converted to dissolved and sorbed mass now residing in the rock matrix and therefore DNAPL flow no longer occurs." Deep penetration of TCE in sandstone at the Santa Susana Field Laboratory is attributed to DNAPL flow in fractures Hurley, Cherry, Parker DNAPL Disappearance "Large TCE mass as dissolved and sorbed phases occurs at and in the local vicinity of TCE DNAPL input locations." TCE DNAPL has been converted to dissolved-phase TCE, and this mass has stayed close to home Hurley, Cherry, Parker 15 2 of 3

95 Site Conceptual Model for the Migration and Fate of Contaminants in Groundwater at the SSFL SCM Document 0-4 DRAFT December SCM Element Diffusion Element Description Title Authors No. "The behaviour of all soluble chemicals and radionuclides is strongly influenced by diffusion into and out of the porous rock matrix." Evidence for Strong Matrix Diffusion of VOCs, Perchlorate and Tritium in Fractured Sandstone near Simi, California Summary of Tortuosity Measuremetns on SSFL Samples Parker, Chapman, Hurley, Andrachek, Clark, and Cherry Chapman, Parker, Cherry, Hurley Vadose 18 Monitorable 19 Retardation "Contaminant mass resides in the vadose zone but vadose zone mass DNAPL Fate and Transport in the Vadose Zone, Santa Susana Field now has minimal or no influence Laboratory, Ventura County, California on plume fronts." "Contaminant plumes can be characterized and monitored." "Matrix diffusion causes plume fronts to be strongly retarded relative to the mean groundwater velocity in the fracture network and along with dispersion, causes maximum plume concentrations to continually decline (be attenuated)." Examination of Long-Term Monitoring Well Data for Plume Monitorability Comparison of blended samples caused by well purging with profiles from non-purged water columns in fractured sandstone Examination of Spatial Relations Between Vadose Zone (Vapour and Soil Contaminant Analyses) and Groundwater Monitoring Well Data and Rock Core Contaminant Profiles Evidence for TCE Attenuation in the Chatsworth Formation Based on Long Term Monitoring Well Records Chlorinated Solvent Plume Front Retardation due to Matrix Diffusion in Fractured Sandstone (Constant Source) Chlorinated Solvent Plume Front Retardation due to Matrix Diffusion in Fractured Sandstone (Finite Source) Field validation of strong plume attenuation in a fractured porous sandstone Simulating Plume Transport in a Complex 3-D Bedrock Flow System using FEFLOW and FRACTRAN McWhorter 17-1 Pierce, Parker, Cherry 18-1 Zimmerman, Parker, Britt 18-2 Cherry, Kennel, Pierce, Parker Pierce, Parker, Cherry Chapman, Parker, Cherry 19-2 Chapman, Parker, Cherry 19-3 Parker, Kennel, Chapman, Cherry Chapman, Parker, Abbey, Martin, Cherry, Andrachek, Woessner Biotic and Abiotic Anaerobic Transformations of Trichloroethene and cis-1,2-dichloroethene in Fractured Sandstone (Environmental Science and Technology, 2008, 42(12), ) Darlington, Lehmicke, Andrachek, Freedman Attenuation Anaerobic Abiotic transformations of cis-1,2-dichloroethene by "Degradation is contributing to minerals in fractured sandstone some degree in the attenuation of Field Evidence for Trichloroethylene Degradation Mechanisms in all contaminant plumes in Fractured Sandstone addition to the attenuation caused Contribution of multiple depth point sampling of unpurged well water by matrix diffusion, sorption, and columns to the understanding of TCE degradation in fractured dispersion." sandstone Numerical Modeling of Chlorinated Solvent Plume Attenuation due to Matrix Diffusion and Degradation in Fractured Sandstone Possibility of Natural Attenuation of Perchlorate with Pyrite as Energy Source: A Literature Review Darlington, Freedman 20-2 Pierce, Aravena, Cherry, Parker 20-3 Zimmerman, Parker, Britt 20-4 Chapman, Parker, Cherry 20-5 Robertson of 3

96 SITE CONCEPTUAL MODEL ELEMENT 1: The Discrete Fracture Network (DFN) Approach was required to characterize contaminant distributions and behavior in the Chatsworth Formation. Overview The reporting on this SCM element includes eleven documents. The first document (1-1) presents a rationale and broad overview of the DFN approach. The development of this approach began at the SSFL in 1997 with the drilling of RD-35B and RD-45B where the rock core VOC contaminant analysis method, as reported on in 1-8, for detailed core hole profiling was first applied. Application of the method at these two locations showed nearly all contaminant mass resided in the rock matrix blocks between fractures rather than in the fractures. Also, comparison of the rock core profiles with groundwater concentrations from the conventional monitoring found groundwater was often impacted at the well by construction practices as reported in document 1-2. These observations lead to evaluation of contaminant source zones to be conducted using rock cores and plume characterization using groundwater monitoring. The DFN approach includes many types of data acquisition, some of which are relatively new. Five of the new methods are described in detail in documents including the use of FLUTe liners and FLUTe multi-level monitoring systems (1-3 and 1-7), high resolution temperature profiling in lined holes (1-4 and 1-5), Westbay multi-level systems equipped with a large number of monitoring ports to obtain detailed hydraulic head and geochemical profiles (1-6), downhole passive sampling using the Snap Sampler system (1-9), and high resolution packer testing (1-10). The application of these new methods at the SSFL has been essential for the development of the SSFL SCM. However, the proof of the performance and value of these methods has come from applications not only at SSFL but other fracture rock sites currently under intensive characterization. When the comprehensive investigations of the Chatsworth Formation began in 2000, the DFN approach was in the earliest stage of development and some of the components had not yet been fully developed or proven. However, within a few years the effectiveness of this approach became clear, hence its application to SSFL. Of the eleven documents that support this SCM element, five are complete and published (1-2, 1-3, 1-4, 1-5, 1-6) and five documents (1-1, 1-7, 1-8, 1-9, 1-10) need revisions and expansions done with a planned submittal in Early 2010 where they will be subjected to scientific peer review. The last document on the list (1-11) describes very briefly a data storage and management system recently developed that now serves as an important tool to access and manage the SSFL data base. This system is tailored specifically to the exceptionally large and diverse types of data obtained from drill holes when the full DFN approach is applied. Much of the SSFL subsurface data has been entered into this system and completion is projected for The goal of this system is to facilitate interpretation of complex data sets through relational displays using existing commercial software for data graphing/display.

97 DRAFT The Discrete Fractured Network (DFN) Approach for Characterization of Contaminated Fractured Rock Sites Beth L. Parker For Submittal to: Groundwater Monitoring and Remediation, Draft: November 16, 2009 ABSTRACT An investigation is in progress at a contaminated site where several TCE plumes occur in fractured sandstone with mudstone/shale interbeds near Simi, CA. The goal is to characterize the plumes using new methods because the previous approach based on conventional borehole geophysics and monitoring wells provided inadequate resolution of contaminant distributions. The first new method, initiated in 1997, involved measurement of volatile organic compounds (VOC) concentrations at closely spaced intervals in rock samples from continuous cores drilled at former (1940s -1990s) TCE DNAPL input locations. These core analyses showed that diffusion has caused nearly all contaminant mass (TCE and degradation products) to reside in the rock matrix blocks between fractures rather than in open fractures. Therefore, contaminant delineation must be accomplished using such rock core analysis rather than monitoring wells. However, to understand the existing contaminant distributions and calibrate flow and transport models to predict future plume behaviour, controls on transport and fate must be quantified. This quantification requires, in addition to rock core contaminant analyses, application of a broad suite of recently developed field methods, including high resolution temperature profiling of static columns inside boreholes temporarily sealed with FLUTe flexible liners, continuous hydraulic conductivity profiling using these liners, high resolution borehole imaging and multi level monitoring systems (Westbay, FLUTe) equipped with large numbers of ports to produce detailed profiles of hydraulic head and water chemistry. Data from all of these new methods and straddle packer tests provide information concerning the fracture network (fracture density and aperture distributions) and rock matrix properties (e.g. porosity, permeability, tortuosity, foc) used in discrete fracture network (DFN) models for contaminant transport and fate. This suite of complementary investigation methods, referred to as the DFN approach, has been applied intensively at the Simi site, a second sandstone site in the USA (Wisconsin) and two dolostone sites in southern Ontario, Canada. At these four sites VOC plumes have been characterized with emphasis on both the fracture network and the rock matrix. The DFN approach is used to characterize plumes and once this characterization has reached a satisfactory stage, the subsequent endeavor can be the design of an efficient long term monitoring network for the plumes.

98 DRAFT BACKGROUND AND INTRODUCTION The behaviour of contaminants in fractured rock is now one of the few remaining scientific frontiers in physical hydrogeology. The status of knowledge concerning groundwater flow and contaminant migration in fractured rock has been reviewed by the U.S. National Research Council (NRC, 1996), Lapcevic et al. (1999), Berkowitz (2002) and Neuman (2005). These reviews encompass considerable published literature concerning the conceptual nature of fractures and hydraulic conditions in fracture networks based on borehole investigations in uncontaminated fractured rock (primarily based on work by the petroleum industry, USGS studies of the Mirror Lake granitic system in New Hampshire and investigations of prospective radioactive waste repositories). Furthermore, many publications concern mathematical models representing groundwater flow and/or contaminant transport in hypothetical or idealized fracture networks for contaminant behaviour in fractured rock systems (e.g., Smith and Schwartz, 1984; 1993; Sudicky and McLaren, 1992; Therrien and Sudicky, 1996; and many others). However, these modelling endeavours generally do not represent actual field sites or any particular type of rock. Furthermore field data of actual contaminant distributions and contaminant behaviour in fractured rock, particularly sedimentary rock, are almost non-existent. Concepts for the nature of contaminant plumes in fractured rock have been speculative and generally parameterization of model inputs inadequately supports field data. Although many techniques for borehole logging and hydraulic testing exist for use at contaminated sites (e.g. review by Sara, 2003), there is a general agreement in the literature that these techniques are severely limited in their prospects for providing the necessary Page 2 of 25

99 quantitative information about fracture networks for understanding contaminant transport and fate in rock (NRC, 1996; Berkowitz, 2002). Unlike behaviour in igneous rock, contaminants in sedimentary rock can reside predominately in the porous rock matrix while downgradient transport occurs in the fractures (e.g. Foster 1975, Lipson et al. 2005). Therefore, determination of the contaminant distribution in sedimentary rock requires measurement of contaminant concentrations in both the fracture network and the rock matrix. Lawrence et al (1990, 2005) and Sterling et al. (2005) provide field examples of measurement of chlorinated solvent concentrations in the rock matrix. However, much of the literature pertaining to groundwater flow and solute behaviour in fractured rock concerns igneous rock such as granite. Several countries have proposed creation of deep repositories for radioactive waste in granitic rock and the search for and assessment of prospective sites has involved intensive field studies. However, these studies have not involved existing contaminant plumes because such plumes do not (yet) exist in these environments (i.e., radioactive waste has not yet been disposed of in this type of rock). Fractured rock research has included tracer experiments but their spatial scale is small in relation to the relevant plume scale. The literature contains no well-documented delineations of industrial contaminant plumes in any type of fractured rock. The difficulty of the challenge posed in the quest to delineate and understand plumes in fractured rock is much greater than that posed by granular media because the scale of variability and complexity imposed by fracture networks is much greater and costs per borehole are much larger due to the greater depths of contamination common in fractured rock. An important contribution to the understanding of contaminant behaviour in granular aquifers has been large-scale, natural-flow tracer experiments with detailed 3-D monitoring to examine effects of heterogeneity on dispersion (e.g. Sudicky 1986, Garabedian et al. Page 3 of 25

100 1991). Such natural-gradient experiments at relevant spatial scales have not been conducted in fractured rock and are generally cost-prohibitive. Therefore, for fractured rock there is no alternative but to rely on intensive studies of actual contaminated sites to gain insights concerning plume formation and evolution and quantify the influences of the various processes such as advection, dispersion and degradation. In essence, the plumes represent long-term, large-scale tracer experiments, which require application of appropriate investigative methods to understand the factors governing the plumes. The purpose of this paper is to describe a new investigative approach for characterization of contaminated fractured sedimentary rock sites and show the applicability of the approach by means of examples of results primarily from a site in southern California where the approach has been applied intensely. The ultimate goal of the DFN approach applied at most sites is the acquisition of site data for use in DFN numerical models for groundwater flow and contaminant transport and fate in fracture networks. Several such models are available, including FRACTRAN (Sudicky and McLaren, 1992), FRAC3DVS (Theirren and Sudicky, 1996), SMOKER (Molsen, 200X) and FRACMAN (reference), which require assignment of parameter values to represent physical and chemical properties of the rock matrix blocks and also parameter values for the fractures and fracture network. The rock core contaminant analysis method with closely spaced measurements applied at numerous sedimentary rock sites where VOCs entered the rock decades ago shows that nearly all contaminant mass resides in the rock matrix where groundwater flow is minimal and not in the fractures where groundwater flow is most active. Therefore, application of the rock core contaminant analysis method is essential to determining the Page 4 of 25

101 nature and extent of contamination. However, application of the other components of the DFN approach is necessary to improve confidence in both the characterization of the contamination and prediction of future conditions. The degree of application needed depends on site specific conditions and the desired reliability of the predictions. The DFN approach is much different than the conventional approach used for determining the nature and extent of contamination in bedrock, which relies primarily on data collected from monitoring wells. In the conventional approach the boreholes are commonly subjected to open-hole borehole geophysical measurements and open hole flow metering and then a monitoring well or multilevel system with minimal monitoring ports is installed. In the DFN approach, the time allocated to open hole measurements is short to minimize the effects of cross contamination caused by vertical flow up and/or down the hole between fractures. As soon as possible after drilling the hole, a flexible liner is installed. However in the time between the end of drilling and the installation of the liner, it may be appropriate to conduct borehole geophysical measurements to determine rock properties, locate geological features and identify fractures by borehole imaging and televiewing. In Figure 2, these measurements are displayed under the open hole designation. The liner, composed of impervious flexible urethane coated nylon is installed as the liner is filled with water to a level well above the blended water level in the open corehole. During the installation of the liner, a method has been developed by Keller et al. (in submittal) in which the measured rate of liner descent is used to calculate a profile of borehole transmissivity versus depth denoted as FLUTe K profiling in Figure 2. This method is different from borehole packer testing in that it requires a relatively short period of time to complete and the measurements encompass the entire borehole. Therefore, relatively high transmissivity zones are not missed. Page 5 of 25

102 Straddle packer testing can play a role in the DFN method, whereby the testing is done on a small number of borehole intervals selected based on results of the open-hole geophysics. Straddle packer testing is minimized in boreholes in contaminated zones to reduce borehole cross contamination. More comprehensive packer testing may be done in holes where contamination is minimal. After the FLUTe liner is in place, geophysical logging and temperature logging are done inside the liner where a static water column exists. Pehme et al (2006 and 2009 in press) describe high resolution temperature logging inside FLUTe lined holes. In the DFN approach, site characterization and monitoring are two different but related endeavours with different goals. Site characterization pertains to the nature and extent of the contamination as it exists today, understanding of how the distribution developed, and prediction, predicting future migration and fate using conceptual and mathematical models. In contrast, site monitoring concerns primarily the design and operation of groundwater sampling networks over time to assess the predictions of contaminant migration and accomplish early warning monitoring at the locations deemed most appropriate based on the results of the DFN approach for site characterization. Page 6 of 25

103 ORIGIN AND COMPONENTS OF THE APPROACH Ten years ago I initiated use of chemical analyses (rock core VOC analyses) done at very closely spaced vertical intervals on contaminated sandstone core in the style presented in Figure 1 to determine the nature of the contaminant distribution at a location in California where TCE had entered sandstone decades earlier. This evolved into a systematic way to investigate contaminated bedrock and is now referred to as the discrete - fracture network (DFN) approach represented in the Figure 2. This core-focused field study grew out of conceptual modeling supported by analytical modeling concerning dissolution and diffusion effects on chlorinated solvent DNAPL in fractured porous geologic media represented by fractured clay and fractured sandstone with literature derived parameters (Parker et al., 1994; ). From the initial rock core analyses done at the California sandstone site mentioned above (Sterling 1999) it was evident that the DNAPL had initially flowed primarily downward through a network of many interconnected fractures, spaced 1-5 m apart, and that over the subsequent years or decades, all or nearly all of the immiscible phase liquid has been transferred by dissolution and diffusion into the rock matrix blocks between the fractures where the mass now resides in the dissolved and sorbed phases. Comparison of the rock core contaminant profiles with groundwater analyses done on samples from conventional monitoring wells and multilevel systems (MLSs) showed that these water analyses gave misleading results because of effects of vertical flow in the holes when the holes were open, allowing cross contamination between fractures with different initial concentrations (Sterling et al., 2005). Conventional methods of borehole geophysics and hydrophysics also gave misleading results about flow in the sense that the aim is to Page 7 of 25

104 understand the flow in the fracture network during ambient conditions not the disturbed flow imposed by the open borehole. From this initial experience a decade ago and subsequent experiences shortly thereafter at other sites where the early version rock core VOC analyses method was applied, the DFN approach was designed to determine the distribution, transport and fate of contaminants in sedimentary rock. The DFN approach has now been applied to some degree at more than 18 sites with chlorinated solvent contamination, from which four were selected as the focus of a long-term intensive field studies. Two of these are in the USA (California and Wisconsin) and two in Canada (Cambridge and Guelph in the Province of Ontario). Table 1 provides a summary of the hydrogeologic conditions at these four sites. Chlorinated solvents have been in the subsurface beneath many industrial properties for several decades allowing plumes to migrate down-gradient several hundreds to thousands of metres or more. These contaminants can now serve as tracers to study contaminant migration over the relevant large spatial and time scales most relevant in contaminant hydrogeology. Chlorinated solvent compounds are not naturally occurring in the environment; hence, even extremely low level detects (possible due to exceptional measurement sensitivity) serve as reliable evidence of contamination over several orders of magnitude. The physical and chemical properties of the common chlorinated solvents make them good indicators of the physical hydrogeologic system characteristics, including the fracture network connectivity and distribution of groundwater flow. The DFN approach encompasses several types of measurements made on samples from continuous rock core and then several types of measurements made in coreholes. The results of these many types of measurements are used in combination within the framework of mathematical models to develop the site conceptual model that serves as the framework Page 8 of 25

105 for site characterization. Figure 2 shows the elements of the DFN approach pertaining to cores and coreholes. At some sites where the bedrock outcrops, surface mapping of geology and fractures can be an important contributor of information. The DFN approach as outlined in Figure 2 pertains only to site characterization and contaminant delineation and does not directly concern site monitoring. For contaminant delineation in sedimentary rock, chemical analysis of rock core samples provides the primary data with analysis of water samples from wells and/or multilevel monitoring systems being secondary and in some case not used at all. Therefore, essentially all of the other activities included in the DFN approach serve the goal of development of understanding the transport and fate of the contaminant distributions determined from the rock core contaminant analysis. This version of the DFN approach is intended for application at contaminated fractured sedimentary rock sites where the contaminants entered the rock years or decades ago and diffusion halos are expected to be large enough to be identified using the rock core analysis method. If this is not the case or if the site is situated on crystalline rock where diffusion of contaminants from fractures into the rock matrix is minimal, than the importance of the rock core measurements of all types is much diminished, in which case all of the other components of the approach must be relied on for site characterization. More detailed dispersion of the applicability of the DFN approach to such sites is beyond the scope of this paper. Page 9 of 25

106 ROLE OF ROCK CORE CONTAMINANT ANALYSES The spatial distribution of contaminants within chlorinated solvent plumes in fractured sedimentary rock exhibits strong variability due to heterogeneity in source zone contaminant mass distributions, fracture network and matrix characteristics accompanied by temporal variability in groundwater flow. To measure the scale of these variabilities requires application of a specific combination of unconventional and conventional field and laboratory methods. The research activities involve the recent development and testing of several methods which complement the existing array of tools and techniques to advance the depth-discrete data sets collected using the DFN field approach (Figure 2). One major reason why so little is known about contaminant migration and fate in fractured sedimentary rock is that traditional research approaches involve only sampling water from the fractures. However, field studies using the rock core VOC analysis method show contaminant mass storage is dominated by the rock matrix rather than the fractures, and the contaminant concentrations in the fractures and the matrix are not in equilibrium (Hurley and Parker, 2002; Sterling et al., 2005; Parker et al., in review). This disequilibrium between fracture and matrix zones is evident in the rock core concentration profile from the California site shown in Figure 3. Therefore, sampling only the groundwater from the fractures cannot provide the overall mass distribution. Furthermore, when conventional boreholes are drilled, the water from a fracture in one section of the borehole migrates to another section of the borehole due to differences in head between the two sections. This creates an un-natural flow and contaminant transport condition within the system known as borehole cross-connection. This condition will also persist across the screened interval of a conventional monitoring Page 10 of 25

107 well; consequently, results from sampling the well do not reflect the natural system (Price and Williams, 1993; Sterling et al., 2005). Rock core analyses provide contaminant mass and phase distributions more relevant to contaminant behaviour than those obtained from monitoring wells or other types of borehole water sampling alone. The determination of the nature and extent of the contamination, with emphasis on elucidating the internal anatomy of contaminant plumes (including contaminant distribution in the rock matrix where groundwater is nearly immobile due to low permeability), is the foundation for understanding the processes governing the contaminant distribution. The rock-core based approach has several advantages over conventional methods for contaminant investigations in fractured sedimentary rock. For example, it provides a timeintegrated finger print of plume behaviour. In the rock matrix block, the extent of the halo evolving outward from each fracture can increase over several decades, depending on the duration of the dense, non-aqueous phase liquid (DNAPL) source. This allows the halo extent to be used as an indicator of the age of contamination (time since contaminant arrival) on a fracture by fracture basis. In contrast to analyses pertaining to the rock matrix, which generally has low permeability, groundwater sampling in the borehole using depth-discrete multi-level groundwater monitoring systems allows the current chemical concentrations in the hydraulically active factures to be determined and permits evaluation of plume variability over time. However, drilling and related borehole cross connection effects can influence the results of groundwater sampling. The rock core analysis method avoids this problem because the low permeability matrix is not easily cross connected during drilling and core retrieval prior to sample collection (Sterling et al., 2005). In addition, the rock core contaminant Page 11 of 25

108 analyses provide a direct measure of contaminant mass storage because the pore space in the rock matrix constitutes nearly the entire contaminant mass storage volume except when DNAPL persists in the fractures. However for the rock core analyses to show the actual mass distribution with useful accuracy, the samples must be collected from the core at closely spaced interval (Lawrence et al., 2006). CONTINUOUS HYDRAULIC CONDUCTIVITY PROFILING IN FRACTURED ROCK HOLES Various methods are available to acquire insight from boreholes concerning the locations and nature of fractures including borehole imaging, geophysical logging, hydrophysical logging, flow metering, hydraulic tests using single or double packers, and core logging. Neuman (2005) draws attention to the importance of assessing the numerous fractures in each hole potentially involved in groundwater flow rather than just the few fractures appearing to be the dominant ones in borehole observations. This implies that the identification of all potential flow features and the measurement of the flow capacity of each of those features are necessary for accurate characterization of flow in fractured rock. There are substantial limitations of existing methods for identifying transmissive borehole features. Borehole televiewing (optical, acoustic, or electrical) qualitatively identifies variable sized fractures but does not discern which of these are transmissive under natural or forced gradient conditions. Flow metering and profiles of fluid resistivity and temperature in open holes rely on open-hole cross connection gradients, which typically are dominated by the flow through the fractures with the largest flow capacity thereby masking the effects of Page 12 of 25

109 smaller fractures (e.g., Pehme et al 2007). Hydraulic straddle packer tests involving water injection or withdrawal measure the transmissivity of specific intervals that are isolated from the rest of the borehole by packers, and when these tests are done throughout the entire borehole using short test intervals, the locations of all major transmissive zones are known. However, to test an entire borehole in this manner is time intensive and therefore is typically very costly. A new hydraulic conductivity profiling, K profiling, method developed in conjunction with the DFN method is capable of identifying and measuring the K of all permeable features (fractures and fracture zones) in a borehole much more quickly than packer testing, thereby enabling the entire borehole to be tested at a much lower cost. The method utilizes a sock-like borehole liner formed of a flexible, water-tight, urethanecoated nylon fabric which is attached to the top of casing and filled with water in a manner to cause the liner to evert (the reverse of invert) down the borehole due to a constant excess hydraulic head inside the liner. The everting liner pushes water out of the borehole beneath it into the formation and the measured descent rate is governed by the bulk transmissivity of the features below the liner. The liner descent rate decreases each time the bottom of the everting liner passes and seals a transmissive feature. Each change in velocity identifies the depth of a permeable feature and the magnitude of each change is related to the transmissivity of the specific feature. The detailed transmissivity distribution is calculated from the liner velocity profile using the Thiem equation for radial flow. The profiling method is performed in holes 89mm (3.5 inches) to 280mm (11 inches) or larger and commonly takes less than a few hours making it much faster than conventional straddle packer testing or borehole flow metering when these are applied throughout the entire borehole. The utility of this method was demonstrated by profiling boreholes in a dolostone Page 13 of 25

110 aquifer where straddle packer tests were done for comparison. The sealing liner is usually left in the borehole after the K profiling episode to provide a temporary continuous seal of the hole minimizing cross connection. TEMPERATURE MEASUREMENT IN SEALED BOREHOLES FOR IDENTIFICATION OF ACTIVE FRACTURES The detection of all fractures in which significant flow occurs is a critical data requirement for the DFN approach. Furthermore, the capability of quantifying the amount of fracture flow in individual fractures over a range of several orders of magnitude is necessary. Various possibilities to acquire such data down-hole have been assessed and the most insightful method proved to be high-resolution temperature logging inside lined holes. The power of this approach is gained from two independent advances: 1) improved sensitivity to monitor temperature variability to o C (probe capability) under ambient temperature conditions and when heat is added (active line source) and its dissipation is monitored (Pehme et al., accepted 2006; Pehme et al., 2007a; 2007b); and 2) utilization of FLUTe liners to temporarily seal the borehole but provide access to static water column inside the liner, allowing measurement of ambient temperature distributions. When temperature logging is done in open bore holes (i.e., holes without liners), as is conventional practice, the vertical flow in the open hole typically swamps the minor but important, temperature signals and therefore larger number of hydraulically active fractures that likely occur in many holes are not identified. The temperature profiling research provides substantial supporting data concerning hydraulically active fractures at each of the four field sites (Pehme et al., Page 14 of 25

111 submitted). Most recently, an important advance in this temperature logging technology has been initiated involving a prototype for identifying hydraulically active fractures in lined boreholes to resolve flow direction and flow rate under natural flow conditions. SUMMARY AND CONCLUSIONS The development of the DFN approach has been in progress since 1997 and has achieved an advanced stage of application at four sites contaminated with organic contaminants that entered the fractured sedimentary rock decades ago as DNAPL. The approach now includes advanced methods for analysis of rock core volatile contaminant concentrations and primarily conventional methods for data acquisition from the cored holes and non-cored holes including: FLUTe K profiling, high resolution temperature profiling inside lined holes, and high resolution straddle packer testing. Conventional and advanced methods of borehole geophysics can contribute in important ways and flexible liners are used to minimize effects of borehole cross connection, in the use of boreholes. The DFN approach differs from the conventional approach for investigations of contaminated bedrock sites in that the time allocated to measurements in the open hole is minimal and emphasis is directed at measurements made while installing the flexible liners in the holes and inside the lined holes. Commercially available multilevel monitoring systems have been available for more than two decades, however in the DFN approach these systems are generally used with many more monitoring ports than conventional use. The DFN approach is used for site characterization focused on determination of the nature and extent of contamination and on development of a reliable site conceptual model. Page 15 of 25

112 Comprehensive characterization and a robust site conceptual model are necessary prerequisites for designing a reliable groundwater contamination monitoring system and predicting future plume behaviour. Although the DFN approach has attained a relatively comprehensive status since the development began more than a decade ago, substantial improvements and component additions are in progress. For example, completion of the relational data base system capable of efficient storage and manipulation of all of the diverse types of data obtained from the core and borehole measurements is necessary. In addition, continued development of methods for identifying individual fractures showing active groundwater flow and measuring their flow rates is needed. Also, downhole methods for measurement of contaminant flux need to be advanced. Cross hole borehole geophysical and hydraulic tests may also contribute useful information. Acknowledgements The development of the DFN approach has benefited continually from: the collaboration and encouragement of John Cherry, Distinguished Professor Emeritus, University of Waterloo (UW), the skillful field and modeling efforts of Steven Chapman, MSc Research Associate (UW) and the dedicated analytical laboratory efforts of Maria Gorecka, MSc Research Associate (UW). The rock core VOC analysis method has steadily advanced due to the collaboration with Tadeusz Górecki (Professor, Chemistry, UW). Innovations in borehole temperature logging were brought to the research program by Peeter Pehme and John Greenhouse (Professor Emeritus, UW) and insights on karst hydrology provided by Jérôme Perrin, post-doctoral fellow (UW). Physical property measurements including measurements of effective diffusion coefficients of rock core samples were performed in a laboratory supervised by Frank Barone (Golder Associates) and Edison Amirtharaj (UW) has done additional matrix property measurements. Robert Ingleton and Paul Johnson provided skilled technical assistance in the field at the Ontario field sites, field work at the California site benefits greatly from the assistance of MWH and Haley and Aldrich, and at the Wisconsin site by GeoTrans. Important advances were made by the following students who completed MSc theses within the research program: D.C. Austin, L.S. Burns, J.C. Hurley, J.R. Meyer, J.H. Plett, S.N. Sterling and C.M. Turner. Page 16 of 25

113 The research program is funded by the Natural Sciences and Engineering Research Council of Canada (NSERC), the University Consortium for Field-Focused Groundwater Contamination Research and the site owners: The Boeing Company, Hydrite Chemical Company, Syngenta Crop Protection Canada and Guelph Tool Inc. Also important are the collaborations with and technical assistance from the following groundwater technology companies: Westbay Instruments (Schlumberger), Flexible Liner Underground Technologies (FLUTe) and Solinst Canada; and most recently, additional support from R.J. Burnside & Associates, AquaResource Inc., GeoSyntec, City of Guelph, Regional Municipality of Waterloo and Schlumberger Canada. References Austin, D.C., Hydrogeologic controls on contaminant distribution within a multicomponent DNAPL zone in a sedimentary rock aquifer in south central Wisconsin. M.Sc. thesis, Department of Earth Sciences, University of Waterloo. Berkowitz, B Characterizing flow and transport in fractured geologic media: a review. Advances in Water Resources, 25(8): Britt, S.L., B.L. Parker and J.A. Cherry. Field testing the Snap Sampler TM A comparison with low flow, volume purging and the polyethylene diffusion bag sampler. For submission to Ground Water Monitoring and Remediation. Cherry, J.A., B.L. Parker and C. Keller, A new depth-discrete multilevel monitoring approach for fractured rock. Ground Water Monitoring and Remediation, 27(2): Cooke, M.L., J.A. Simo, C.A. Underwood, P. Rijken Mechanical stratigraphic controls on fracture patterns with carbonates and implications for groundwater flow. Sedimentary Geology, 184: Foster, S.S.D The chalk groundwater tritium anomaly a possible explanation. Journal of Hydrology, 25: Freeze, R.A. and J.A. Cherry, Ground Water, Englewood Cliffs, NJ: Prentice-Hall Inc. Garabedian, S.P., D.R. LeBlanc, L.W. Gelhar and M.A.Celia, Large-scale natural gradient tracer test in sand and gravel, Cape Cod, Massachusetts. II. Analysis of spatial moments for a nonreactive tracer. Water Resources Research WRERAQ, 27(5): Goldstein, K.J., A.R. Vitolins, D. Navon, B.L. Parker, S.W. Chapman, and G.A. Anderson Characterization and pilot-scale studies for chemical oxidation remediation of fractured shale. Remediation, 14(4): Graham Wall, B.R Influence of depositional setting and sedimentary fabric on mechanical layer evolution in carbonate aquifers. Sedimentary Geology, 184(3-4): Grisak, G.E. and J.F Pickens, Solute transport through fractured media: 1. The effect of matrix diffusion. Water Resources Research, 16(4): Hurley, J.C. and Parker, B.L Rock core investigation of DNAPL penetration and TCE mobility in fractured sandstone. In: Ground and Water: Theory to Practice, Proceedings of the 55th Canadian Geotechnical and 3rd Joint IAH-CNC and CGS Groundwater Specialty Conferences. Eds. Stolle D., A.R. Piggott and J.J. Crowder, Southern Ontario Section of the Canadian Geotechnical Society. Page 17 of 25

114 Keller, C., Liners and packers: Similarities and differences. Proc. Of EPA/NGWA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Sept , 2007, Portland, ME. Keller, C.E., J.A. Cherry and B.L. Parker. A new method for fracture identification and continuous hydraulic conductivity profiling in fractured rock boreholes. For submission to Groundwater Monitoring and Remediation. Lapcevic, P. A., K. S. Novakowski and E. A. Sudicky, Groundwater flow and solute transport in fractured media. In: The Handbook of Groundwater Engineering, J. W. Delleur ed., chap.17, Lawrence, A.R., PJ. Chilton, R.J. Barron and W.M. Thomas A method for determining organic solvents in chalk pore water (southern and east England) and its relevance to the evaluation of groundwater contamination. Journal of Contaminant Hydrogeology, 6: Lawrence, A., M. Stuart, C. Cheney, N. Jones and R. Moss, Investigating the scale of structural controls on chlorinated hydrocarbon distributions in the fractured-porous unsaturated zone of a sandstone aquifer in the UK. Hydrogeology Journal, 14: Lipson, D.S., B.H. Kueper and M.J. Gefell, Matrix diffusion-derived plume attenuation in fractured bedrock. Ground Water, 43(1): Meyer, J.R., B.L. Parker, and J.A. Cherry. Use of detailed hydraulic head versus depth profiles for defining hydrogeologic units in flat laying fractured sedimentary rock. Submitted to Journal of Environmental Geology. Molson, J.W., E.O. Frind, HEATFLOW, A 3D groundwater flow and thermal energy/mass transport model for porous and fractured porous media, version 2.0, University of Waterloo. Molson, J.W., P. Pehme, J.A. Cherry and B.L. Parker, Numerical analysis of heat transport within fractured sedimentary rock: Implications for temperature probes. Proc. Of EPA/NGWA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Sept , 2007, Portland, ME. Neuman, S.P Trends, prospects and challenges in quantifying flow and transport through fractured rocks. Hydrogeology Journal, 13: Pankow, J.F. and Cherry, J.A. (eds.) Dense Chlorinated Solvents and other DNAPLs in Groundwater: History, behaviour, and remediation. Waterloo Press, Guelph, Ont., 522 p. Parker, B.L., L.S. Burns, C.M. Turner and J.A. Cherry. DNAPL origin for deep herbicide and TCE contamination in a dolostone aquifer. Journal of Contaminant Hydrology, Submitted Parker, B.L., D.B. McWhorter and J.A. Cherry, Diffusive loss of non-aqueous phase organic solvents from idealized fracture networks in geologic media. Ground Water, 35(6): Parker, B.L., R.W. Gillham and J.A. Cherry, Diffusive disappearance of immiscible phase organic liquids in fractured geologic media. Ground Water, 32(5): Pehme, P.E., J.P. Greenhouse and B.L. Parker. The active line source temperature logging technique and its application in fractured rock hydrology. Journal of Environmental & Engineering Geophysics, Accepted Pehme, P., J. Greenhouse and B.L. Parker, The Active Line Source (ALS) technique, a method to improve detection of hydraulically active fractures and estimate rock thermal Page 18 of 25

115 conductivity. Proceedings, 60th Canadian Geotechnical Conf. & 8th Joint CGS/IAH-CNC Groundwater Conf, Oct , Ottawa, ON. Pehme, P., B.L. Parker and J.A. Cherry, The potential for compromised interpretations when based on open borehole geophysical data in fractured rock. Proc. Of EPA/NGWA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Sept , 2007, Portland, ME. Pehme, P., B.L. Parker and J.A. Cherry. Improved Resolution of Ambient Flow Through Fractured Rock 1 with Temperature Logs. In press in Groundwater. Pierce, A.A., Isotopic and hydrogeochemical investigation of major ion origin and trichloroethene degradation in fractured sandstone. M.Sc. thesis, Department of Earth Science, University of Waterloo. Pierce, A.A., B.L. Parker, R. Aravena, J.A. Cherry. Field evidence for TCE degradation mechanisms in fractured sandstone. For submission to Journal of Contaminant Hydrology. Price, M. and A. Williams, The influence of unlined boreholes on groundwater chemistry: A comparative study using pore-water extraction and packer sampling. Journal of the Institute of Water and Environmental Management, 7(6): Sara, M.N, Site Assessment and Remediation Handbook. 2nd Edition. New York: Lewis Publishers Smith, L., and F.W. Schwartz An analysis of the influence of fracture geometry on mass transport in fractured media. Water Resources Research, 20(9): Smith, L., and F.W. Schwartz Solute transport through fracture networks. In Flow and Contaminant Transport in Fractured Rocks, J. Bear, C. F. Tsang, and G. de Marsily, eds. New York: Academic Press. Sterling, S.N., B.L. Parker, J.A. Cherry, J.H. Williams, J.W. Lane Jr., and F.P. Haeni, Vertical cross contamination of trichloroethylene in a borehole in fractured sandstone. Ground Water, 43(4): Sudicky, E.A., A natural gradient experiment on solute transport in a sand aquifer: Spatial variability of hydraulic conductivity and its role in the dispersion process. Water Resources Research, 22(13): Sudicky, E.A., and R.G. McLaren The Laplace transform Galerkin technique for large scale simulation of mass transport in discretely fractured porous formations. Water Resources Research, 28(2): Therrien, R. and E. Sudicky A three dimensional analysis of variably-saturated flow and soluble transport in discretely fractured porous media. Journal of Contaminant Hydrology, 23, (1-2): U.S. National Research Council Rock Fractures and Fluid Flow: Contemporary Understanding and Applications. National Academy of Sciences, 551 pp. (see chap. 6, Field scale flow and transport models: p ) Vanderkwaak, J.E. and E.A. Sudicky Dissolution of non-aqueous phase and aqueous phase contaminant transport in discretely fractured porous media. Journal Contaminant Hydrology, 23(1-2), Page 19 of 25

116 Table 1: Summary of contaminant types, site hydrogeology and causes of the contamination at the four field research sites. Entered Water table Degradation Overburden ground depth and Cause of Field Site/ Major Parent Products (in Thickness Rock Type water Max. cont. Contamination/ Owner Chemicals order of and Type (main depth (meters Comments abundance) (meters) period) bgs) Cambridge Ontario Guelph, Ontario Dolostone aquifer Metolachlor, on shale aquitard; TCE flat lying Dolostone aquifer on shale aquitard; flat lying TCE, minor PCE Sandstone with Simi, TCE, minor siltstone, shale California TCA interbeds; 30 dip Wisconsin Sandstone and minor dolostone with minor siltstone; flat lying PCE, TCE, TCA, Ketones None cis-dce, VC cis-dce, 1,1-DCE, t-dce, VC cis-dce, 1,1-DCA, 1,1-DCE, VC s 1950s- 1960s 1950s- 1960s 20m 150 m into shale 3-4 m 50 m but may be deeper m >300 m above grade- 25 m Nearly all mass shallower than 60 m Glacial; sand and silt, and thin basal till 3-5 Till 0-5 Alluvium 7-40 Glacial sand, silt and clay layers Agricultural chemical packaging; no DNAPL found; metolachlor plume goes to a municipal well; below MCLs; Auto-parts manufacturing; small lateral plume extent expected; no DNAPL found Rocket engine testing, research; many plumes from many different source areas; no DNAPL found NE area focus Solvent recycling; plume extends ~3km from source zone; 35,000L DNAPL pumped out and residual DNAPL remains Page 20 of 25

117 Figure 1: Schematic diagram illustrating use of rock core contaminant analyses to identify migration pathways by identification of diffusion haloes associated with active fractures Page 21 of 25

118 Figure 2 Summary of DFN Approach Page 22 of 25

119 Figure 3: Example of rock core analysis results for TCE in sandstone at a location near TCE DNAPL source zone at the California site. All analyses are much below TCE solubility indicating lack of DNAPL presence. (modified from Sterling et al. 2005) Page 23 of 25

120 Figure 4: The discrete fracture network (DFN) approach for investigating contaminated sites on fractured sedimentary rock, includes intensive data acquisition from contaminated cores and from the corehole. Open hole conditions are minimized. Illustration of conceptual stages in the time evolution of source zone and plume at chlorinated solvent DNAPL sites on fractured porous sedimentary rock: a) DNAPL flows in fracture network and begins to dissolve and diffuse into rock matrix. DNAPL flow ceases soon after DNAPL input to the rock ceases. b) All DNAPL mass has dissolved completely and the contaminant mass now exists almost entirely in the rock matrix as dissolved and sorbed mass due to diffusion driven mass transfer. Therefore, the source zone no longer has DNAPL and there is not distinct difference in contaminant state between the zone initially referred to as the source zone and the plume. c) Groundwater flow through the initial DNAPL source zone has caused complete mass translocation from much of the initial source zone into the downgradient plume ; the plume front is migrating only slowly or is stable or shrinking due to the combined effects of matrix diffusion and degradation. Page 24 of 25

121 Figure 5: Example of DFN simulations using FRACTRAN (2-D numerical model; cross section display) of TCE plume in fractured sandstone with fracture and matrix properties consistent with those of the California site: a) source location on fracture network domain; aperture distribution shown, b) TCE plume (no degradation) after 50 years, and c) stylistic comparison of rock core TCE profiles 75m from source at 50 years. Page 25 of 25

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139 A New Depth-Discrete Multilevel Monitoring Approach for Fractured Rock by John A. Cherry, Beth L. Parker, and Carl Keller Abstract A new approach for monitoring in fractured rock was demonstrated in a contaminated (trichloroethylene and metolachlor) dolostone aquifer used for municipal water supply. The system consists of two related technologies: a continuous packer for temporary borehole seals (Flexible Liner Underground Technologies Ltd. [FLUTe] blank liner) and a depthdiscrete multilevel monitoring system (MLS) (the Water FLUTe) for temporary or permanent monitoring. The continuous borehole liner consists of a urethane-coated nylon fabric tube custom sized to each hole. The FLUTe MLS consists of the same liner material with many depth-discrete intervals for monitoring hydraulic head and/or ground water quality. The FLUTe blank liner seals the entire borehole, prior to FLUTe multilevel installation, to prevent vertical cross connection while allowing borehole logging and testing. The FLUTe multilevel system also seals the entire borehole with the exception of each monitoring interval where the formation water has direct hydraulic connection to the pumping system via a thin permeable mesh sandwiched between the liner and the formation. The blank sealing liners and the multilevel systems were used in five boreholes ranging in diameter between 9.6 and 14.5 cm in the dolostone aquifer to depths of 150 m. The systems were custom designed for each borehole and included between 12 and 15 monitoring intervals. The application demonstrated the ease of installation and removability and facilitated obtaining large data sets with minimal labor. The system offers unique and versatile design features not possible with other bedrock monitoring devices and has been used at many bedrock contamination sites across North America. Introduction Depth-discrete multilevel monitoring in a borehole is the use of an engineered system for measurement of hydraulic head and/or ground water sampling at several or many depth-discrete intervals. A seal is positioned above and below each monitoring interval to prevent hydraulic connection between intervals. The goal of such multilevel monitoring is to achieve, using a single hole, what would otherwise be accomplished with a cluster of monitoring wells each completed to a different depth and each installed in a separate borehole. Multilevel monitoring minimizes formation disturbance by drilling fewer holes and decreases cost per monitoring zone. Depth-discrete multilevel monitoring, henceforth termed multilevel monitoring in this article, has a long history in contaminant hydrogeology, starting with systems used in studies of ground water contamination in permeable unconsolidated deposits (Merritt and Parsons 1960; Pickens et al. 1978; Cherry et al. 1983). Copyright ª The Author(s) Journal compilation ª 2007 National Ground Water Association. Multilevel monitoring systems (MLSs) are particularly advantageous for investigations in fractured rock because the need to maximize the quantity and diversity of data acquired from each hole is driven by high drilling costs. Also, the expectation of fracture network complexity drives a need for data from many depths in each hole. In the late 1970s, Westbay Instruments Inc. ( developed the first commercially available MLS suitable for fractured rock. The utility of this system, described by Black et al. (1986), has been well established by many fractured rock applications around the globe. Cherry and Johnson (1982) developed a second type of MLS for fractured rock, which was subsequently redesigned and made commercially available globally by Solinst Canada Ltd. in the late 1980s as the Waterloo-Solinst system (Pianosi and Weaver 1991; Dunnicliff 1988; This article describes a new approach for monitoring contaminated sites in fractured rocks. This approach comprises two related technologies: the first is a continuous flexible packer (i.e., borehole liner) used temporarily for complete borehole sealing to prevent borehole cross connection prior to installation of the MLS, and the second, which incorporates attachments to the liner, is a removable MLS for temporary or long-term monitoring of hydraulic Ground Water Monitoring & Remediation 27, no. 2/ Spring 2007/pages

140 head and/or water quality. This new system is different than the Waterloo-Solinst and Westbay systems, allowing it to accomplish multilevel monitoring in ways and circumstances not otherwise possible. This approach was initially developed by Flexible Liner Underground Technologies Ltd. (FLUTe) in the late 1990s and evolved to its present design by refinements based on experience at many field sites ( The design and operation of the borehole liner and FLUTe MLS is presented herein along with a description of their performance at a site where trichloroethylene (TCE) and a pesticide, metolachlor (MET), occur in a fractured dolostone aquifer used for municipal water supply in Cambridge, Ontario. The precursor for the FLUTe ground water method was a patented version of the flexible liner technology known as the SEAMISTÔ system developed by Carl Keller for vadose zone monitoring beginning in The FLUTe method has been used at many fractured rock sites; however, the technology was pushed to its limit at the Cambridge, Ontario site, which resulted in major design improvements in response to the field experience. At this site, exceptionally large data sets were produced by taking full advantage of the capabilities of the method. The Cambridge site is used by the University of Waterloo to develop and assess methods for investigating ground water contamination in both overburden and bedrock. At the same site, a modified version of the Waterloo-Solinst system is being used for investigations in the overburden (Parker et al. 2006). An early version of the FLUTe system was first used in the dolostone at the site in 2000 to An improved version was used beginning in 2002, and a version with additional improvements was used beginning in By 2005, five holes were equipped with FLUTe multilevel systems. System Components and Installation Procedure Each component of the FLUTe multilevel system, also called the Water FLUTe Ò system, is attached to the borehole liner, making it the fundamental component of the multilevel system. The liner is made from impermeable, tubular, and flexible nylon fabric slightly larger than the borehole and it extends from the top to the bottom of the hole. The liner is inflated when the water level inside the liner is positioned a few meters above the hydraulic head in the zone of highest head in the hole. On its own, without any attachments, the blank liner performs as a complete seal like a continuous packer in the hole. When used in this manner, the blank liner prevents hydraulic cross connection in the hole. When ports for water sampling and/or hydraulic head monitoring are attached to the liner, the system performs as a depth-discrete MLS. The procedures for installation and removal are the same for both the blank liner and the MLS. The blank liner and also the multilevel system are installed using a procedure known as eversion: a process whereby the tubular liner turns from being inside out to being right side out as it descends down the hole. Figure 1 shows the four main stages in the installation procedure. In the first stage (Figure 1A), the reel on which the liner is shipped from the manufacturing facility to the field site is positioned close to the surface casing onto which the top of the liner is clamped. The surface casing normally extends through the overburden and/or rock rubble into competent rock. In this first stage, the liner is pushed by hand down into the casing (~1 m) to form an annular pocket in which water is then added to drive the liner farther down the hole (Figure 1A). Addition of water into this initial handformed pocket causes continuous propagation of the liner down the hole. A) Liner is clamped on and installation begins B) Liner begins advancing below standing water in hole C) Liner is halfway down the hole D) Liner is completely everted Inverted liner Water Clamp Surface Casing Reel Inverted liner Water Reel Water To reel Inverted liner Excess headin liner Tether tied to liner bottom Open borehole static water level Liner Open fractures Open hole Liner installation sequence Figure 1. Stages in installation of blank liner: (A) liner is clamped to casing head and pushed by hand a few feet down the casing before water is added, (B) the liner goes below the standing water level in the hole, which pushes water into the formation, (C) the liner is halfway down the hole at which point only the tether remains on the reel, and (D) the liner is completely everted in the hole and the water level inside the liner is raised a few meters or more above the static water level measured in the open hole prior to liner installation; this excess head causes inflation of the liner to seal the borehole. 58 J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2: 57 70

141 In the second stage, continued addition of water causes the liner to descend below the standing water level in the hole, and as it goes deeper, the liner causes expulsion of water from the hole into the formation (i.e., into fractures; Figure 1B). In effect, the installation procedure is a progressive large-scale slug test because at each point in time, the rate of liner descent is governed by the transmissivity of the open-hole segment below the bottom of the liner. The length of this open-hole segment decreases as liner installation proceeds, and if the hole has substantial transmissivity all of the way to the bottom, installation of the liner is typically completed within 1 to 2 h. However, if the hole has low transmissivity throughout or in the bottom parts, the installation can take many hours or even several days. This longer time is usually not detrimental because only minimal labor is required during periods of slow liner descent, which involves periodic checking and perhaps topping off the water level inside the liner. If faster installation is desired, a water removal tube, typically 1- to 2-cm inner diameter (ID), is inserted deep in the hole before liner installation starts. With this tube in place, liner descent is accelerated by pumping water from this tube situated between the liner and the borehole wall. The tube is withdrawn after the liner reaches the bottom of the hole. To allow tube removal, the water level inside the liner is lowered to cause liner deflation and then the water level is raised to reinflate the liner after the tube is withdrawn. The halfway point of the liner installation is achieved when the bottom end of the liner (i.e., the tubular roll) is completely off the reel and the bottom end of the liner is positioned at the top of the casing (Figure 1C). At this halfway point, only the tether line remains on the reel. As the liner continues to go down the hole, the tether rolls off the reel and also goes down inside the liner. The main purpose of the tether is for removal of the liner from the hole, if removal is desired later. The liner installed in each hole is generally constructed to the hole length; and therefore, when the bottom of the liner reaches the hole bottom, it is fully everted (Figure 1D). Because the liner is everted in the hole rather than just lowered down the hole, no part of the outside of the liner contacts the borehole wall until it everts. Therefore, there is no rubbing or scraping along the borehole wall. Rubbing or scraping is also avoided when the liner is removed because removal is just the opposite of the installation process, except that water is pumped from the liner interior while tension is applied to the tether. The MLS is constructed by creating attachments outside and inside the liner to allow formation water from depth-discrete segments to pass through the liner into pump systems. Specifically, monitoring intervals are formed by bonding different materials to the liner. For simplicity, Figure 2 shows a single monitoring interval with a double check valve pump system for sampling the interval. In this version of the system, the hydraulic head in the monitoring interval is determined manually using a standard water level tagline. In another version of the system, a pressure transducer is positioned on the port-pump tube, just below the first check valve, with the cable extending to ground surface inside the liner. Each monitoring interval is formed by a thin flexible spacer (Figure 3), attached by Surface Casing Pump quick connect (Single port system shown for clarity) Tether support of tubing bundle Formation head in pump Spacer defining monitoring interval Sealing liner Port behind spacer thru liner Sample tube (0.17 in ID) Pump tube (½ in ID) Second check valve Bottom of the U First check valve Port to pump tube (0.17 in ID) Figure 2. The port and pump system in the FLUTe multilevel system; for clarity, only one port and pump system are shown; however, actual FLUTe systems contain multiple ports. In the system illustrated here, hydraulic head is determined by lowering a conventional water level probe down the pump tube. Or, in an alternative design, a dedicated pressure transducer is connected to each port. heat welding on the outer surface of the liner, to create a segment of the borehole not sealed by the liner. The spacer is constructed using a permeable mesh cut to the desired length of the monitoring interval and that runs continuously around the circumference of the borehole. The mesh forms a very pervious but thin (1- to 2-mm) annulus between the liner and the hole wall, and it intersects all ground water flow paths that encounter the borehole wall in the interval. Figure 3 shows the perforated tube that conducts the formation water from those flow paths to the port through the liner. A thin outer filter fabric composed of finely woven nylon or polyester is stitched to the spacer to prevent particles of 200 lm orlarger(e.g.,sandorcoarsesilt)from going through the port into the pumping system. Because the combined thickness of the spacer composite is so small (less than 2 mm), the spacer has almost no storage volume, and therefore, minimal purging is required to draw fresh formation water into the plumbing in the interior of the liner. Figure 4 shows the port with a dedicated gas-driven, double-valve pump. The port behind the permeable spacer connects through the liner to a tube called the port tube that descends to the bottom of the liner in an interior sleeve of the liner. The port tube rises up to the bottom end of the J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2:

142 A Side cut view Central tubing bundle Formation Spacer material between liner and filter fabric Filter fabric on outer surface of spacer Gas bottle 3 way valve Sample container Gas/water interface before purging stroke Liner Pump tube Monitoring interval Perforated tube in spacer material B Water filled liner Cross cut view Feed-through sealing tube through liner Port-to-pumptube Sleeve on liner containing tube to pump Liner Spacer Filter fabric Port tube Buffer against aeration Sample tube Gas/water interface at end of sample stroke Bottom of the U First check valve (closed) Liner water fill Port-to-pump tube in spacer Central tubing bundle containing pumps, etc.. Figure 3. Details of the spacer design and other aspects of the monitoring port: (A) vertical cross section through centerline and (B) horizontal cross section. Figure 4. Procedure for purging and sampling the FLUTe gas-driven, double-valve pump system. The pump system yields water when gas pressure is applied to the pump tube (large diameter), which drives the water in the pump tube to surface via the sample tube (smaller diameter). During this water yield period, the first check valve is closed and the second one is open. When the gas pressure application stops, the flow stops and the second check valve closes, causing storage of water in the sampling tube until the next application of gas pressure. pump at the first check valve. The port tube does not go directly up from the port to ground surface but rather goes down to the bottom of the liner and then up to the pump. This check-valve pump design is unique to the FLUTe system when compared to other gas-driven, check-valve pumps used in other multilevel systems and monitoring wells. The advantage of this down-then-up configuration is that, regardless of the port elevation, the pump length is the same for each port. Therefore, the pumping volume per unit stroke is much larger than that produced by other types of dedicated gas-driven downhole pumps. The FLUTe pumping system allows simultaneous purging of FLUTe pumps connected to several ports. The water flowing from the spacer enters the pump tubes through the first check valve and fills the pumping system to the level of the hydraulic head in the formation at each spacer. The check valve has a Teflon Ò ball, but no spring, to provide the minimum impedance to the equilibration of the water level in the pump tube with the water level in the formation after purging. This equilibration allows for the manual water level measurements of hydraulic head. Thus, calibration of pressure transducers attached to each port can be checked after installation, or the system can be used without pressure transducers. The pump system consists of the two tubes to the surface forming a very long U shape (Figure 4). The large tube, called the pump tube, is typically 0.5-in (1.27-cm) ID. The other leg of the U is a smaller diameter tube, typically 0.17-in (0.43-cm) ID, referred to as the sample tube. A second check valve is often included in the sample tube near the bottom of the pump U bend. The second check valve usually contains a stainless steel ball without a spring. The primary advantage of the second valve is the improved pumping efficiency by preventing backflow into the pump tube as it is refilling with water. The second valve is also useful for relatively shallow holes and for use with flow-through devices where water quality monitoring occurs at ground surface prior to sample collection. The disadvantage of the second check valve is that it increases the complexity of the pumping system and it requires that the system be purged of water prior to obtaining a manual water level measurement. The pump tube systems for the Water FLUTe ports are connected and supported with KellumÔ grips attached to the central tether. The entire tubing bundle is wrapped in a diagonally woven sheath. In this compact, snakelike form, it passes easily into the liner during the installation procedure. A 1.27-cm-ID tagline tube is included in the tubing bundle to facilitate manual measurement of the water level inside the liner. The entire tubing system from the spacer to the surface is generally made of polyvinylidene fluoride (PVDF) tubing but other tubing, such as nylon, has been used on some systems. All systems produced before 2002 used nylon tubing. The fittings are usually brass unless stainless steel or another material is specified. 60 J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2: 57 70

143 Liner Provides Borehole Seal The liner is pressed tightly against the borehole wall when the water level inside the liner is kept a few meters or more above the highest head in the formation. The liner is custom built to the diameter specified for each hole and is generally sized slightly larger (i.e., 3 to 4 mm) than the nominal hole diameter so that the liner, when inflated, can easily expand into borehole wall irregularities. Borehole televiewer images taken inside a liner at the Cambridge site and other sites show that the liner conforms to wall irregularities, as shown in Figure 5. The video snapshot shown in Figure 5A is the borehole wall at approximately 53-m depth without the liner. The second snapshot is with the liner at about the same elevation (Figure 5B). The slight elasticity of the liner and the eversion process of installation allow the liner to conform very well to the irregularities of the borehole wall as evidenced by the similarities of the two photographs. Some boreholes have enlargements or cavities that are too large for the liner to fill by expansion. While the zones themselves are not sealed by the liner, seals do form at the necks above and below these zones (Figure 6), preventing the enlargements or cavities from jeopardizing the overall integrity of the borehole seal. In the cavities or enlargements, the liner is not in contact with the borehole wall, i.e., unsupported liner. An essential property of the liner is that it is sufficiently strong to prevent rupture in those zones where the borehole wall has a much larger diameter than the liner. Liner burst tests conducted in the laboratory show that the typical unsupported liner can withstand an excess internal pressure up to 448 kpa (~45 m of head differential between inside and outside the liner). Though, failure pressure is inversely proportional to the diameter of the borehole. The capability of the liner for crossing cavities makes both the blank liner and FLUTe multilevel system suitable for use in karst. Reinforced sections of the liner or a stronger fabric can be used where video logs indicate large voids or sharp ledges that could cut the fabric. If the depth to water in the borehole is too close to ground surface to allow the recommended 3-m differential for normal inflation, the interior water level can be increased by extending the liner as a column above ground surface or a weighted bentonite slurry can be used instead of water to fill the liner. Laboratory tests show that a 3-m head differential provides sufficient seal for the typical range of borehole diameters (< 10 inches) and hazards, such as sharp edges, without overloading the liner. The heavy bentonite slurry fill was used in 13 liners in 2004 and 2005 to deal with shallow water table or artesian conditions. The liner is still removable where the bentonite slurry fill has been used. Application of a head differential within the range recommended previously is not always sufficient to achieve a complete seal in each interval between monitoring ports because, in some situations, hydraulic head much higher than the open-hole static level can occur at particular zones. The open-hole water level is a blended water level dependent on the hydraulic conductivity and hydraulic head distribution throughout the hole. Prior to installation of the multilevel system, the depth of highest head in the hole and the magnitude of this head are unknown. However, after the multilevel system is installed and head measurements are obtained, the presence of zones of higher head can generally be deduced from the head profile, and higher inflation pressure can then be applied incrementally until the head differentials for each of the ports are approximately 3 m. In the case of hydraulic testing where large drawdowns are created by pumping nearby wells, the potential for seals to be compromised and for liner ruptures to occur is increased and needs to be heeded during designs of such tests. Water Sampling Using the Double-Valve Pump The most commonly used version of the FLUTe multilevel system has four or more monitoring ports with a double-valve pump attached to each port. The largest number of double-valve pumps used in a FLUTe multilevel system is 15 at the Cambridge site. For simplicity, Figure 2 shows a system with only one port. In FLUTe multilevel systems using dedicated pumps for multiple ports, the pump system of Figure 2 is replicated in the liner interior to match the number of ports. At sites where the formation hydraulic head is close enough to ground surface for suction pumps (e.g., peristaltic pumps) to be effective, the double-valve pumps are not required but provide a more convenient sampling method. The double-valve pump operates as follows. After the pump tube has filled with water, a pressure source such as Figure 5. Borehole television images taken in a hole in the dolostone aquifer at the Cambridge site: (A) image at a depth of approximately 53 m bgs without the FLUTe liner and (B) image inside the liner at approximately the same depth. Both images clearly show irregularities on the borehole wall, indicating that the liner conforms to the irregular wall surface (liner expands slightly to press into the borehole wall depressions). J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2:

144 Figure 6. Illustration of the behavior of a FLUTe liner during installation; as the liner descends through a breakout, it expands but cannot expand sufficiently to press against the breakout wall. The liner is then unsupported in this interval, but once the liner goes below the breakout, it once again presses against the entire circumference of the borehole. E1 ¼ liner in normal size hole, E2 ¼ liner expands where hole is larger, E3 ¼ liner cannot expand to fill breakout, so liner is unsupported but does not break unless excess head is applied, E4 ¼ liner entering normal size hole, and E5 ¼ liner proceeds down normal size hole. a nitrogen tank (Figure 4) is connected to the top end of the pump tube. A purge pressure is applied from the standard nitrogen tank to the gas/water interface at the standing water surface in the pump tube. The applied purge pressure is great enough to force the interface to the bottom of the U and up the sample tube to the surface, thereby expelling essentially all of the water from the entire pumping system for that port. Only residual droplets are left in the sample tube. Then the pressure is dropped in the pump tube and the system is allowed to refill with formation water from the spacer via the port tube. The residual droplets are swept up by the recharge of the pump and sample tubes. The volume of the pump tube is usually far greater than that of the port tube plus the interstitial space in the spacer. Hence, upon the first recharge of the pump, some formation water adds to the fill of the pump tube through the first check valve during this first recharge. A second application of the purge pressure, referred to as a stroke, forces all of the water out of the pumping system again. This second purge stroke volume is discarded along with the first purge volume. The pressure in the pump is dropped again to atmospheric pressure, allowing the pump tube to refill for the second time. This second recharge volume is essentially all formation water. In the next stage, called the sampling stroke, the pressure at the gas source is lowered to the sampling pressure. This pressure is prescribed for each Water FLUTe system and is sufficiently low that the gas/water interface can only be driven down to within 6 to 10 m of the bottom of the U bend. As the water for this stroke is driven out of the sample tube at the surface, a recommended volume of the first flow is discarded because it may contain aerated droplets from the last purge stroke. The sample is collected from the subsequent flow. As the interface in the pump tube approaches its lowest level, the flow rate from the sample tube slows dramatically for easy sample collection. The typical continuous flow volume for a single stroke from which the sample can be obtained exceeds 6 L, including 0.5 to 1 L for the discard, depending on the depth of the borehole and water table. The water sampled using this procedure essentially comes from the lower three quarters of the pump tube. If more sample water is desired, the pump pressure is dropped for another recharge. When applying the sample pressure again, there is no need to discard the first flow. The sample cycle can be repeated until no more sample water is desired or until the pump has extracted a sufficient volume of water to collect a sample from the desired distance away from the borehole. If desired, the flow from the sample tube can be directed under near steady-state conditions through a chamber with probes for measurement of chemical parameters such as ph, dissolved oxygen, and electrical conductance as commonly done during low-flow sampling. Water from the two initial purge strokes should not be used because it includes water that may have resided in the pump tubing for long periods of time and could have contaminant concentrations nonrepresentative of the formation water. A distinct advantage of the FLUTe system compared to all other multilevel systems is that all ports can be purged and sampled at the same time with ease. Due to the U-tube design, the tubing length and purge volumes are essentially the same regardless of port elevation except for the hydraulic head differences between ports. Nevertheless, the tube systems can be purged and sampled with the same purge pressure and sample pressure requiring about the same time to sample all the ports simultaneously. In other words, it takes nearly as much time to purge and sample one port as it does to purge and sample all the ports in the system. Five to 15 minutes of continuous flow is available depending on how rapidly the pressure is applied during one sample stroke. By purging all ports at the same time, the hydraulic heads in all ports are disturbed by a similar amount minimizing the accentuation of transient vertical gradients and resultant vertical redistribution of water in the formation due to sampling. The tubing diameters and simple geometry make the system resistant to clogging even for sites with high turbidity. Application of the FLUTe Method to the Cambridge Site The field site is located in an industrial part of Cambridge, Ontario (Cambridge site), where pesticide contamination, MET, in the overburden and the bedrock (Silurian dolostone) was discovered in 1993 beneath an agrochemical packaging and distribution facility. A network of conventional 62 J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2: 57 70

145 monitoring wells was established in the overburden and in the bedrock during a major investigation in 1993 to 1995 (Carter et al. 1995). All of these wells are in the upper half of the 94.5-m-thick dolostone unit. During this initial investigation, TCE was also discovered in the ground water. A follow-up study showed that the source of the TCE is a nearby industrial site. The investigations reported here, which began in 1999, focus on monitoring in the dolostone. In the current study in which multilevel monitoring is the thrust, the goal is to monitor ground water quality from top to bottom in the dolostone aquifer and also in the upper part of the underlying shale aquitard. The previous study showed that MET and TCE contamination is present in the upper half of the dolostone (Carter et al. 1995) and no information was obtained from greater depth. The study area is situated in the vicinity of several municipal wells that supply water to the city of Cambridge. Therefore, comprehensive knowledge of the spatial distribution of chemical constituents relevant to water quality in the aquifer is needed. Five Water FLUTe systems have been installed in the bedrock to date and more installations are anticipated. In each hole in which a Water FLUTe has been installed, a FLUTe liner was used prior to the installation to minimize borehole short-circuiting and to allow high-resolution temperature logging inside the liner. Borehole logging inside FLUTe liners offers opportunities for characterization of natural-gradient flow conditions without the openhole dominating flow conditions. Furthermore, it ensures an open stable hole for logging with downhole tools, but this must be done with care to avoid entanglement with the tether line. This is particularly important to protect against potential downhole loss of active radioactive sources associated with neutron and gamma-gamma logs. The liner system does not interfere with the use of high-resolution temperature and natural gamma tools, and prospects for application of the full wave form of acoustic televiewer appear to be good. In addition, transparent liner material has been used to facilitate video logging of borehole walls. At the Cambridge site, the liners were removed on occasions when borehole geophysics, straddle-packer testing, and borehole-flow metering were conducted but reinstalled to minimize the cross connection due to open borehole flow. Thus, comprehensive data sets pertaining to lithologic and ground water flow conditions were used to custom design the Water FLUTe system for each hole. The resultant FLUTe depth-discrete monitoring systems were used to determine the spatial distributions of a wide variety of chemical constituents, including chloride from road salt, nitrate from sewage, TCE, and MET. Each contaminant has a different source location and input condition, resulting in different contaminant concentration distributions. There is no basis for presupposing where the highest concentrations of each contaminant will likely occur in any hole. Therefore, achieving a maximum number of monitoring intervals in each hole is desirable. The length and position of each port, and hence seals, should be based on geologic and hydrologic information obtained from each hole (i.e., core descriptions, geophysical, and/or hydrogeophysical logs that provide information on lithology, fracture, and flow distributions). The diameter of the borehole is the factor limiting the maximum number of monitoring intervals that can be accommodated in the FLUTe system. In this investigation, the first hole was drilled using an air-rotary, water-well rig. The borehole had a nominal diameter of 14.3 cm (5.63 inches). The Water FLUTe installed in this hole contained 15 dual-purpose monitoring intervals. A nitrogen drive, double-valve pump, and a pressure transducer (Solinst /Geokon 4500H, Georgetown, Ontario, Canada) were attached to each of the 15 ports. This first hole extended to a depth of 100 m below ground surface (bgs) (70 m below top of rock), and the 15 monitoring intervals were distributed over the depth interval between 9 and 73 m below top of rock. Figure 7 shows the multilevel system coming off the reel and going down the hole and a view of the wellhead after the system was fully installed and ready for use. From the sampling of this Water FLUTe system and rock core contaminant analysis performed in 2000 to 2001, it was found that MET and TCE contamination occurred at all monitoring depths. It was then decided to obtain borehole logging data and rock core analyses in the bottom part of the dolostone and in the upper part of the underlying shale. Therefore, the FLUTe system was removed after more than a year of monitoring. The deeper drilling was done in 2003, after which the reconstructed Water FLUTe system with the same configuration of ports was replaced in the hole to continue collection of temporal data. The replacement Water FLUTe used the Figure 7. Photographs taken at the first installation at the Cambridge site in 2000 showing (A) the FLUTe multilevel system coming off the reel and going down the hole, while water from a hose is added to the system and (B) the wellhead unit for monitoring the 15 ports set up after installation of the multilevel system. J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2:

146 pumps taken from the removed system and new pressure transducers (In Situ PXD-261, Fort Collins, CO). This experience demonstrated easy installation and removal of the FLUTe system with the maximum practical number of dual-purpose monitoring intervals (i.e., maximum number of ports is 15 for 14.3-cm-diameter hole). Although there is space remaining inside the liner and tubing bundle, a larger number would strongly diminish prospects for successful installation and removal. However, a larger number of monitoring intervals could be achieved if each port were not dual purpose. For example, if fewer pressure transducers are used, more pumping ports can be included. Four additional holes were drilled at the Cambridge site. The first of these was drilled using a PQ coring system producing a 5.0-in (12.7-cm) borehole. The FLUTe system incorporated 10 dual-purpose monitoring intervals (sample pump and pressure transducer [In Situ PXP-261]). The remaining three holes were HQ coreholes (3.78-in [9.6-cm] diameter). To achieve the most comprehensive monitoring configuration specific to the needs of the site conditions, these systems have 12 ports with pumps and 6 of these 12 ports have pressure transducers. If each port were to have both a pump and a pressure transducer, the maximum number of ports would be eight or nine. Assessment of the results from the 15-port system installed in the first hole using a combination of 12 pumping ports and 6 transducer ports was deemed optimal. Fewer pressure transducers than pumping ports were used in the HQ holes because the variability with depth in the chemical distributions in each hole was much greater than that for hydraulic head distribution at any given time. Furthermore, a negligible vertical component of hydraulic gradient was observed within the aquifer. Pressure transducers, however, were deemed essential to monitor temporal head variability due to intermittent pumping of water supply wells and hydraulic response to recharge. The 12 water supply wells located within 3 km of the multilevel systems are pumped at varying rates according to user demand. Figure 8 shows representative results from pressure transducers recorded over a 2-week period in Only the shallowest port and the deepest port are shown; all of the other ports provided the same temporal trends with hydraulic head at positions between the two shown on this figure. The transducer records show that there is a large diurnal and weekly change in hydraulic head with the lowest to the highest head in each port differing by as much as 5 m within a week interval. In addition, even the daily variations in head are large. The alternative to the use of pressure transducers in the multilevel systems would be to use the open pump tubes for manual measurements. However, because of the rapid daily fluctuations, manual measurements do not provide data suitable for determining head differences at a particular snapshot in time between the various monitoring locations. Moreover, it is impractical to simultaneously measure hydraulic head manually in the 12 wells to give reliable hydraulic gradients. Figure 9 shows the vertical profiles of head for the 15-port multilevel system for a point in time on October 7, 2000, and on the September 2, The shapes of the two profiles are nearly identical. This shows that the change in pressure transducers between the two monitoring years had no substantial influence on the results and that all of the transducers were operational in both systems. The manner in which the water FLUTe system is installed in the holes is sufficiently protective of the transducers to avoid damage during installation. The uppermost three ports in the multilevel system (Figure 9) show the largest head differentials. For example, on September 2, 2000, the head drop between the uppermost port and the port below is 1.2 m, resulting in an estimated vertical component of the gradient of 0.7. The head drop between the second and third port is 0.8 m, resulting in an estimated vertical component of the gradient of 0.3. The length of sealed borehole section between these ports is 1.8 and 2.4 m, respectively. The smaller amplitude of the Hydraulic Head (masl) Friday 12:00 AM Monday 12:00 AM UW16B - Port 1 (shallow; ~273 masl) UW16B - Port 15 (deepest; ~210 masl) Friday 12:00 AM Monday 12:00 AM /28/00 7/30/00 8/1/00 8/3/00 8/5/00 8/7/00 8/9/00 8/11/00 Time (vertical gridlines are spaced 24 hours apart) Figure 8. Temporal variation in hydraulic head caused by pumping from nearby municipal wells. Shallowest and deepest ports in the Cambridge site UW16B FLUTe multilevel system are shown: (a) each day shows a cycle from low to high to low reflecting more pumping during business hours and (b) each week shows a cycle with recovery during the weekends reflecting diminished demand by industry. 64 J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2: 57 70

147 Figure 9. Hydraulic head profiles from the FLUTe depth-discrete MLS in the Cambridge site dolostone aquifer, comparison of two representative profiles for the FLUTe system in 2000 and head variations with time shown by the uppermost three ports (e.g., upper port in Figure 8) and the strong component of the vertical hydraulic gradient shown by these upper ports indicate that this upper part of the dolostone aquifer has greater resistance to vertical flow than the deeper part of the aquifer. Without the large number of ports and the pressure transducers, it would not be possible to discern such characteristics of the hydrogeologic system. Figure 10 shows representative results for several contaminants (MET, TCE, Cl, NO 3, 3 H) obtained from the 15-port system. The large number of data points for each profile establishes distinct features for each contaminant distribution with depth. If fewer ports had been used, important maxima or minima for each contaminant would likely have gone unobserved. For example, in the bottom half of the aquifer, the deepest maximum values for TCE, MET, and chloride are present in only one port and, therefore, would have been missed if this point were absent. The nitrate profile shows the presence of nitrate limited to the uppermost part of the aquifer with an abrupt lower boundary of nitrate contamination. Consistency between sampling results for MET and TCE and the lack of other organic compounds when comparing 2000 and 2004 results provide confidence that the results are representative of the formation without significant influence by any biases due to interactions between the contaminants and the tubing or liner material. Each of the 15 ports was sampled monthly in 2000 and The time for the purging and sampling episodes was approximately 3 h. A single stroke of the double-valve pump supplied approximately 3 L of water from each sample tube. The single stroke provided sufficient sample volume for the analytes required for the project, and sample collection time did not increase appreciably when the list of analytes was expanded. Experience at Other Sites, Failure Mechanisms, and System Longevity The Water FLUTe for ground water monitoring was first installed in the summer of 1999 at a fractured sedimentary rock site in Valley Forge, Pennsylvania. Four-port systems with open tubes for manual water level measurements and double-valve pumps for sampling were installed in six 12.7-cm-diameter vertical holes to 45 m bgs. These systems have performed without problems since their installation (T. Bergling, personal communication, 2005). In addition to the Pennsylvania site and the Cambridge site, the system has been used at 63 sites across North America in many types of rock, including granite in New England, sandstone in California, shale in New York and New Jersey, basalt in Idaho, and karst in Tennessee and Alabama. FLUTe systems have been installed in boreholes with diameters between 7.62 and 45.7 cm and to depths ranging between 15 and 270 m bgs. The depths to static water level have ranged from flowing artesian conditions to 150 m bgs. Installations in karst formations have been successful, even when the borehole passed through substantial caverns and solution channels. The effectiveness of the seal provided by the FLUTe liner in a borehole in a sedimentary rock sequence was confirmed in comparisons of hydraulic head measurements made using packer systems and buried pressure transducers (Bradbury et al. 2007). At the test location, hydraulic head in upper and lower sandstone formations separated J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2:

148 290 A 290 B :<0.05 mg/l Tritium Units (TU) Nitrate + Nitrite (expressed as mg/l N) Trichloroethene (µg/l) 290 C D Trichloroethene Metolachlor :<1.0 ug/l Chloride (mg/l) Metolachlor (µg/l) Figure 10. Examples of ground water quality data obtained from the 15-port multilevel system at the Cambridge site: (A) atmospheric-derived tritium, (B) nitrate, (C) chloride, and (D) TCE and MET. Sampling at closely spaced vertical intervals identifies abrupt changes in concentration with depth with distinct maxima and minima. by a shale unit of approximately 1 m in thickness differed by approximately 8 m. The hydraulic head measurements from adjacent ports in the FLUTe system that straddled the shale unit were in excellent agreement with those made using the packer tests and buried transducers. The capability to maintain this magnitude of head differential over such a short vertical interval is indicative of an excellent seal between the FLUTe liner and the borehole wall. The Water FLUTe system has been used to investigate a range of contaminants at many chlorinated solvent sites as well as sites with petroleum hydrocarbons (e.g., benzene, toluene, ethylbenzene, and xylene; methyl tertiary-butyl ether), polychlorinated biphenyl compounds, or radionuclide contamination. Failures of the Water FLUTe system have been infrequent and nearly all of those that occurred failed at the time of installation due to liner damage during installation or a manufacturing defect. With only two exceptions, each problem was fixed by removal of the liner and repair of the system. For one exception, a small leak was sealed by filling the liner with a bentonite slurry. Failure of the system after it has been in use is very rare (1 in 70), and that failure was attributed to chemical attack caused by contact with a zone of high-concentration chlorinated solvents. At the Cambridge site, 1 of 15 ports stopped producing water a few months after the system had been working well. When the liner was removed (about 1 year later), it was observed that the filter fabric was torn on a spacer located at the bottom of the casing, and the port tube had become clogged with sand. The problem was solved by installation of a reconstructed Water FLUTe system. A major part of the cost of a comprehensive FLUTe multilevel system is the purchase of the dedicated pump and/or pressure transducer for each port. However, in many hydrogeologic situations, the necessary data can be acquired without attachment of a pump and pressure transducer to each port. For example, at sites where the hydraulic head at all depths in the zone of investigation is less than the practical suction depth (i.e., less than ~7.5 to 9 m), open tubes (e.g., 1.2-cm diameter) and a peristaltic pump can provide water samples. If the formation head is somewhat deeper, inertial pumping can be used for purging and sampling. Standard water level probes can be used for manual water level measurements in the open tube as long as temporal fluctuations are not excessive. Small diameter (~1-cm) pressure transducers are available 66 J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2: 57 70

149 commercially (e.g., Druck Model PDCR 35/D). In some situations, only hydraulic head or alternatively water chemistry data are required, in which case cost is reduced by tailoring the systems to this single purpose. For example, we recently used single-purpose FLUTe systems in a fractured-bedrock study where pumping tests were used to examine fracture network interconnectivity. At this site, FLUTe systems with only pressure transducers attached to each port were installed in each hole. Each system included 16 pressure transducers distributed from 4.6 to 45 m bgs in 10.1-cm-diameter holes. After the pumping tests were completed, the multilevel systems were removed and the pressure transducers were used for other purposes. Although the FLUTe systems are normally inflated using water, in a few cases, a bentonite slurry was used to achieve the necessary positive fluid pressure differential in the liners when the static borehole water level was close to or aboveground surface. For these artesian conditions, the bentonite slurry was made heavier by addition of barium sulfate, which is a common drilling mud additive. At sites where the longest possible longevity of the FLUTe system is needed with minimum maintenance, it may be desirable to inflate the system using bentonite or bentonite/cement slurry. In cases where a slurry is used, a tremmie tube (1.8- to 2.5-cm outside diameter) is built into the tubing bundle. The tube extends to the bottom of the liner so that the internal water column can be displaced as the slurry is injected. The addition of weighted mud inside the FLUTe liner has become a common practice for artesian conditions. Tests of the preferred bentonite/barite mixture for a stable mud column were conducted on a vertical 4.25-m column to assure that the barite would not settle out of the bentonite slurry. The simple test for settlement of the mixture was to lower the mud-filled pipe into a horizontal position and determine the balance point of the pipe. The pipe mass center was within 0.3 cm of the center of the pipe. There was no settlement of the 1.26 g/cm 3 mixture over a 2-year period. The performance of the Water FLUTe system can be compromised by failure of the liner or failure of individual ports. If a leak develops in the liner, a loss of adequate head differential to maintain the seal between the borehole wall and the liner will result. In this situation, the entire system has failed and must be removed for repair and reinstallation or installation of a new system. Because the system is removable, both options are available. Leaks are easily detected by monitoring the water level inside the liner over time. Failures of individual ports may arise as a consequence of the failure of the pressure transducer, malfunction of the pump likely as a result of sediment in the ball valves, or clogging of the filter material in the spacer. If the pressure transducer or pump fails, it may be possible to manually measure the water level in or obtain a water sample from the pump tube. Clogging of the screen material in the sampling interval has yet to be observed in system applications. A failure of one port does not influence the integrity of other ports in the system. In all cases, however, the system can be easily removed and repaired as necessary. Potential for Water Quality Sample Bias All ground water sampling systems have potential to cause particular types of bias in the contaminant concentrations and the FLUTe system is no exception. Five potential possibilities for bias are considered here for organic contaminants: (1) bias due to contaminant sorption/ desorption from the tubing conveying the water from the port to surface; (2) contaminant diffusion from the formation through the liner to cause contamination accumulation in the interior water column and then diffusion through the tubing conveying the water sample to surface; (3) contributions of chemicals leached from the system materials (i.e., from the urethane-coated liner and from the tubing conveying the sample); (4) mixing of stagnant water in the port spacers with fresh formation water during the sampling; and (5) inclusion of aerated water droplets from the purge cycle into the water sample if initial water from the sample stroke is not discarded. Drag-down of contaminants sorbed on the liner is not a potential bias source because the liner is everted into the hole. Potential biases (1), (2), and (5) would cause the contaminant concentrations measured in the sample to be less than (e.g., sorption or volatile loss) or greater than (e.g., desorption) the actual contaminant concentrations in the formation, and (3) and (4) would likely cause the sample to show presence of contaminants not occurring in the formation water. These five effects would be very situation dependent, and the FLUTe system and operational procedures are designed to minimize them. First, the tubing used in the pumping system is PVDF (changed from nylon tubing used prior to 2002). Parker and Ranney (1997, 1998) conducted laboratory studies of sorption/desorption to various tubing materials using eight organic compounds: nitrobenzene trans-1, 2-dichloroethylene, m-nitrotoluene, TCE, chlorobenzene, o-dichlorobenzene, p-dichlorobenzene, tetrachloroethylene (PCE), and 20 types of tubing ranging from low-density polyethylene through Teflon. They concluded that among the polymers, fluorinated ethylene propylene (FEP), FEPlined polyethylene (PE), and PVDF were the least sorptive materials tested; PVDF was the best tubing tested, by a good margin, for minimal sorption of TCE and PCE. The Parker and Ranney studies measured the contaminant uptake (sorption) and leaching (desorption) by the tubing materials in a way that includes all of the uptake and leaching mechanisms (i.e., mass transfer) between the water and the tubing, including diffusion into and out of the tubing during alternate clean and contaminated water flows. Therefore, PVDF has been used in the FLUTe multilevel system since 2002, although other types can be used to suit special circumstances (i.e., contaminant types) to achieve lower costs where appropriate. The sample bias indicated in the study of Ballestero et al. (2002) was caused by failure to follow the recommended purging and sampling procedure. Also, those errors were aggravated by the nylon tubing design used in systems built in 2002 and earlier. The other aspect of the FLUTe system design that minimizes bias from processes (1) and (2) is the relatively large yield for each pump stroke. This yield causes water from the formation to pass quickly from the formation J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2:

150 through the entire pumping system to the surface. Typical flow rates from the system average about a liter per minute. The recharge rates are similar for the uppermost ports but faster for the lower ports in a high transmissivity interval. Because the first recharge water is also the last to flow out the sample tube, the longest residence is 10 to 15 min, with a more typical time for the water collected after the discard of 5 to 7 min. If the interval transmissivity is low, the recharge time will be dominated by the formation transmissivity. The use of PVDF tubing further minimizes the effect of longer residence times. Issue (5) is addressed by the sampling procedure, which discards the first flow from the sampling stroke. The third possible source of bias, the leaching of compounds of potential concern from the liner fabric and/or tubing, can be minimized through rigorous purge and sampling practices and can be expected to decrease with time. Trace levels of benzene have been detected in ground water samples from systems where nylon tubing was used, but it is not a concern where PVDF was used. The leaching of toluene, total organic carbon (TOC), and arsenic from the liner material has been documented in field systems and laboratory leach tests. These compounds are seen in the sample water to varying degrees depending upon time and whether the prescribed purge procedure was performed. Toluene, which is used in the production of the urethane coating, has been found in the ground water samples at concentrations of several hundred micrograms per liter, with more typical values of 10 to 70 lg/l soon after the liner installation. The concentrations of toluene have been shown to decrease with time to near nondetectable levels after several months to a year. Concentrations of TOC in ground water obtained from FLUTe systems have ranged from nondetect to several milligrams per liter immediately following installation but typically decrease with time to less than 1 mg/l. A recent side-by-side comparison of a FLUTe system and three cluster wells showed good agreement for TOC concentrations ranging from 1 to 14 mg/l in sampling intervals at the elevations of the three well screens (T. Roeper, personal communication, 2005). The urethane-coated fabric was developed for military use. As part of its general specifications, the material contains arsenic to prevent mildew formation. The liner material may leach some arsenic to ground water. In recent leach tests, arsenic was present at concentrations of as much as 0.2 mg/l in water that had contacted the samples of standard fabric for periods of several weeks. The volume of water in the monitoring interval in contact with the liner material adjacent to each port is small in field systems. Ground water is flushed through the interval immediately following installation and in subsequent pumping cycles associated with purging and sampling. FLUTe liners and MLS systems installed to date have used the standard liner material, but the potential for use of nonstandard, arsenic-free liner fabric has been explored with the manufacturer. Concentrations of arsenic leaching from the material will also likely decrease with time. In field systems, most of the ground water in contact with the liner will be removed during the sample purge cycles. The formation ground water that enters the spacer during the pumping cycles for sample collection contacts the liner material for periods of minutes. Although potential contributions of arsenic to the ground water may be very small during the postpurge sampling process, the use of nonstandard, arsenic-free fabric would be preferable in projects where arsenic is a contaminant of concern. The fourth source of bias, the mixing of stagnant water from the sampling interval with formation ground water during sampling, is expected to be minor. The volume of water in the sampling interval between the liner and borehole wall is small. For example, in a 10-cm-diameter borehole, the storage reservoir for each meter of length of spacer is approximately 300 ml. This is orders of magnitude smaller than the sampled reservoir of the other multilevel systems used in fractured rock, where there is a large annulus between the port casing and borehole wall. The small reservoir of the FLUTe system results in rapid replacement of the reservoir water volume during a single purge stroke. Prevention of Open Borehole Cross Connection Depth-discrete MLSs of all types, including the FLUTe system, provide more insightful data sets when they are designed to suit the particular features of each hole based on core descriptions, borehole geophysics, and/or hydraulic testing. This borehole information is acquired after each hole is drilled; therefore, there is a period of time between completion of drilling and finalization of the multilevel system design that requires sealing to minimize vertical flow conditions in the hole. The potential for cross contamination caused by vertical flow in the open hole may cause sample bias. Sterling et al. (2005) provide a field example involving TCE cross contamination in sandstone. At the Cambridge site, such cross contamination was minimized through use of blank liners installed immediately after the drill rods were removed from the completed hole. The blank liners were removed when straddle-packer hydraulic testing and particular types of borehole geophysics (e.g., acoustic televiewer and formation resistivity) were done in the holes and reinstalled thereafter. Other types of geophysics (e.g., gamma, electromagnetic [EM] conductivity) were done inside the liner. The blank liners were installed using the procedure described for the Water FLUTe system. Removal was easily accomplished by lowering the water level 1 to 2 m inside the liner (close to but not below the open borehole static water level) to facilitate its inversion during removal. With the blank liner in each hole, the multilevel system for that hole was designed and built according to hole-specific features and conditions. Upon arrival of the system at the site, the blank liner was removed and immediately thereafter the multilevel system was installed. System Removability The ability to remove the FLUTe system from holes is advantageous in four ways. First, if problems arise during the installation of a FLUTe system, which occurred at the Cambridge site when a new version of the system was 68 J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2: 57 70

151 being tried in 2005, the system can be readily removed for repair back at the manufacturing facility (Santa Fe, New Mexico). In the meantime, a FLUTe blank liner is installed in the hole to prevent hydraulic cross connection until the repaired system is returned to the site for reinstallation. Because no drill rig is used for installation or removing of the FLUTe system, the cost of using FLUTe systems does not include any rig expense. In the unlikely event that blockage or caving of the hole occurs during installation, removability avoids costs of stuck equipment or rig time for multilevel system removal. Once the liner everts into the hole, it supports the hole wall everywhere against slough of loose rock into the hole. The second advantage of removability is that the borehole is available for other uses. For example, at the Cambridge site, the first FLUTe system was installed in a 100-m-deep borehole. No contamination was expected deeper in the aquifer, which turned out to be incorrect. Therefore, after using the FLUTe system in the 100-mdeep hole to collect data for a year, the system was removed to allow the hole to be drilled deeper. Thus, this first FLUTe system was used as an exploratory tool to acquire information to design the longer-term monitoring system. The third advantage of removability is that the system can be purposefully used for temporary monitoring to avoid the financial risk of potential long-term compliance monitoring. The fourth advantage is that the system allows for decommissioning in a reliable manner. The system can be removed easily and then the hole can be securely sealed with grout. Summary and Conclusions The FLUTe technology is a new and innovative approach for hydrogeologic characterization in bedrock boreholes because the blank liner allows temporary borehole sealing to minimize hydraulic cross connection and the multilevel system can be custom designed for each hole with seals in all segments of the hole not used for monitoring. The use of the FLUTe systems in five dolostone holes at the Cambridge site demonstrated that the system accommodates a maximum of 15 dual-purpose monitoring ports in a 14.3-cm-diameter borehole and a maximum of 10 dual-purpose ports in HQ size holes (9.7-cm diameter). Site-specific designs can accommodate additional ports. For example, 12-port systems were installed in HQ holes at the Cambridge site when optimizing ports for sampling and/or hydraulic head monitoring. Also, at the Cambridge site, the pressure transducers recorded the strongly transient characteristics of the hydraulic head distribution in the water supply aquifer, confirming the advantage of continuous hydraulic head measurements to accurately determine hydraulic gradient in such a transient system. The easy removability of the system was demonstrated when the 15-port system was pulled out of the first monitoring location to allow deeper drilling, and blank liners were installed and withdrawn many times. Removability is an asset allowing alternative borehole uses and low cost borehole decommissioning when the need arises. Problems encountered in the first trial in an HQ size hole led to an improved design. The latest design uses PVDF tubing to conduct the formation water from the port to the surface, which minimizes sampling bias due to tubing effects. The five FLUTe systems at the Cambridge site contain a total of 63 ports. The recording systems for ground water pressure and the short time required for system purging and sampling have facilitated the collection of large data sets with minimal effort by the field personnel. The FLUTe multilevel system has several unique advantages, including easy removability, suitability for application in boreholes with a range of diameters, and the small volumes of water required for complete purge of the pump tube system. Because installation of FLUTe systems does not require use of a drill rig, and because the blank liner is used to seal the hole immediately after drilling, the time between completion of drilling and installation of the custom-built multilevel system need not be subjected to difficult or expensive scheduling constraints. Acknowledgments We thank the many people who played essential roles during the installation of the FLUTe liners and multilevel systems at the Cambridge site, most notably Bob Ingleton, Paul Johnson, and Erik Storms, and those who acquired the data from the systems, specifically Leanne Burns, James Plett, and Chris Turner. Funding for the Cambridge site investigations was provided by Syngenta Crop Protection Canada Inc., Canada Foundation For Innovation, and the Natural Sciences and Engineering Research Council of Canada. Editor s Note: The use of brand names in peer-reviewed papers is for identification purposes only and does not constitute endorsement by the authors, their employers, or the National Ground Water Association. References Ballestero, T.P., G. Pulido-Silva, and K.S. Newman Comparison of bedrock well water sampling methods. In Proceedings of National Ground Water Association Conference on Fractured-Rock Aquifer 2002, March, 13 15, Denver, Colorado, Dublin, Ohio: National Ground Water Association. Berglard, T Personal communication. Baltimore, Maryland. June 23. Black, W.H., H.R. Smith, and F.D. Patton Multiple-level ground-water monitoring with the MP system. In Proceedings of the Conference on Surface and Borehole Geophysical Methods and Ground Water Instrumentation. Dublin, Ohio: National Water Well Association. Bradbury, K.R., M.B. Gotkowitz, D.J. Hart, T.T. Eaton, J.A. Cherry, B.L. Parker, and M.A. Borchardt Contaminant Transport Through Aquitards: Technical Guidance for Aquitard Assessment. Report 91133B. Denver, Colorado: American Water Works Research Foundation. 170 pp. Carter, R.S., W.H. Stiebel, P.J. Nalasco, and D.L. Pardieck Contamination of ground water by pesticides in J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2:

152 Canada. Water Quality Research Journal of Canada 30, no. 3: Cherry, J.A., and P.E. Johnson A multilevel device for monitoring in fractured rock. Ground Water Monitoring Review 2, no. 3: Cherry, J.A., R.W. Gillham, E.G. Anderson, and P.E. Johnson Migration of contaminants in ground water at a landfill: A case study: 2. Ground water monitoring devices. Journal of Hydrology 63, Dunnicliff, J Geotechnical Instrumentation for Monitoring Field Performance. New York: John Wiley and Sons. Merritt, E.A., and P.J. Parsons Sampling devices for water and soil. In Disposal of Radioactive Wastes, vol. II, Vienna, Austria: IAEA. Parker, B.L., J.A. Cherry, and B.J. Swanson A multilevel system for high resolution monitoring in rotosonic boreholes. Ground Water Monitoring and Remediation 26, no. 4: Parker, L.V., and T.A. Ranney Sampling trace-level organic solutes with polymeric tubing part 2. Dynamic studies. Ground Water Monitoring and Remediation 18, no. 1: Parker, L.V., and T.A. Ranney Sampling trace-level organic solutes with polymeric tubing: Part I. Static studies. Ground Water Monitoring and Remediation 17, no. 4: Pianosi, J.G., and T.R. Weaver Multilevel groundwater assessment of confining units in a bedrock sequence near Sarnia, Ontario. In: Hydrology and Hydrogeology in the 90s, American Institute of Hydrology, St. Paul, Minnesota: American Institute of Hydrology. Pickens, J.F., J.A. Cherry, G.E. Grisak, W.F. Merritt, and B.A. Risto A multilevel device for ground-water sampling and piezometric monitoring. Ground Water 16, no. 4: Roeper, T Personal communication. Mahwah, New Jersey. October 24. Sterling, S.N., B.L. Parker, J.A. Cherry, J.H. Williams, J.W. Lane Jr., and F.P. Haeni Vertical cross contamination of trichloroethylene in a borehole in fractured sandstone. Ground Water 43, no. 4: Biographical Sketches John A. Cherry has geological engineering degrees from the Universities of Saskatchewan and California, Berkeley, and a Ph.D. in hydrogeology from the University of Illinois and has been a faculty member in the Department of Earth Sciences at the University of Waterloo since Since 1996, he has held the NSERC-GE Research Chair in contaminant hydrogeology. His research is focused on field studies of contaminant behavior in ground water. He may be reached at Department of Earth Sciences, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada; (519) ; fax (519) ; cherryja@uwaterloo.ca. Beth L. Parker has a Bachelors degree in environmental science/economics from Allegheny College, a Masters degree in environmental engineering from Duke University, and a Ph.D. in hydrogeology from the University of Waterloo. She joined the faculty of the Earth Sciences Department at the University of Waterloo in 1996 and became a professor in the School of Engineering at the University of Guelph in April Her research involves field studies of transport, fate, and remediation of chlorinated solvents in diverse hydrogeologic environments including fractured rock, clayey aquitards, and sandy aquifers. She may be reached at Department of Earth Sciences, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada; She may be reached at the School of Engineering, University of Guelph, 50 Stone Road East, Guelph, Ontario, N1G 2W1, Canada; (519) ext ; fax (519) ; diffusionfx@sympatico.ca. Carl Keller, corresponding author, has Bachelors and Masters degrees in math, physics, and engineering science from Valparaiso University and the Rensselaer Polytechnic Institute. He was employed with the U.S. Department of Energy and Department of Defense from 1966 to 1984 as an underground nuclear test containment scientist, developing a variety of models for multiphase flow in the earth. In 1989, he developed the first everting flexible liner system for collection of pore water samples. He holds 12 patents concerning vadose zone and ground water monitoring and other flexible liner methods. He established Flexible Liner Underground Technologies in 1996 where he is owner and chief scientist/engineer. He may be reached at Flexible Liner Underground Technologies Inc., 6 Easy Street, Santa Fe, NM 87506; (505) ; fax (505) ; carl@flut.com. 70 J.A. Cherry et al./ Ground Water Monitoring & Remediation 27, no. 2: 57 70

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169 Improved Resolution of Ambient Flow through Fractured Rock with Temperature Logs by P.E. Pehme 1,B.L.Parker 2,J.A.Cherry 3, and J.P. Greenhouse 4 Abstract In contaminant hydrogeology, investigations at fractured rock sites are typically undertaken to improve understanding of the fracture networks and associated groundwater flow that govern past and/or future contaminant transport. Conventional hydrogeologic, geophysical, and hydrophysical techniques used to develop a conceptual model are often implemented in open boreholes under conditions of cross-connected flow. A new approach using high-resolution temperature (±0.001 C) profiles measured within static water columns of boreholes sealed using continuous, water-inflated, flexible liners (FLUTe ) identifies hydraulically active fractures under ambient (natural) groundwater flow conditions. The value of this approach is assessed by comparisons of temperature profiles from holes (100 to 200 m deep) with and without liners at four contaminated sites with distinctly different hydrogeologic conditions. The results from the lined holes consistently show many more hydraulically active fractures than the open-hole profiles, in which the influence of vertical flow through the borehole between a few fractures masks important intermediary flow zones. Temperature measurements in temporarily sealed boreholes not only improve the sensitivity and accuracy of identifying hydraulically active fractures under ambient conditions but also offer new insights regarding previously unresolvable flow distributions in fractured rock systems, while leaving the borehole available for other forms of testing and monitoring device installation. Introduction Fractured rock studies aimed at understanding contaminant transport have been in progress for many decades, prompted initially by proposals for creation of deep underground nuclear repositories and stimulated 1 Corresponding author: Department of Earth Sciences, University of Waterloo, 200 University Avenue West, Waterloo, Ontario, Canada N2L 3G1; (519) ; fax (519) ; ppehme@waterloogeophysics.com 2 School of Engineering, University of Guelph, 50 Stone Road East, Guelph, Ontario, N1G 2W1, Canada; (519) ext ; fax (519) ; bparker@uoguelph.ca 3 Department of Earth Sciences, University of Waterloo, 200 University Avenue West, Waterloo, Ontario, Canada N2L 3G1, Canada; (519) ; fax (519) ; cherryja@rogers.com 4 PO Box 122, Tobermory, ON N0B 1N0; (519) jpgreenh@amtelecom.net Received May 2008, accepted September Copyright 2009 The Author(s) Journal compilation 2009 National Ground Water Association. doi: /j x more recently by the prevalence of contaminants in bedrock aquifers at industrial sites. In the quest to achieve better understanding and predictions of contaminant behavior in bedrock, greatly improved characterization of groundwater flow in fracture networks is widely desired (e.g., National Research Council 1996; Berkowitz 2002; Sara 2003). In a recent summary of the state of knowledge concerning groundwater flow and solute migration in fractured rock, Neuman (2005) indicates the need not only to identify dominant discrete fractures but also the hundreds or thousands of fractures having a wide range of sizes. This article focuses on the use of temperature measurements (i.e., temperature profiling) in boreholes sealed with removable liners to improve identification of fractures that are hydraulically active under ambient (noncross-connected) conditions. Davis (1999) summarizes the early work of Humboldt and Arago in the mid-1800s using temperature profiles to describe groundwater flow, hot springs, and geothermal energy resources. Prensky (1992) summarizes the subsequent expansion of NGWA.org GROUND WATER 1

170 temperature measurements into other aspects of earth sciences and subsurface hydrogeology. Anderson (2005) provides a recent review of the use of heat and temperature measurements in groundwater science, indicating initial applications beginning in the 1960s. Trainer (1968) was one of the first to use temperature profiles to investigate groundwater flow in bedrock fractures. He traced major laterally continuous bedding plane fractures in a carbonate rock aquifer for several hundred meters by correlating inflections in open-hole temperature profiles. Drury (1984), Drogue (1985), Silliman and Robinson (1989), and Malard and Chapuis (1995), among others, provide field examples using open-hole temperature profiles to identify hydraulically active fractures. Bidaux and Drogue (1993) compared hydrochemical dilution profiles with temperature profiles at a fractured carbonate rock site and found temperature profiles worked well for the identification of high flow zones, but not low flow zones. These authors and also Robinson et al. (1993) identified two limitations to the usefulness of temperature profiles for identifying hydraulically active fractures: vertical flow in open boreholes and inadequate temperature probe sensitivity. The limitation due to the low resolution of temperature measurements has since been overcome. Genthon et al. (2005) used thermistors in karst studies that resolved temperatures to 0.01 C, concluding high-precision temperature logging is required because the details of the signal s thermal variations could not have been detected with 0.1 C precision. Greenhouse and Pehme (2002) and Pehme et al. (2007a) used improved temperature probes to show repeatable borehole log variations of a few thousandths of a Celsius degree. To avoid the adverse impacts of vertical flow in boreholes because of cross connection between fractures, temperature profiles have been measured in the static water columns inside water-filled steel or PVC pipes (e.g., Keys and Brown 1978; Ferguson et al. 2003). Another approach involves permanently embedding sensors, most recently fiber optic temperature measurement in grout columns outside boreholes casings (e.g., Henninges et al. 2005) sealed in the rock. However, these approaches are rare because dedicated boreholes are expensive and prevent borehole use for other purposes. Fiber optic temperature measurements currently have limited resolution (Wisian et al. 1998) and embedding individual temperature probes in grout becomes prohibitively expensive when a large number of probes are required to produce the very detailed vertical resolution necessary for identifying fractures at numerous depths, which is our focus in this article. To obtain the most useful temperature profiles, avoiding flow within the open borehole caused by cross connection is necessary so that the measurements reflect the natural groundwater system (i.e., ambient flow). However, the practical means for accomplishing this has not previously been achieved. Consequently, temperature profiling has not become an essential or important technique in fractured rock hydrogeology, despite the many examples During installation Reel Everted liner Excess head in liner Liner Fully lined hole Tether Figure 1. Schematic representation of the FLUTe liner installation (from Cherry et al. 2007). presented over past decades. The approach used in the work present here avoids the borehole cross-connection effects typical of open holes in rock by way of an inexpensive, removable, flexible liner (FLUTe, Santa Fe, New Mexico, used for temporary borehole seals (Figure 1). Cherry et al. (2007) introduce the various forms of the FLUTe liner and provide a detailed discussion supporting the assumption that the liner creates a good seal. They provide three forms of evidence: visual (video data), multilevel head data, and a test by Bradbury et al. (2007) comparing the hydraulic head data from a FLUTe multilevel installation with nearby buried pressure transducers. Once installed, these liners perform as continuous water-inflated packers, providing a static water column that takes on the ambient temperature distribution of the rock surrounding the borehole (Pehme et al. 2007a). In this context, an ambient flow system has groundwater movement uninfluenced by the presence of the open borehole, and therefore the flow governing the temperature distribution should also control containment transport in the fracture network. For the temperature profile obtained from the static water column to be useful in identifying fractures, where groundwater flow occurs, the water in the fractures must be in thermal disequilibrium with the surrounding rock; if not, the temperature profile represents the ambient geothermal gradient and is not useful in detailed fracture network studies. Pehme et al. (2007a) conducted temperature profiling in two lined boreholes in a dolostone aquifer to measure thermal dissipation in response to heating the entire static water column inside the liner (known as Active Line Source [ALS] logging). Although their primary purpose for temperature profiling using heating was to determine the thermal conductivity of the rock, they note the method offers potential for identifying fractures with active groundwater flow. This article is an assessment 2 P.E. Pehme et al. GROUND WATER NGWA.org

171 of this potential, whereby we examine high-resolution temperature profiling inside both lined and open boreholes and assess the ability of this method to improve identification of hydraulically active fractures under ambient groundwater flow conditions without reliance on the ALS heat source. Our goal is to extend the capability of temperature logging to discern many more hydraulically active fractures than currently possible, using conventional temperature techniques with emphasis on ambient flow conditions. Within this process, we present the utility of (1) high-pass filtering to emphasize short wavelength variability that may be associated with fracture flow, referred to here as variability logs ; and (2) the comparison of the changes in temperature logs run at different times, referred to here as change logs. We present data collected from a borehole drilled through a dolostone aquifer in Cambridge, Ontario (UW1) as a detailed example of our approach and procedures in comparison to conventional temperature logging. For comparison, temperature profiling was completed in many of the same holes with the liners removed. In several of the lined holes, the profiling was done on multiple occasions spread over weeks or months to determine whether temporal variations provide additional evidence of active fractures. Temperature profiling and related measurements have been conducted in more than 25 other lined holes in fractured sedimentary rock at four contaminated sites: two in Canada and two in the United States. Selected data from these other sites are introduced to show that our conclusions are not unique to the Cambridge dolostone and illustrate a range of responses, including one example showing the effects of a leaking liner. The details of the four boreholes discussed herein are provided in the Supporting Information (Table S1). The boreholes selected to demonstrate the lined borehole method were also subjected to many other types of data acquisition, including geological logging of continuous core, rock core contaminant analysis, other borehole geophysics, flow metering, straddle packer tests, and continuous hydraulic conductivity profiling. Select data from these other methods are used to establish the hydrogeologic context for the temperature measurements and demonstrate that the interpretations from temperature are consistent with and add value to other relevant data. We present many types of data pertaining to fractures and the different meanings associated with the term fracture must be distinguished. Fracture can refer to the geometric discontinuities identified visually in rock core or televiewer borehole imaging. These discontinuities may or may not be hydraulically transmissive or interconnected with other discontinuities. Also, the presence of fractures can be inferred by hydraulic tests or induced flow activity in the absence of other types of information. In summary, identifying fractures having flow under ambient groundwater conditions is essential to understanding contaminant transport, and these fractures are not necessarily those interpreted from hydraulic tests or by imaging open boreholes. Cambridge Open-Hole Data and Interpretation As a consequence of releases of the pesticide metolachlor in the 1970s, the Cambridge site has been the subject of extensive investigation, initially by consulting companies (e.g., Carter et al. 1995) and more recently by researchers at the University of Waterloo. Perrin et al. (2009) describe the hydrogeology of the general area and the site. The facility undergoing investigation is the only local site in Cambridge known to have handled metolachlor. UW1 was drilled at a location where a stratigraphic window in the overburden is believed to have allowed the contaminant to enter the dolostone (Carter et al. 1995). This facility has no history of trichloroethene (TCE) use; however, the property lies within an industrial area and releases are suspected to have occurred creating a TCE source for bedrock contamination 200 to 300 m upgradient of UW1. Figure 2 provides a suite of conventional open-hole data collected in borehole UW1 at the Cambridge site. This information is typical of what might be available for interpretation of flow and the planning of a multilevel monitoring system installation as part of a contaminated site investigation. Figure 2 includes the general stratigraphy (Figure 2a) intersected by UW1 as interpreted from continuous core, a natural gamma log (Figure 2c) collected from within the 150-m deep, 10-cm (4-inch) diameter open borehole, and a virtual caliper profile (Figure 2j) based on the average travel time calculated from a FAC40 (Advanced Logic Technologies, ALT redange sur attert, Luxembourg) acoustic televiewer (ATV) log. The geologic sequence at UW1 is typical of Cambridge and the surrounding areas. The uppermost bedrock units are relatively flat-lying fractured dolostones of the Guelph and Lockport formations that overlie the Rochester shale. Based on the gamma log, the dolostone has a relatively uniform, low clay content with the exception of the argillaceous Eramosa member, which forms the upper part of the Lockport formation. The Eramosa is recognized regionally as a laterally extensive horizon, with a gamma signature readily distinguishable from other dolostone units above and below; in some parts of the region, it is considered to be an aquitard, but not in the study area. Groundwater flow in the dolostone aquifer occurs in fractures and is controlled by several pumping wells surrounding the site (the closest is 900 m south). The pumping wells are open from just below the bedrock surface ( 15 to 30 mbgs) to depths between 60 and 100 mbgs. In a nearby multilevel installation, the hydraulic head near surface is higher than encountered at depth, inferring overall downward flow through the dolostone aquifer. The hydraulic head levels across the aquifer typically fluctuate by almost a meter over 1-week cycles, due to municipal pumping in the area. Perrin et al. (2009) show evidence of karst features in the dolostone but conclude that although karst channels have local influence, they generally do not govern the groundwater flow system and contaminant distributions in the Cambridge area. NGWA.org P.E. Pehme et al. GROUND WATER 3

172 Figure 2. UW1 open-hole data showing (a) the stratigraphic column (from Burns 2005), (b) hydraulic conductivity from straddle packer testing (m/s at a log scale), (c) gamma log (cps), (d) temperature logs collected in the open borehole on December 1, 2003 and January 12, 2004 ( C), (e and f) the thermal gradient (over 10 cm) of the December 1, 2003 and January 22, 2004 temperature logs, respectively ( C/km), (g and h) variability logs for December 1, 2003 and January 22, 2004, respectively ( C), (i) the stationary heat pulse flowmeter tests (L/min), and (j) virtual caliper (mm) from travel time of acoustic televiewer data (FAC40 manufactured by ALT Ltd.). Straddle packer tests were conducted at 2.2-m-wide intervals through the length of UW1 (Figure 2, b). All of the intervals tested are interpreted to have bulk hydraulic conductivity above the method detection limit of 10 8 m/s and several zones of elevated hydraulic conductivity exist throughout the borehole (e.g., 174, 189, 212, 226, 235, 249, 254, 272, and 278 m above sea level [masl]). Numerous rock core samples (cylinders with 38 mm diameter and 40 to 70 mm length) cut from the larger core were tested in the lab for rock matrix permeability, and results consistently show hydraulic conductivity values much below the lower limit of the packer tests. Visual inspection of the larger core specimens suggests the effect of anisotropy is small. Comparing the results of rock core permeability tests against packer tests leads to the conclusion that either individual large aperture fractures or numerous smaller fractures with substantial combined hydraulic conductivity occur within many of the tested intervals, and therefore abundant potential for ambient groundwater flow exists. We use the irregularities in the borehole diameter, represented by the virtual caliper calculated from the ATV travel time (Figure 2j), as a convenient representation of geometric fractures for comparison with other open-hole data, acknowledging the pulse width of the probe limits the resolution of discontinuities on the borehole wall that are less than 3 mm (ALT 2002). While recognizing that the drilling process can increase the apparent fracture aperture at the borehole wall, given the competency of the dolostone and the scales at which the data herein are compared, any such enlargement is likely inconsequential to this discussion. The ATV data show a higher fracture frequency above the zone at 235 masl than below. As well as identifying several large fractures, the ATV data indicate numerous smaller potential discontinuities of varying aperture. Other notable characteristics in the ATV data are a large void immediately below the bottom of the casing at masl and the scarcity of irregularities between and masl. The size and frequency of the irregularities in the ATV data generally, but not always, correlate with zones of elevated hydraulic conductivity measured in packer tests. Obvious exceptions are the presence of two zones of hydraulic conductivity above 10 5 m/s (at 188 and 210 masl), within a portion of 4 P.E. Pehme et al. GROUND WATER NGWA.org

173 the borehole having relatively few and small fractures, as well as the other more severely fractured zones with lower hydraulic conductivity. Inconsistencies in the correlation between the ATV and packer tests are expected because the ATV does not distinguish between permeable and impermeable fractures, and cannot detect very small aperture fractures. Variations in vertical flow in UW1 under open-hole conditions (Figure 2i) were measured on December 4 and 5, 2004, using a Mount Sopris model HFP2293 heat pulse probe. Measurements were made at 1-m intervals starting from the bottom and moving up, with the probe kept stationary at each test location. At each position, the instrument was allowed to stabilize before measurement and as many as four readings were taken to confirm repeatability of the results. The heat pulse flowmeter data could not be interpreted above an elevation of 238 masl, because the responses were either too irregular to differentiate a single pulse or entirely flat. Given the large downward gradients measured in two multilevel monitoring installations 8 to 10 m away (Perrin et al. 2009), the most plausible interpretation for the poor data quality is that the borehole is not stagnant above 238 masl but rather has high flow, dominantly downward and possibly with a horizontal component, beyond the measuring capability of the heat pulse probe. From 238 to 193 masl, the flow decreases through a transitional zone that is interpreted to have many fractures, the majority of which act as minor drains or outflow points along the borehole. Relatively low flow occurs below 193 masl with the exception of two tests just below 177 masl. This zone of low flow correlates with the portion of the borehole, where the virtual caliper log indicates only a few fractures. Overall, the heat pulse flowmeter data show strong downward flow entering the open borehole just below the casing with much of this water exiting at or near 238 masl. The open borehole temperature logs at UW1 (Figure 2d) were collected in December of 2003 and again in January of 2004, with a BMP04 temperature probe manufactured by Instruments for Geophysics Corporation (IFG) of Brampton, Ontario. This probe measures the water temperature with an accuracy of 0.1 Candresolution on the order of C (IFG 1993; Greenhouse and Pehme 2002; Pehme et al. 2007a, 2007b), with a time constant of approximately 1 s (Blohm 2007). The borehole water column was allowed to stabilize undisturbed for several days before collecting data to avoid thermal disturbance due to drilling, other geophysical logging, hydraulic testing, or in some cases, liner removal. The temperature was measured while downward logging at a nearly constant speed between 0.5 and 0.7 m/min and the system recording raw data at a rate of 2 Hz. The results were subsequently splined and resampled to convert the data set to a constant depth interval, consistent with the nominal raw sampling distance (0.005 m for Figure 2d). To highlight small-scale irregularities and variability, common practice is to calculate a thermal gradient profile from the temperature data (Figure 2e and 2f), in this case as the difference in temperature over a vertical distance of 0.1 m, reported in units of C/km. The basic premise for the interpretation of a temperature profile is that below the near-surface, environmentally influenced, heterothermic zone, the spatial variation of temperature in a relatively uniform medium, such as solid rock, should be a reflection of the very gradual regional geothermal gradient. The thermal conductivity of rock is typically 2.5 to 5 times that of water (Bejan 1993), and therefore, without annular flow, stagnant water in a borehole will not facilitate vertical heat conduction faster than the surrounding material. Furthermore, the only mechanism available to perturb the uniform geothermal gradient is the transport of heat by groundwater movement through flow pathways. Anderson (2005) summarizes many of the efforts at estimating broad-scale recharge and discharge using heat as a tracer. In rock where the water movement is primarily in fractures, narrow aberrations (either positive or negative) in the temperature profile measured in a borehole, devoid of any cross connection, are expected to be the result of water flow in fractures. Molson et al. (2007) use a numerical model to test the conceptualization of detailed temperature variations caused by flow in fractures as described earlier. In simulations of a fractured rock system where seasonal surface temperature variations typical of the Cambridge study area were invoked, they found numerous temperature variations of magnitudes similar to the measured temperature profiles presented in this article for flow in a network of many interconnected hydraulically active fractures. The only other mechanism that may cause thermal disequilibrium between the water in the borehole and the surrounding formation is convection, a potential occurrence when the temperature of the borehole environment increases with depth. Sammel (1968) addressed convection within a cased borehole and showed that within a 4-inch (10-cm) well at water temperatures of 10 C to 15 C, convection should be anticipated at critical temperature gradients of approximately to C/m. Pehme et al. (2007c) show readings at the Cambridge site varied by less than one hundredth of a degree Celsius over a 1-week period, where the gradients were less than C/m. The variability over the same period increased with the thermal gradient to approximately C, where gradients were above C/m. No evidence indicates convection in the profiles presented or indicates it is a factor in these interpretations. The gradient log (Figure 2), also referred to as the differential temperature, is a derivative of the temperature log commonly used to emphasize short wavelength events that could represent water movement (Keys 1989). However, when calculations are conducted over vertical spans that highlight small-scale features such as fracture flow, the result becomes bimodal and distorts the shape of a peak or trough anomaly. An alternative method for examining the variations within a temperature log that represent hydraulic flow through fractures while avoiding this distortion is to subject the data to a simple high-pass filter to produce what is referred to here as a variability log. The NGWA.org P.E. Pehme et al. GROUND WATER 5

174 raw data are smoothed with a box-car filter, with a typical window length of 5 m. Subtraction of this smoothed base log from the original data creates the variability log that emphasizes features with length scales less than the filter window, while suppressing broad features such as the geothermal gradient and the shallow environmental (seasonal) temperature variations, with minimal distortion of the shape of the small-scale anomalies. Figure 2 (g and h) shows the variability logs for the open-hole temperature data collected in UW1. The two open-hole temperature logs were collected approximately a month apart, and the data sets calculated from them (Figure 2e to 2h) show the same basic patterns, indicating similar hydraulic conditions existed during (and immediately preceding) the two measurements. The water in a shallow fractured zone (285 to 290 masl) is warmer than the water deeper in the hole. Below the shallow zone, the temperature of the water in the borehole decreases only minimally, and almost linearly, with minor deviations to 238 masl coincident with the depth where flowmeter measurements changed from uninterpretable to interpretable. The flowmeter results and the temperature profiles in the open hole are both consistent with the view that these open-hole data are dominated by downward flow above a fracture at 238 masl, which acts as a major hydraulic drain (sink), accepting much of the water flowing down the open borehole. The absence of variability above 238 masl suggests very little downward flow leaves the borehole at these depths, despite the fracturing evident on the virtual caliper log. On the other hand, the variability may depend only on downward flow velocity below some threshold value (exceeded above 238 masl). From 238 to 200 masl, the temperature decreases steadily but with notably larger variability, implying flow out of the borehole into fractures. This variability is not coherent between the two temperature logs in its short-scale detail, but the decline in temperature and the level of variability correspond well to the progressively decreasing downward flow velocity indicated by the flowmeter. Although the details of flow into individual fractures cannot be inferred, the decreasing downward flow velocities recorded by the flowmeter in this section support the premise of outward flow of water into fractures, with the temperature variability possibly the result of minor flow complexities at the fracture entrance. The only distinct deep feature in the temperature data that is indicative of flow occurs at approximately 174 masl, near the top of the Rochester shale, where the ATV log indicates a series of fractures, the packer tests measure higher hydraulic conductivity, and slightly below where the heat pulse flowmeter indicates downward flow increases over a short interval. Although the majority of the flow occurs higher in the borehole, some water movement is detected throughout the dolostone to the top of the Rochester shale. In summary, although the heat pulse flowmeter results and the temperature logs provide a mutually consistent interpretation of flow in the open borehole, there is little correlation between most of the features identified from these techniques and the numerous (geometric) fractures identified with the ATV or the high-permeability tests measured with the straddle packer. Only the zone from 235 to 238 masl is distinguishable in all data sets, as the lower limit of high downward flow, and a zone of elevated hydraulic conductivity with several large aperture fractures. The data related to water flow show little indication of the high-permeability zones or numerous distinct fractures above 238 masl. In general, these openhole data do not provide a data-consistent basis for ranking the importance of sampling zones and designing a multilevel monitoring installation. Cambridge Lined-Hole Data and Interpretation After completion of the open-hole logging of UW1, a FLUTe liner was installed to a depth of 135 m (171 masl) to prevent cross-connected flow. The liner was filled to achieve a head of approximately 3 to 5 m above the standing water level in the open borehole, thereby inflating the sleeve so it presses against the borehole wall and seals the fractures. The borehole was then temperature logged several times over a 2-month period. Figure 3 displays a broader set of data collected in UW1, including open and lined temperature profiles, alongside other geophysical logs, straddle packer testing, and rock core analyses to be discussed further later in this text. Figure 3f displays the three lined-hole temperature logs collected on February 16, March 1, and April 12 of 2004, alongside the two open-hole logs from Figure 2d. Figure 3 (i to k) shows the corresponding variability logs calculated from the respective temperature logs. The different scales used for the variability logs are set so as to display a similar level of apparent variability, where the response in all three logs is similar and yet fractures are anticipated (190 to 220 masl). The portions of the temperature profiles collected above 279 and below 187 masl inside the liner are similar to those collected in the open hole. However, with the vertical cross-connected flow restricted, the lined-hole temperatures between these elevations are cooler than their open-hole counterparts, and large, time-varying peaks in temperature at 235 to 240 and 253 to 259 masl clearly identify broad zones of ambient groundwater flow. The variability logs emphasize small-scale anomalies within these broad peaks, which we believe to identify individual fractures showing evidence of substantial ambient flow. To interpret these small-scale variations relevant to groundwater flow, we look for irregularities that stand out on individual logs and are consistently present in some fashion on all three data sets (recognizing the actual shape of the anomaly may vary with time as discussed earlier). On that basis, in addition to the two major flow zones, and ignoring thermal irregularities in the vicinity of the casing, several smaller variations (e.g., at 226, 230, 250, 263, 272, and 282 masl) also meet these criteria. The features in the lined-hole temperature logs vary in character with time and often the variability logs are required to identify the most subtle features. The temperature variations in the April 12, 2004 logs are on the order of 0.01 C, and 6 P.E. Pehme et al. GROUND WATER NGWA.org

175 Figure 3. UW1 data showing (a) the stratigraphic column, (b) hydraulic conductivity from packer testing (at a log scale), (c and d) TCE and metolachlor rock core analysis (μg/l) (red, quantifiable; blue, low order quantification; and green, below detection limit), (e) gamma log (blue), (f) passive temperature logs collected in the lined boreholes from left to right on April 12, 2004; March 1, 2004; and February 16, 2004 and open boreholes on December 1, 2003 and January 22, 2004, (g) change logs from lined boreholes on March 1, 2004 (blue) and February 16, 2004 (green), (h, i, j, and k) variability logs January 22, 2004; April 12, 2004; and February 16, 2004, (l) heat pulse flowmeter, and (m) virtual caliper from travel time of acoustic televiewer data, with interpretation of flow zones (blue shading) and limits of temporal change in flow. (Columns a, b, c, and d from Burns [2005]). barely discernable on the original log. Nevertheless, these repeatable perturbations are strong evidence of groundwater flow effects at these depths. All of the variability logs, both open and lined holes, show the temperature is highly uniform and steady below an elevation of 196 masl. Between 222 and 196 masl, the temperature in the lined borehole continues to be relatively uniform and steady with only minor variations (e.g., at 210 masl), while in contrast, the open-hole temperature profile is irregular implying some flow. In this portion of the open-hole logs, the fractures are disproportionately emphasized as the temperature of the borehole fluid transitions from the deepest hydraulic outflow of consequence at 222 masl to near stagnant conditions below 196 masl. Figure 4 contrasts the key characteristics of water flow interpreted from the temperature profiles collected in UW1 while both open and lined. Arrows of varying size are used to indicate the relative amounts of flow in the major zones and differentiate predominantly vertical flow along the axis from flow which is into or out of the open borehole, or across and around when lined. In the open hole, downward flow originating from shallow fracture(s) near the water table at 284 masl dominates the upper part of the temperature profile. The linearly varying temperature with a low gradient is consistent with a large amount of water moving down the open borehole, gradually equilibrating with the formation. There are no indications of either additional sources or outflows until much, but not all, of the downward flow exits the borehole at 235 masl. The water that continues to flow downward below 235 masl is warmer than the formation and gradually reaches ambient temperatures at 200 masl. Over the interval from 235 to 200 masl, water is distributed to various small fractures as confirmed by the heat pulse flowmeter results (Figure 3h). The temperature does not vary in time (is thermally stabile) and hydraulic activity below 200 masl is not otherwise indicated, with the exception of a single inflection in the profile at 174 masl that implies some flow across the borehole. The lined-hole temperature profiles provide a very different perspective. Over the interval between 284 and 235 masl, where the open-hole profile shows two flow zones, the lined hole shows at least seven (Figure 4). However, the importance of the differences goes beyond the number of flow zones. The lined-hole data (Figure 4) indicate the major flow is at 258 masl, but the strong NGWA.org P.E. Pehme et al. GROUND WATER 7

176 Figure 4. Comparison of basic interpretations of temperature logs collected in open and lined borehole UW1. Blue arrows indicate major and minor flow zones. Red arrows are lower limits of shallow flow based on temperature variability (all C ). vertical cross-connected flow makes this zone indistinguishable on the open-hole temperature profile. Although the flow zone at 258 masl is not as obvious in the April lined-hole data (Figure 4) compared with other open- and lined-hole data collected earlier in the year, it creates an irregularity on the variability log that is comparable to the other large flow zones in the borehole. The lined-hole temperature logs allow for more flow zones to be identified on a consistent (repeatable) basis and a very different interpretation of the relative amounts of water movement in the zones that are identified or in some cases masked by vertical flow in the open hole. Temporal Assessment of Temperature Profiles Visual comparison of Figure 4a and 4b indicates that major flow occurs under ambient conditions at 285, 255, and 235 masl. However, repetitive logging spaced by days to weeks or longer is required to identify additional flow zones. The change log procedure (Figure 5) was developed to improve the representation of the temporal variations in temperature response. This involves assigning one logging event as the reference profile and, in a manner similar to the calculation of the variability log described earlier, smoothing it to create a base log representing the geothermal gradient and the environmental variations. The change log is then calculated by subtracting the base log from other logging events in the same borehole and normalizing the result by dividing by the time span (in days). The use of time to normalize the change is intended to improve comparison between boreholes at a site when logged on different days over periods of less than a month and has a decreasing value when looking at long-term changes over several months or years. Logs collected prior to the base log are represented by negative time and later data sets by positive time, resulting in decreases in temperature over time being negative and increases positive. The long-term geothermal gradient and medium-term surface effects common to both logs are suppressed in the change log, accentuating the 8 P.E. Pehme et al. GROUND WATER NGWA.org

177 Figure 5. Repeated temperature logs in FLUTe lined borehole UW1 highlighting temporal changes: (a) lined-hole temperature logs, (b) change logs, and (c, d, and e) variability logs, with interpretation of key flow zones (blue arrows) and example of subtle anomalies (purple asterisk), (all C ). relatively short-term temporal variations in the borehole. Short-term changes in temperature reflect a redistribution of the groundwater temperature as a result of, for example, groundwater recharge events or changes in the flow system caused by municipal well pumping cycles. However, depending upon the time between the acquisition of the data sets, the change log may include some broader scale seasonal or environmental temperature variations. Figure 5b shows the UW1 change log for the lined borehole. Superimposed on Figure 5b is the interpretation of the major flow zones previously presented in Figure 4, refined based on the change log. The qualitative assessment of the relative amount of flow as depicted by the size of the flow arrows is best rationalized in the temporal temperature variation represented by the change logs, as are the limits of apparent major and apparent active change. Notably, the flow zones dominating the February and March temperature profiles also manifest as aberrations in the comparatively smooth temperature log collected in April, although they are considerably subdued and are best depicted in this case in the variability log, demonstrating the enhanced value of the combined use of variability and change logs. Numerous other small-scale irregularities in the change logs are present in the individual variability logs. For example, the small feature highlighted by purple asterisk on Figure 5 is apparent in all three independent data sets. The correlation of individual small-scale features is best when the system is in the most disequilibrium, in this case, February and March. Where polarities reverse or features exist in two, but not three, data sets, the most plausible explanation is that the details of groundwater flow have changed, possibly by variations in pumping or NGWA.org P.E. Pehme et al. GROUND WATER 9

178 infiltration. Although many of these fine-scale features likely represent water flow with minimal flux, neither the size nor polarity is always consistent, and a better understanding of the details of the system would be required for these to be individually interpreted. Although the details vary, the temperature logs collected in six other boreholes at the Cambridge site under lined- and open-hole conditions have many of the same features and similar implications from the comparisons drawn from the UW1 data. Applications at Other Sedimentary Rock Sites Although the general nature of the differences in results between lined and open holes for three other sites, where the techniques have been extensively applied (Guelph, Ontario; Simi, California; and near Madison, Wisconsin), are similar to the differences observed at the Cambridge site, the temperature profiles from these other sites provide important additional insights. The hydrogeology of the Guelph site is similar to the Cambridge site, including bedrock with the same geologic units, primarily dolostone. At the Simi and Madison sites, bedrock is mostly sandstone, flat lying at Madison and dipping about 30 at Simi. At the Cambridge site, municipal pumping wells are operating close to the study area, but at the other three sites pumping wells are comparatively distant from the holes. The Simi site is on a ridge and topography creates a strong downward hydraulic gradient to drive groundwater flow. Each of these sites provides very different hydrogeologic and/or geologic conditions than the Cambridge example. The decreasing temperature with depth throughout most of the Guelph and Madison boreholes and the variability through the entire length of the lined-hole logs in the three data sets suggest all are in a state of thermal disequilibrium to the depths drilled. Similar to the Cambridge site, the two most important aspects to examine for these other three sites are the number of hydraulically active fractures apparent from open and lined holes and the differences in relative importance assigned to particular flow zones under cross-connected (open) and ambient (lined) conditions. Further discussion of the Guelph site follows, while data and details of the Simi and Madison sites (Figures S1 and S2, respectively) are provided in Supporting Information. Figure 6 displays examples of open- and lined-hole temperature profiles from the Guelph site, including arrows of varying size, representing relative amounts and direction of flow using a consistent color scheme (red for open hole and blue for lined hole) along with gamma logs. For the Guelph site, heat pulse flowmeter data and a caliper log derived from the ATV are also displayed (Figure 6). The open- and lined-hole temperature profiles from the Guelph site are similar in general characteristics to those from the Cambridge site, indicating overall consistency of the groundwater flow systems at the two locations. The open-hole temperature data collected in Guelph indicate water entering the borehole at approximately 325 masl, and although intermediate perturbations are present in the logs, the flow is dominantly downward and exits at 261 masl. The open-hole variability logs are for the most part mutually consistent, but with poor correlation between the fractures identified in the ATV virtual caliper or the flowmeter data and the openhole temperature results. In contrast, although lined-hole temperature logs appear uniform below 320 masl, with a single bulge deep in the dolostone, the variability logs display numerous irregularities throughout the borehole that are highly coherent at amplitude levels of a centidegree or less, indicating many fractures with flow exist. Peaks in the variability log (negative and positive) coincide with most geometric fracture zones identified from the ATV virtual caliper. The interpretation of the lined-hole data provides a total of 10 major flow zones. These data suggest the main contributors to the temperature bulge (ambient ground water flow) are fractures at 263 and 267 masl, but only to a limited degree at 261 masl, the elevation at which the open-hole temperature data suggest the major outflow occurs. This example highlights the usefulness of the variability log in representing very small, yet repeatable irregularities that correlate well with other data sets. It also provides another instance where the most distinctive feature in the lined-hole temperature data profiles presents as one of several irregularities on the open-hole profile that would be unlikely to warrant particular attention in the form of additional testing or allocation of a port in a multilevel installation. The results from all four sites indicate the lined-hole temperature logs provide identification of hydraulically active fractures down to substantial depth at each site, ranging from 110 m at the Guelph site to 155 m at the California site. In all cases, both open- and lined-hole logs converge to a common temperature at the bottom of the borehole. Investigations of many contaminated sites on bedrock have most emphasis within a hundred meters of ground surface, suggesting the lined-hole temperature method will likely have widespread usefulness. However, in this depth context, it is relevant to consider whether the four sites have provided a biased impression. For the lined-hole method to show active fractures, the temporal temperature variations occurring at the surface must be transmitted relatively deep into the bedrock fracture network. Each of the four sites considered in this article has two conditions that are highly favorable for propagation of temperature disequilibrium: a vertical component to the general hydraulic gradient causing downward flow to the bottom of the domain of interest, nor has the overburden created a barrier to recharge by being excessively thick or having a low vertical hydraulic conductivity. At the Simi site, the overburden is thin or absent over much of the area; at the other sites, the overburden is thin or lacking a substantial aquitard unit where moderate in thickness. If either of these two major factors were absent at any of the four study sites, the maximum depth of sensitivity of the lined-hole method may have been shallower, which could render the method less useful. 10 P.E. Pehme et al. GROUND WATER NGWA.org

179 Figure 6. Data from MW24 (Guelph, Ontario): (a) stratigraphy (ob, overburden; G1-4, Guelph formation subunits; and LH, Lions Head; from Perrin et al. [2009]). Note that two naming conventions exist for the same stratigraphic sequences through portions of southern Ontario (i.e., the Amabel is equivalent to the Lockport Formation), dependent on location relative to the Algonquin geologic arch, which separates the Appalachian and Michigan sedimentary basins. The geological community interprets the arch to be between the Cambridge and Guelph sites, and we have adopted the local conventions chosen by others (Coniglio 2007); (b) gamma log (cps); (c) temperature profiles (open holes shades of red, lined hole blue) (C ); (d) lined-hole variability logs (C ); (e) open-hole variability logs (C ); (f) change log (C ); (g) heat pulse flowmeter flow (L/min); and (h) virtual caliper from travel time of acoustic televiewer data (mm). Arrows schematically represent zones of flow, amount (by size), and direction (orientation), and are referred to in the text (red for open hole and blue for lined hole). Additional Evidence for Numerous Hydraulically Active Fractures In all of the cases presented, the temperature profiles inside lined holes show many more hydraulically active fractures than are indicated by open-hole logging. Other independent lines of field evidence, including rock core contaminant analysis, borehole hydraulic tests, and ATV logging, support the concept of a large number of fracture pathways for groundwater flow. Parker et al. (1994, 1997) used analytical models and representative sandstone parameters from the literature to predict that with sufficient residence time, as a contaminant moves through the fractures of a sedimentary rock, chemical diffusion can cause a considerable amount of the contaminant mass to transfer from within fractures into the adjacent rock matrix. VanderKwaak and Sudicky (1996) confirmed the potential for the contaminant halo effect in fractured porous geologic media using a numerical model. Parker (2007) describes the most recent iteration of a methodology wherein analyses of closely spaced rock samples collected from continuous cored holes are used to create contaminant mass vs. depth profiles, thereby inferring contaminant transport within fractures and consequently active groundwater flow. Sterling et al. (2005) provide an example of such a profile from sandstone at the Simi site. Rock core analyses were conducted for the volatile organic contaminant TCE and the pesticide metolachlor at the Cambridge site. In Figure 3, the most hydraulically active fractures identified in UW1 based on temperature profiles are shown as blue shading alongside other geophysical logs, straddle packer testing, and the rock core analyses, thereby providing the framework for examining NGWA.org P.E. Pehme et al. GROUND WATER 11

180 all lines of evidence concerning fractures in this hole. The TCE and metolachlor profiles show that although the highest concentrations exist above 253 masl, numerous contaminant occurrences, particularly for TCE, are distributed across the thickness of the dolostone aquifer indicating that the fracture network is vertically interconnected (Perrin et al. 2009) and groundwater flow has occurred throughout the aquifer under ambient conditions. The straddle packer tests (Figure 3b) indicate measurable hydraulic conductivity exists in all intervals confirming that fractures exist from the top to the bottom of the dolostone. The ATV log also confirms that the fracture density varies with depth, but there are numerous fractures throughout most of this borehole (Figure 3m). Although neither the packer test nor the ATV results provide direct evidence of groundwater flow under ambient conditions, the presence of numerous fractures is consistent with the interpretation of many active flow zones from the linedhole temperature logs and rock core profiles. The degree of correlation of the hydraulically active fractures as identified by the lined-hole temperature profile, rock core contaminants, and implication of fractures identified from packer tests and the ATV log varies in different portions of the borehole. Above approximately 250 masl, the flow zones as indicated by irregularities in temperature are at the same elevations as the peaks in the other three data sets and also match in relative size. The coincidence between the major metolachlor and TCE contaminant concentration peaks in the rock core and the irregularities within the various forms of lined-hole temperature data in its raw (Figure 3f), processed (Figure 3g to 3k), and interpretation (shading) are generally very good. The variations and changes observed to dominate the lined-hole thermal profile at 254 and 270 masl coincide with the zones of highest contaminant concentrations. Although many of the lesser rock core peaks can also be related to aberrations in the thermal profiles, inconsistencies remain due to an inherent difference in the nature of the results from the two methods. From 222 (the limit of major change) to 250 masl, the temperature irregularities coincide with the zones of high hydraulic conductivity and fractures in the ATV, but relatively few rock core peaks exist; and below 222 masl, the peaks identified in all four data sets match inconsistently. However, there is no reason to expect strong positive correlations between the four types of evidence for fracture occurrences because each of these techniques measures a different aspect of the system; the presence of an opening on an ATV log implies nothing about water movement, nor does water dissipation under pressure in a packer test confirm water flows under ambient conditions. In addition, the fracture network is three dimensional and therefore the borehole directly encounters only some of the fractures involved in contaminant migration near each hole. Although elevated contaminant levels in rock core analyses indicate migration pathways nearby, the actual groundwater flow may not intersect the borehole but instead be within a meter or so. The temperature profiling indicates the hydraulic activity at the moment of profiling; when done more than once, it indicates variation over a fixed, but relatively short, time interval. At the Cambridge site, the contaminants have been in the dolostone aquifer for at least two decades (Carter et al. 1995) while the number and pumping rates of municipal wells have varied. Therefore, the contaminant concentrations now found in the rock core represent the cumulative influence of diffusion into and then later out of the rock matrix blocks between fractures over decades; exact correlation between the degree of hydraulic activity in the fractures identified by temperature profiling and the strength of the contamination in the rock core is not a reasonable expectation. Lastly, the absence of a chemical peak infers little about water movement if the water is not contaminated. Although the ATV caliper and straddle packer tests in UW1 indicate potential for flow below 200 masl and the TCE peaks confirm water movement has occurred, the temperature profiles are smooth and uniform over time, implying little ambient flow. To create an aberration in the temperature profile, the water moving through the fracture must be at a sufficiently different temperature than the rock to cause a detectable change. Raw temperature logs vary temporally according to source (surface) temperature changes, the relative size and connection of flowpaths, and the changes in driving forces (pumping in Cambridge). Although the variability logs improve the interpretation of the flowpaths, the size of the variations also varies over time. As the degree of thermal disequilibrium between the water and the rock would decrease with depth, so also would the ability to resolve flowing fractures from temperature profiles. Importantly, the ALS technique (Pehme et al. 2007a, 2007b), presented as a method for estimating the thermal conductivity of the formation, artificially creates thermal disequilibrium in the water column and has the potential to improve the detection of ambient flow at depth. This study presents the hypothesis that variations in thermal energy, primarily originating at surface, can propagate through the overburden and be transported to substantial depth by groundwater flow in a fractured bedrock system before the temperature contrast is attenuated below the detectable limits of the field equipment. The temperature profiles provided are consistent with that hypothesis and no other alternative hypothesis has been identified to explain these field results. Although the lack of alternative explanations supports the hypothesis, it cannot be taken as definitive corroboration on its own. However, there is independent support from Molson et al. (2007) wherein a numerical model of heat transport and groundwater flow in fractured rock examined the plausibility of this premise. The model includes density-dependent groundwater flow through discrete, stochastically generated fracture networks coupled with thermal advection, conduction, and retardation within the porous rock matrix. The model boundary conditions and media properties/parameters were selected to stylistically represent the setting and dolostone aquifer underlying the City of Guelph, Ontario and are also generally characteristic of the same aquifer in Cambridge, Ontario. In the model, 12 P.E. Pehme et al. GROUND WATER NGWA.org

181 natural heat energy pulses are generated based on seasonal air temperatures and applied as a thermal flux condition uniformly over the upper boundary surface. Molson et al. concluded that ground source thermal pulses can propagate deep into a fractured rock system and appear as weak thermal anomalies within the fractures on the order of a few tenths or hundredths of degrees. Therefore, this modeling indicates that the use of high-resolution temperature profiling to identify hydraulically active fractures, as we presented in this article, is consistent with what is known based on the physics of heat transport in rock fracture networks. In the assessment of lined-hole vs. open-hole temperature results, the lined-hole data clearly provide identification of substantially more active fractures under ambient flow conditions. However, the open-hole data can also give a misleading view of the ambient flow system in the fracture network. The peaks in contamination measured from the rock core data confirm fractures in the vicinity of 254 masl (Figure 3) are important for understanding contaminant migration. This same zone dominates the lined-hole temperature data, yet is rendered relatively unremarkable in the open-hole temperature data by downward flow in the borehole. In the examples provided, and in every borehole where we have compared open- and lined-hole data, the open-hole temperature logs differ distinctly from the lined-hole data above what would be interpreted from the open-hole results as the deepest major fracture. Below that point, the open-hole logs approach and eventually match the lined-hole profiles. This observation is consistent with the premise that the bottom of an open borehole typically acts like a cup, holding nearly stagnant water similar to the breached liner example (Figure S1, Supporting Information). A slightly deeper borehole could intersect another hydraulically conductive fracture, rendering the higher conduit inconsequential with regard to borehole flow, invisible to the open-hole temperature log, and leading to a different interpretation of which fractures control flow well above the limit of the borehole. In contrast, a lined-hole temperature log is independent of the intersection of deeper fractures and provides a superior representation of ambient conditions. Conclusions and Implications The high-resolution temperature profiling in lined holes presented in this article makes use of flexible liners that are relatively inexpensive and are usually easy to install and remove. Used alone, the liners serve to prevent borehole cross connection, and therefore create the static water columns suitable for the temperature profiling described in this article. At all four of the sedimentary rock sites presented, the lined-hole temperature profiles indicate many more hydraulically active fractures and a different interpretation of which fractures facilitate the most flow than do the open-hole profiles. Larger numbers of active fractures indicated in the lined holes are consistent with independent types of borehole information. The comparatively fewer number of active fractures detected in the open holes is most reasonably attributed to the masking effects of vertical flow resulting from hydraulic cross connection between fractures. The value of lined-hole temperature profiling is enhanced when profiling is done multiple times separated by days or weeks. The profile repetitions allow for application of the change log procedure and improve the potential for ranking the fractures in terms of degree of hydraulic activity. Open-hole temperature profiles usually identify two to five active fractures per hole regardless of hole depth. In conventional interpretations, these fractures are typically envisioned as the dominant conduits for groundwater flow; however, data collected from lined holes indicate they are commonly not the most hydraulically active fractures governing the ambient groundwater flow system. Therefore, at sites where the goal of borehole measurements is to understand contaminant distributions and transport, a conceptual model for the fracture network based solely on open-hole data is prone to misrepresentation. Furthermore, such misrepresentations may result in the selection of the wrong intervals for monitoring when using wells or multilevel systems and erroneous fracture densities for discrete fracture models of groundwater flow and transport. The temperature profiling method we have described currently has some limitations. First, in some unusual hydrogeologic circumstances, extreme hydraulic head variations over short intervals can cause inadequate sealing of the liner along parts of the hole. This loss of seal should be easily identified from the nature of the temperature profiles. Second, the method is not sufficiently sensitive to identify all hydraulically active fractures, and therefore the number of fractures identified in a given hole should be regarded as a minimum. Various independent lines of evidence at each of the four sites suggest more, in some cases many more, hydraulically active fractures are present than we have highlighted in the lined-hole data. Attention is now being directed at improving detection limits and confidence in the interpretation of the smaller temperature variations as well as working toward a quantitative analysis using numerical models to calibrate against the field data. Third, the method is effective only if the temperature of the water in the fractures is in disequilibrium with the surrounding rock. Thermal disequilibrium at depth requires a combination of natural or man-made influences that create both a temperature differential (e.g., near-surface heterothermic effects) and a driving force to move the water that is in thermal contrast through the fracture system, such as recharge or regional pumping. Although the ambient flow regime that drives the propagation of thermal disequilibrium will also be one of the dominant influences on contaminant transport, at sites where the maximum depth of interest is beyond the depth of thermal disequilibrium, the method is limited. However, the ALS technique (Pehme et al. 2007a) offers potential to overcome this limitation. NGWA.org P.E. Pehme et al. GROUND WATER 13

182 Acknowledgments Detlef Blohm, president of IFG, developed the temperature probe and made improvements as needs arose. Carl Keller, chief scientist at FLUTE, developed the flexible liner and provided assistance concerning liner use. The temperature profiles presented in the article were obtained in concert with M.Sc. thesis research by Leanne Burns, James Plett, Chris Turner, Jonathan Kennel, Diane Austin, and Jessica Meyer. Personnel from Montgomery Watson Harza assisted at the California site. Comments provided by Mary Anderson and three reviewers (Grant Ferguson, Gary Wealthall, and an anonymous individual) resulted in manuscript improvements. Funding support was provided by the Canada Foundation for Innovation and the Natural Sciences and Engineering Research Council of Canada via grants to B.L.P. and J.A.C. and the University Consortium for Field Focused Groundwater Contamination Research. Supporting Information Supporting Information may be found in the online version of this article: Table S1. Detailed characteristics of the boreholes discussed. Additional Application at Other Sedimentary Rock Sites: Additional descriptive text and analysis associated with the Simi and Madison data. Figure S1. Examples of temperature data from Simi, California. Figure S2. Examples of temperature data from Madison, Wisconsin. Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. References Advanced Logic Technologies (ALT) FAC40 Televiewer, Notes on use and operation, Advance Logic Technologies sárl. 31p. Anderson, M.P Heat as a ground water tracer. Ground Water 43, no. 6: Bejan, A Heat Transfer. New York: John Wiley & Sons. Berkowitz, B Characterizing flow and transport in fractured geological media: A review. Advances in Water Resources 25, no. 8 12: Bidaux, P., and C. Drogue Calculation of low-range flow velocities in fractured carbonate media from borehole hydrochemical logging data comparison with thermometric results. Ground Water 31, no. 1: Blohm, D President, IFG Corporation Inc. Phone interview, September 12, Brampton, Ontario. Bradbury, K.R., M.B. Gotkowski, D.J. Hart, T.T. Eaton, J.A. Cherry, B.L. Parker, and M.A. Borchardt Contaminant transport through aquitards: Technical guidance for aquitard assessment. Report 91133B. Denver, Colorado: American Water Works Foundation. Burns, L.S Fracture network characteristics and velocities of ground water, heat and contaminants in a dolostone aquifer in Cambridge, Ontario. Unpublished M.Sc. thesis, University of Waterloo. Carter, R.S., W.H. Steibel, P.J. Nalasco, and D.L. Pardieck Investigation and remediation of ground water contamination at a pesticide facility A case study. Water Quality Research Journal of Canada 30, no. 3: Cherry, J.A., B.L. Parker, and C. Keller A new depthdiscrete multilevel monitoring approach for fractured rock. Ground Water Monitoring and Remediation 27, no. 2: Coniglio, M Professor, University of Waterloo. Personal communication, March 22, Waterloo, Ontario. Davis, S.N Humboldt, Aargo, and the temperature of groundwater. Hydrogeology Journal 7, no. 5: Drogue, C Geothermal gradients and ground water circulations in fissured and karstic rocks: The role played by the structure of the permeable network. Journal of Geodynamics 4, no. 1 4: Drury, M.J Borehole temperature logging for the detection of water flow. Geoexploration 22, no. 3 4: Ferguson, G., A.D. Woodbury, and G.L.D. Matile Estimating deep recharge rates beneath an interlobate moraine using temperature logs. Ground Water 41, no. 5: Genthon, P., A. Bataille, A. Fromant, D. D Hulst, and F. Bourges Temperature as a marker for karstic waters hydrodynamics. Inferences from 1 year recording at La Peyrére cave (Ariège, France). Journal of Hydrology 311, no. 1 4: Greenhouse, J.P., and P.E. Pehme Monitoringthetemperature in a sleeved borehole: Implications for fracture detection. In Proceedings of the 55th Canadian Geotechnical and 3rd Joint IAH-CNC and CGS Ground Water Specialty Conferences,October 20-23, Niagara Falls, Ontario: Southern Ontario Section of the Canadian Geotechnical Society. Henninges, J., G. Zimmermann, G. Büttner, J. Schrötter, K. Erbas, and E. Huenges Wireline distributed temperature measurements and permanent installations behind casing. In Proceedings World Geothermal Congress Antalya, Turkey, April. Instruments for Geophysics Corporation (IFG) Temperature Logging, Tech Note Ref: Brampton, Ontario: IFG Corporation. Keys, W.S Borehole geophysics applied to groundwater investigations. In Techniques of Water-Resources Investigations of the United States Geological Survey, chap. E2, , Denver, Colorado: USGS. Keys, W.S., and R.F. Brown The use of temperature logs to trace the movement of injected water. Ground Water 16, no. 1: Malard, F., and R. Chapuis Temperature logging to describe the movement of sewage-polluted surface water infiltrating into a fractured rock aquifer. Journal of Hydrology 173, no. 1 4: Molson, J., P. Pehme, J. Cherry, and B. Parker Numerical analysis of HEAT transport within fractured sedimentary rock: Implications for temperature probes. In 2007 NGWA/U.S. EPA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Portland, Maine, September. 14 P.E. Pehme et al. GROUND WATER NGWA.org

183 National Research Council (NRC) Rock Fractures and Fluid Flow: Contemporary Understanding and Applications. Washington, DC: National Academy Press. Neuman, S.P Trends, prospects and challenges in quantifying flow and transport through fractured rocks. Hydrogeologic Journal 13, no. 1: Parker, B.L Investigating contaminated sites on fractured rock using the DFN approach. In NGWA/U.S. EPA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Portland, Maine, September 2007, Parker, B.L., D.B. McWhorter, and J.A. Cherry Diffusive loss of nonaqueous phase organic solvents from idealized fracture networks in geologic media. Ground Water 35, no. 6: Parker, B.L., R.W. Gillham, and J.A. Cherry Diffusive disappearance of immiscible phase organic liquids in fractured geologic media. Ground Water 32, no. 5: Pehme, P., J.P. Greenhouse, and B.L. Parker. 2007a. The Active Line Source temperature logging technique and its application in fractured rock hydrogeology. Journal of Environmental and Engineering Geophysics 12, no. 4: Pehme, P.E., J.P. Greenhouse, and B.L. Parker. 2007b. The Active Line Source (ALS) technique, a method to improve detection of hydraulically active fractures and estimate rock thermal conductivity. In Proceedings of 60th Canadian Geotechnical Conference & 8th Joint IAH-CNC Ground water Conference, Ottawa, Ontario. Pehme, P.E., B.L. Parker, J.A. Cherry, and J.P. Greenhouse. 2007c. The potential for compromised interpretations when based on open borehole geophysical data in fractured rock. In 2007 NGWA/U.S. EPA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Portland, Maine, September. Perrin, J., B.L. Parker, and J.A. Cherry Evidence for a non-karstic flow system within a confined karstic and fractured dolostone aquifer. Journal of Hydrology submitted Prensky, S Temperature measurements in boreholes: An overview of engineering and scientific applications. Log Analyst 33, no. 3: Robinson, R., S. Stilliman, and C. Cady Identifying fracture interconnections between boreholes using natural temperature profiling: II. Application to fractured dolomite. Log Analyst 34, no. 1: Sammel, E.A Convective flow and its effects on temperature logging in small diameter wells. Geophysics 33, no. 6: Sara, M.N Site Assessment and Remediation Handbook, 2nd ed. New York: Lewis Publishers. Silliman, S., and R. Robinson Identifying fracture interconnections between boreholes using natural temperature profiling: I. Conceptual basis. Groundwater 27, no. 3: Sterling, S.N., B.L. Parker, J.A. Cherry, J.H. Williams, J.W. Lane Jr, and F.P. Haeni Vertical cross contamination of trichloroethylene in a borehole in fractured sandstone. Ground Water 43, no. 4: Trainer, F.W Temperature profiles in water wells as indicators of bedrock fractures. Professional Paper 600-B. Washington, DC: USGS. VanderKwaak, J.E., and E.A. Sudicky Dissolution of non-aqueous-phase liquids and aqueous-phase contaminant transport in discretely-fractured porous media. Journal of Contaminant Hydrogeology 23, Wisian, K.W., D.D. Blackwell, S. Bellaniti, J.A. Henfling, R.A. Normann, and P.C. Lynse Field comparison of conventional and new technology temperature logging systems. Geothermics 27, no. 2: NGWA.org P.E. Pehme et al. GROUND WATER 15

184 1 2 3 Improved Resolution of Ambient Flow through Fractured Rock with Temperature Logs (Supplemental Information) P.E. Pehme 1, B.L. Parker 2, J.A. Cherry 3 and J.P. Greenhouse 4 4 Table S1: Detailed characteristics of the boreholes discussed. Site Approximate Depth to (m) Hole Surface Ambient Bottom Bottom Dia. MASL Water of (cm) Level Casing Cambridge, ON Guelph, ON Simi, CA 20 ~ Madison, WI Application at Other Sedimentary Rock Sites Although the nature of the comparisons between the temperature profiles collected in lined and open holes at the sites in Simi, CA and near Madison, WI (Figures S1 and S2 respectively) are similar to those observed at the Cambridge and Guelph sites, these sites provide additional insights. A color and symbol scheme consistent with the main text, including arrows of varying size, representing relative amounts and direction of flow as well as red for open-hole and blue for lined-hole temperature data are used in these figures The general characteristics of the temperature profiles from the Simi site (sandstone with mudstone interbeds; Williams et al. 2002) are much different than those from the Cambridge and Guelph dolostone. The topographically driven downward crossconnected flow in the open hole at the Simi site results in much of the open hole temperature profile appearing as a series of linear segments with gradually increasing values with depth and low variability (less than 0.1 C). The uniform segments are separated by five abrupt temperature increases, or steps. A reasonable interpretation of the open borehole data is that water enters the borehole at each of the steps and 1

185 moves downward, gradually warming but in thermal disequilibrium with the formation until the next major flowing fracture is reached where the process repeats. The abruptness of the temperature changes implies individual fractures dominate water movement into the borehole. In contrast, using the size of the temperature variations as an indicator of the relative contribution of water, none of the variations within the FLUTe liner (Figure S1c, d) are as sharp as in the open-hole case, implying flow occurs over a series of fractures (a zone) rather than individual distinct discontinuities in the rock. Over the 55 m interval of hole at the Simi site, 22 flow zones are indicated by the lined-hole profile, which is many more than can be identified based on the open borehole logs In late September, 2003 open- and lined-hole measurements were obtained in MP6 at the Wisconsin site. The process was repeated and a second set of open- and lined-hole data were collected in mid-december, Meyer et al. (2007) describe the hydrogeology of this study area and indicate the geologic units consist of Cambrian- Ordovicran inter-layered, nearly horizontal sandstone and siltstone with a dolostone unit. Groundwater flow is primarily horizontal in the sandstone beds with downward leakage towards the Mount Simon Formation, which is a regional aquifer. During the September episode, the liner had a leak as indicated by the inability to raise the water level inside the liner above the level in the formation. Comparison of the lined-hole temperature profiles (September vs. December) clearly shows the location of the leak; moreover, the lined- and open-hole profiles in September are identical above the level (roughly 212 masl) where the leak prevents liner seal, while below they appreciably diverge. The actual outflow must be slightly above the level where this divergence occurs because some head differential is needed to create liner seal below the leak. This differential occurs because overall the head in the formations decrease with depth (Meyer et al. 2007). This example demonstrates: (i) a temperature profile inside a liner can be used to 2

186 determine the approximate depth of a liner leak, which is helpful in identifying the point where liner repair is needed, and (ii) useful temperature profile interpretation can be obtained from the sealed portion below the leak. This contrast between unsealed and lined-hole data in the same log supports the premise that a sealed hole provides a much different thermal perspective of the borehole The most distinct feature of the lined-hole temperature profile at the Wisconsin site is a major flow zone indicated at an elevation of 214 masl (Figure S1c). This is likely the zone of most active ambient groundwater flow at this borehole location, yet there is no indication of its existence based on inspection of the open-hole profiles. In addition to this major feature, nine other flow zones were denoted from the lined-hole profiles. Other lines of evidence at this site suggest many more flow zones in addition to those identified using the existing interpretation procedures. The sections of the open-hole temperature logs that have an irregular and highly variable response (e.g., m) are without general coherence between repeat logs and are therefore not interpreted as individual fractures, although some likely occur in this depth range The data from the Simi and Madison sites re-enforce the conclusions presented in the main text, in particular that the improved interpretation of lined temperature data is a general situation rather than specific to Southern Ontario. These examples also show that the techniques have application in varying geologic and hydrogeologic settings. Supplemental References Meyer, J.R., B.L. Parker, and J.A. Cherry Detailed hydraulic head profiles as essential data for defining hydrogeologic units in layered fractured sedimentary rock. Environmental Geology, doi: /s Williams, J.H., J.W. Lane Jr., K. Singha, and F.P. Haeni Application of advanced geophysical logging methods in the characterization of a fractured sedimentary 3

187 bedrock aquifer, Ventura County, California. USGS Water-Resources Investigations Report Boulder, Colorado: U.S. Geological Survey. 4

188 75 76 Figures Figure S1: Examples of temperature data from Simi, CA (data plotted relative to depth in metres from ground surface): (a) gamma log, (b) temperature profiles collected in open holes (red) and lined (blue) hole, and variability logs for lined- (c) and open- (d) hole conditions. Arrows schematically represent zones of flow, amount (by size), and direction (orientation), and are referred to in the text (red for open hole and blue for lined hole). 5

189 Figure S2: Examples of temperature data from Madison, WI, including a, b & c) stratigraphy, hydrogeologic units (HG), and gamma log from Meyer et al. (2007), d) lined hole temperature logs from 27/09/2003 (red) and 9/12/2003 (purple), open hole temperature logs from 30/09/2003 (blue) and 13/12/2003 (teal), e) lined hole variability logs from 27/09/2003 (red) and 9/12/2003 (purple), f) open hole variability logs from 30/09/2003 (blue) and 13/12/2003 (teal), and g) lined hole change log. Arrows schematically represent zones of flow, amount (by size), and direction (orientation), and are referred to in the text (red for open hole and blue for lined hole). 77 6

190 Environ Geol (2008) 56:27 44 DOI /s ORIGINAL ARTICLE Detailed hydraulic head profiles as essential data for defining hydrogeologic units in layered fractured sedimentary rock Jessica R. Meyer Æ Beth L. Parker Æ John A. Cherry Received: 15 September 2007 / Accepted: 12 November 2007 / Published online: 7 December 2007 Ó Springer-Verlag 2007 Abstract This paper describes a study in southern Wisconsin where vertical hydraulic head profiles measured in exceptional detail provided the key data for defining hydrogeologic units (HGUs) in a layered sequence of sandstone, siltstone, shale, and dolostone. The most important data were obtained from corehole MP-6 which was cored 131 m into bedrock and instrumented using a Westbay 1 multilevel system with 36 depth discrete monitoring intervals. The resulting head profile is consistant over time and shows eight distinct inflections in hydraulic head. Several of the inflections occur between adjacent permeable units and are likely due to poor vertical connectivity of fracture sets rather than distinct lower permeability layers or aquitards in the conventional sense. No other type of data was capable of identifying the position of such distinct hydrogeologic features. These zones of abrupt head loss provide the primary dataset for delineation of eleven HGUs at MP-6 and are supported by less detailed head profiles at other locations. Although the detailed head profiles are essential, core logs and geophysical logs from other boreholes are nessessary to fully establish the lateral continuity of the HGUs. Keywords Hydrogeologic unit Fracture Sedimentary rock Hydraulic head Profile J. R. Meyer (&) B. L. Parker J. A. Cherry Department of Earth Sciences, University of Waterloo, 200 University Ave W, N2L 3G1 Waterloo, ON, Canada jmeyer@scimail.uwaterloo.ca Present Address: B. L. Parker School of Engineering, University of Guelph, N1G 2W1 Guelph, ON, Canada Introduction In the 1960s, the focus of groundwater research turned to analysis and modeling of regional scale flow systems (Toth 1962, 1963; Freeze and Witherspoon 1966, 1967). For example, Toth (1963) demonstrated the influence of topography on flow paths which led to the definition of regional, intermediate, and local groundwater flow systems. Toth s (1963) research was expanded on by Freeze and Witherspoon (1966) who introduced a numerical model to investigate idealized regional groundwater flow patterns. Later, Freeze and Witherspoon (1967) showed vertical 2D cross-sections from numerical simulations where lines of equal head are refracted at the interfaces between units of contrasting hydraulic conductivity. This refraction represents poorer hydraulic connection between the units making them hydraulically distinct. Several terms have been used to describe hydraulically distinct units in the subsurface. In the broadest hydrogeologic context, the subsurface groundwater flow system is sub-divided into aquifers and aquitards. Meinzer (1923) defined the term aquifer as a water-bearing unit comprised of a formation, group of formations, or part of a formation. The term aquitard, first appeared in the hydrogeologic literature in the 1960s (Davis and DeWiest 1966) for which Cherry et al. (2006) provide the definition geologic deposits of sufficiently low hydraulic conductivity and sufficient a real extent, thickness and geometry, to impede groundwater flow between or to aquifers. These two terms are relative and thus are most suitable for sitespecific use when groundwater flow and yield are the main issues. Other terms have since been introduced to describe partitions within the groundwater flow regime. For example, Maxey (1964) introduced the term hydrostratigraphic unit to define bodies of rock with considerable lateral 123

191 28 Environ Geol (2008) 56:27 44 extent that compose a geologic framework for a reasonably distinct hydrologic system. Seaber (1988) provides a thorough review on the concept of hydrostratigraphic units as well as their potential place in the formal stratigraphic code. More recently, Anderson (1989) used the term hydrogeologic facies to describe homogeneous but anisotropic units with hydrogeologic meaning both with respect to modeling and field experiments. Later, Poeter and Gaylord (1990) introduced the term hydrofacies to describe 3D aquifer elements that could be used as the framework for simulation of groundwater flow and contaminant transport in numerical models. In this paper, we adopt the term hydrogeologic unit (HGU) to represent partitions of the groundwater flow domain that are hydraulically consistent at a specified spatial scale. The geometries of the HGUs may be as simple as a series of flat lying layers or they may have a much more complex geometry. However, they must have measurable thickness and aerial extent in order to be detected and monitored. Ideally, the geometry of the HGUs should be consistent with the depositional setting and any post-depositional processes which have shaped the geometry and hyraulic charactieristics of the sedimentary deposits. Although HGUs are hydraulically distinct in some manner within the context of the site specific system, each unit is connected to some degree to the adjacent units and the nature of this connection is important. Considering the requirements given above, the term hydrogeologic unit seems appropriate because it describes units whose delineation relies on both detailed hydraulic data and detailed geologic data at a scale appropriate to the problem or question. However, regardless of the terminology adopted, an important goal in hydrogeologic studies is segmentation of the groundwater zone, using field derived data, into units that provide the framework for describing groundwater flow and in many cases, contaminant distribution, migration, and fate. In this paper, we focus on defining and delineating HGUs in an area of groundwater contamination within a layered sequence of sandstone, dolostone, siltstone, and shale common to the midwestern United States with particular emphasis on the key role of vertical hydraulic head profiles. HGUs can be differentiated by hydraulic discontinuities that are generally produced by the influence of geologic variability on the flow system. Therefore, the characterization of this geologic variability in a context that is hydraulically meaningful is important to all studies of groundwater flow. There are many means by which the geological variability and its potential influence on the flow system are measured or observed. Geological methods rely primarily on core or outcrop measurements and observations to provide insight into the geological variability. However, geologic data cannot provide direct hydraulic information. Although hydraulic methods may provide more direct hydraulic information most only provide horizontally-oriented data. Geophysical methods are used to infer information regarding both geologic and hydraulic properties but are indirect data requiring site specific calibration to improve the reliability of interpretation. So the question is, what combination of tools provides the essential data for defining HGUs? Various studies have focused on describing methods for defining the hydrogeologic framework for sedimentary rock. For example, Macfarlane et al. (1994) used many methods to delineate regional scale hydrostratigraphic units within the Dakota aquifer system in Kansas, some of which included: sequence-stratigraphy, borehole geophysical logs, laboratory testing of core samples for hydraulic conductivity, and in situ hydraulic testing using wells. Muldoon et al. (2001) used detailed outcrop and core stratigraphy, short interval borehole packer hydraulic conductivity testing, flow metering, and a variety of other geophysical logs to identify 14 high permeability zones within a section of the Silurian dolomite aquifer in northeastern Wisconsin. Runkel et al. (2006) performed an investigation focused on re-defining the hydrogeologic framework of the Cambrian system in southeastern Minnesota. Their complementary data sets included many tests of porosity and permeability on core plugs; outcrop, borehole, and core observations of secondary porosity features; flowmeter logs and other geophysical logs. Although each study listed above used a variety of data sets to delineate HGUs, some very detailed, none directly utilized hydraulic head profiles to delineate HGUs. The first multilevel systems (MLSs) capable of measuring accurate, detailed head profiles and functioning at substantial depth ([75 m) became commercially available in the early 1980s. Currently, these systems include the Westbay 1 MP system (Black et al. 1986), the Solinst 1 Waterloo system (Cherry and Johnson 1982), and the Water FLUTe TM system (Cherry et al. 2007). These MLSs allow for accurate head measurements in many depth discrete intervals within a single borehole. Einarson (2006) provides an overview of depth-discrete multilevel monitoring methods. Although MLS have long been available, they have not commonly been used to define and delineate HGUs based on the vertical pattern of hydraulic head. The ultimate goal of investigations at this study area is to achieve sufficient understanding of the flow system in order to predict the migration and fate of the organic contaminants in the bedrock using 3D numerical models. Regional numerical models for groundwater flow inclusive of the study area have been constructed previously (McLeod 1975; Krohelski et al. 2000). However, the hydrogeologic framework available for the regional groundwater flow model was not adequate to represent the 123

192 Environ Geol (2008) 56: salient details of the groundwater flow system and observed variability in contaminant mass distribution within the source zone and plume. Consequently, the first objective of the field study was to determine if exceptionally detailed hydraulic head profiles from commercially available MLS would alter or refine the number and/or position of HGUs. The second objective was to better understand the geologic controls on the physical characteristics of the HGUs and the interfaces between them. This understanding provides a means to estimate each unit s lateral continuity, geometry, and thickness in addition to providing information used to estimate the magnitude and direction of groundwater fluxes between such units. Improved delineation and quantification of the flow system should result in better prediction of groundwater flow and plume behavior. Study area The field study area is located in Dane County, south central Wisconsin, in the village of Cottage Grove approximately 21 km east of Madison, Wisconsin (Fig. 1a). The study area occupies approximately 7.3 km 2 of once glaciated terrain characterized by flat areas between broad northeast to southwest trending drumlins (Fig. 1b). Groundwater flow is generally from west to east across the study area (Fig. 1b). This paper focuses on data collected from four research coreholes, MP-5, MP-6, MP-8, and MP-15, shown in Fig. 1b. The study area includes a zone of subsurface DNAPL near a chemical facility and a dissolved-phase plume beneath the wetland areas and agricultural land (Fig. 1b). The contamination is composed of a mixture of organic contaminants estimated to have entered the subsurface between 1950 and 1970 (Hydro-Search Inc. 1989). The study area overlies approximately 175 m of Cambrian and Ordovician age sandstones, siltstones, and dolostones draped by Quaternary glacial deposits, described in Fig. 2, all of which overlie igneous and metamorphic Precambrian rocks. The sedimentary rocks in Dane County generally dip only 2 3 m per kilometer to the southwest, south, and southeast forming the central section of the Wisconsin Arch (Cline 1965). The well-sorted, clean, friable, fine- and medium-grained sandstones (Ostrom 1978)of the Mt Simon Formation serve as a major water supply aquifer in the area (Bradbury et al. 1999). In contrast, the shale and siltstone sections of the Eau Claire Formation are considered to form a regional scale aquitard (Bradbury et al. 1999). At the study site, the lateral extent of groundwater contamination is greatest within the Reno Member of the Lone Rock Formation. The Reno Member is primarily composed of very fine- to medium-grained glauconitic and feldspathic sandstone (Odom 1978) but can be further differentiated into three sublithologies described in Fig. 2. Methods Four boreholes, MP-5, MP-6, MP-8, and MP-15, were drilled through the unconsolidated deposits and into bedrock to total depths ranging between 96 and 152 m below ground surface (bgs). MP-6 was positioned upgradient of the contamination to allow for extensive open borehole testing while avoiding concerns about cross-contamination due to vertical flow (Fig. 1b). MP-5 and MP-15 were placed in the middle of the plume and MP-8 was located at the leading edge of the plume to investigate the internal variability of the contamination (Fig. 1b). Drilling and geologic logging methods Coreholes MP-5, MP-6, and MP-8 were drilled using a Cantera 250 mud rotary drill rig. The unconsolidated deposits, which range in thickness between 7 and 21 m at Fig. 1 a Study area location within Wisconsin, b the study area 123

193 30 Environ Geol (2008) 56:27 44 Fig. 2 Study area stratigraphy if no reference is given, the descriptions are derived directly from core observations MP-5, MP-6, and MP-8, were drilled using a mud rotary technique with a 25.4 cm (10 in.) tricone bit. Once the top of bedrock was reached, a permanent casing was installed from ground surface into the upper meter of bedrock. MP-6 and MP-8 were continuously cored from the top of rock to their completion depths using an HQ3 wireline coring system which provides a 6.1 cm (2.4 in.) diameter core and a nominal 9.7 cm (3.8 in.) diameter corehole. MP-5 was continuously cored using the HQ3 wireline coring system from the top of rock until 81 m bgs when the system was switched to the NQ3 wireline coring system for the remainder of the corehole (to 106 m bgs). The NQ3 wireline coring system provides a 4.6 cm (1.8 in.) diameter core and a nominal 7.6 cm (3 in.) diameter corehole. The continuous coring at MP-5, MP-6, and MP-8 was performed using an air/water mist as the primary drilling fluid. The MP-15 corehole was drilled using an air rotary technique from ground surface to approximately 50 m bgs. The MP-15 corehole was continuously cored from 50 to 65 m bgs. However, due to poor core recovery and time constraints MP-15 was drilled using an air hammer from 65 m bgs to its completion depth of 96 m bgs. Geologic logging of the continuous cores from MP-5, MP-6, and MP-8 was performed by a variety of collaborators including Austin (2005), Aswasereelert (2005), Meyer (2005), Swanson (2007), David LePain of the Wisconsin Geological and Natural History Survey (WGNHS), and GeoTrans, Inc. Lithology, sedimentary structures, and fractures were logged in detail. An effort was made to log 123

194 Environ Geol (2008) 56: characteristics that might help to distinguish between fractures that appeared to be hydraulically-active and those that were likely breaks induced by the drilling. For example, staining indicative of oxygenated water, precipitates, and weathering on fracture surfaces were taken to indicate potentially hydraulically active fractures. In addition, each fracture was classified into three broad categories representing orientation: horizontal, oblique, or vertical to the core axis. Matrix physical property measurement methods In order to characterize the rock matrix, samples were taken from each corehole and each lithology type for physical property measurements. A section approximately 30 cm in length was separated from the core using a hammer and chisel and then subcored resulting in a sample with a nominal diameter of 3.8 cm (1.5 in.) and an approximate length of 5 cm (2 in.). After the samples were subcored, they were dried in a 100 C oven from 24 to 48 h to remove the porewater after which they were stored in a desiccator. The samples were then weighed and the exact diameter and length of each sample was measured using a Vernier caliper. The matrix permeability of each sample was measured in the vertical axial direction using a nitrogen gas permeameter and the ASTM standard method D (ASTM 2004). Porosity was measured using the imbibition method described by Collins (1961). Geophysics and hydraulic testing methods MP5, MP-6, and MP-8 were logged using the combined 2PGA-1000 and 2PEA-1000 probes. The combination of the two probes allows for concurrent measurement of natural gamma, single point resistance, spontaneous potential, and normal resistivity with electrode spacings of 20.3 cm (8 in.), 40.6 cm (16 in.), 81.3 cm (32 in.), and cm (64 in.). A physical caliper log was also measured in MP-5, MP-6, and MP-8 using a CLP-230 threearmed caliper. A HFP-2(4)293 heat pulse flow meter was used to measure the amount of vertical flow within each of the open coreholes. In addition, the MP-6 corehole was logged with an ALT OBI optical televiewer. An MGX II console was used to communicate with each of the downhole instruments described above. Both the downhole probes and the console are available from Mount Sopris Instrument, Inc. Detailed straddle packer slug testing was also performed across intervals 0.7 m in length in the MP-6 corehole using the slug created by the packer inflation, a method similar to that described by Muldoon (1999). Multilevel systems Once all geophysical logging and open hole hydrualic testing was completed in each of the coreholes, MLSs were designed based on the complementary data sets described above and then installed. An open-tube Water FLUTe TM system with 11 monitoring intervals was installed in the MP-5 corehole in October However, the bottom section of the hole collapsed prior to installation preventing access to the two deepest ports. An open tube Solinst 1 Waterloo system with 13 monitoring intervals was installed in the MP-8 corehole in August Water levels were measured manually with a small diameter water level tape in both the MP-5 and MP-8 multilevels. In addition, a Westbay 1 MP38 system with 36 monitoring intervals was installed in the MP-6 corehole in December 2003 and a Westbay 1 MP38 system with 16 monitoring intervals was installed in the MP-15 corehole in December Though each of the multilevels described above provided valuable data, this paper relies heavily on the exceptionally detailed profile of hydraulic head provided by MP-6. Therefore, a detailed description of the MP-6 multilevel installation is provided below, which also applies to the MP-15 multilevel. The MP-6 multilevel design targeted features observed in the continuous core logs as well as the geophysical logs. The HQ3 drill rods were used as a guide tube in order to keep the entire length of the corehole open during installation. The system was assembled one component at a time as it was lowered down the corehole. Each of the joints between components, the measurement ports, and the pumping ports were pressure tested just prior to lowering into the corehole to confirm there was no leakage. Once the entire system was lowered into the corehole to the appropriate depth, inflation of the packers began. The drill rods were pulled up in increments exposing one or more packers. The packers were inflated using water pressure and tests were performed to verify the packers were sealing and that the system casing was not leaking. After installation, pressure measurements were made in each of the 36 monitoring zones on four separate occasions over two years. For details regarding the methods for measuring pressures in a Westbay 1 system see Once each profile of shut-in monitoring zone pressures was measured, water levels (equivalent piezometric levels), with respect to ground surface, were calculated for each monitoring zone using Eq. (1). The water levels are then converted to total hydraulic head, simply referred to hereafter as head, referenced to mean sea level using the surveyed elevation of the ground surface near the well. 123

195 32 Environ Geol (2008) 56:27 44 Equiv. Piezometric level (bgs) ¼ depth of port (bgs) monitoring zone pressure ambient air pressure ð Þ density of water ð1þ Accuracy assessment The uncertainty associated with the heads calculated for the Westbay 1 systems using Eq. 1 are assessed by addressing the relative uncertainty in the density of water, pressure measurements, and port depths between closely spaced monitoring intervals. The relative uncertainty in the density of water between two closely spaced monitoring intervals is taken to be negligible. The density used in the calculations is 1 g/cm 3, the density of pure water at approximately 4 C. The relative uncertainty associated with the measured pressures is ±0.7 cm of head (Westbay 1, personal communication, 2007). Port depths are determined using the measured lengths of the Westbay 1 System casing components. The relative uncertainty in the port depths due to uncertainty in the measured lengths of the Westbay 1 System casing components is approximately ±0.1 cm (Westbay 1, personal communication, 2007). In addition, deviation of the borehole from vertical can also contribute uncertainty to the port depths. However, data derived from the MP-6 borehole televiewer confirms that uncertainty in the port depths between two closely spaced monitoring intervals due to borehole deviation is negligible. Therefore, the total relative uncertainty of the calculated head values between two closely spaced monitoring zones is ±0.8 cm or about ±1 cm. Head values derived from manual water level measurements taken from an open tube Waterloo system, MP- 8, and an open tube water FLUTe TM system, MP-5, were made using a small diameter water level tape graduated every meter and a measuring tape graduated every 0.1 cm with estimated error between ±3 and 5 cm. The estimated error for the manual water level measurements is relatively large due to problems specific to the installations at this study area and not to the multilevels themselves. Results and discussion Detailed vertical head profiles Figure 3 shows each of the four hydraulic head profiles measured from the MP-6 Westbay 1 multilevel. The shape of the profiles, including the major head changes labeled (A) through (H), is consistent for four measurement episodes during two years; therefore, each of the profiles is approximately representative of steady-state conditions at MP-6. The December 2005 profile is used as an example in the following discussion because it is the most recent. In December 2005, a total of 4.3 m of head was lost across the m section monitored by the MLS. The head differential between adjacent ports for each of the inflections labeled in Fig. 3 for December 2005 range between 10 and 148 cm. Each differential is outside the estimated error associated with the calculated head values. The vertical gradient was calculated over the distance between the midpoints of adjacent intervals for each of the eight major inflections in the December 2005 head profile. The minimum and maximum distance between the midpoints of adjacent intervals used in the gradient calculations were 2.4 and 3.7 m, respectively. The vertical hydraulic gradient for the eight inflections labeled A through H ranges between and -0.6 with negative values indicating a downward vertical gradient. Alongside the head profile is the percentage of head loss across each inflection compared to the overall head loss across the system in December The individual inflections contribute between 2 and 35% of the overall head loss across the system. The remainder of the head loss is attributed to the overall downward vertical gradient at MP-6. Approach for delineating hydrogeologic units The distinct inflections in the MP-6 head profiles shown in Fig. 3 suggest that the system is not one hydraulic unit but rather is comprised of many stacked HGUs. In many cases, the HGUs are not separated by conventional aquitards but instead are separated by an interface that represents poor vertical hydraulic connection between the units. Each of these units and the interfaces between the units are described in detail in the following sections. The eight major inflections of the MP-6 head profile shown in Fig. 3 served as the basis for the HGU delineation. However, other types of data including bulk hydraulic conductivity data, rock matrix hydraulic conductivity and porosity, continuous core geologic logs, and geophysical logs provided support and helped to establish the location of each interface between the adjacent monitoring zones. Between the graphical stratigraphic log and the hydraulic head profile in Fig. 3 is a profile showing the HGUs delineated locally at MP-6. The original HGU delineation presented by Austin (2005) and Meyer (2005) included eight bedrock units. The delineation presented here is an extension of the original delineation made possible by additional data and additional inspection/interpretation of the hydraulic head profile, geophysical logs, and continuous core logs. Thirteen bedrock HGUs have been identified at the study area, 11 of which are present at MP-6. In 123

196 Environ Geol (2008) 56: Fig. 3 Head profiles measured from the MP-6 multilevel well in December 2003, July and August 2004, and December

197 34 Environ Geol (2008) 56:27 44 addition, three HGUs identified in the unconsolidated glacial material overlying the bedrock are not addressed in this paper. Bedrock HGUs 12 and 13, both part of the St Peter Formation (Fig. 2), are not discussed in detail because they were not encountered at the MP-6 location. In order to understand flow within and/or between the HGUs, flow is considered to occur in several general domains: the fractures, the matrix blocks between fractures, or a combination of the two. Comparison of bulk hydraulic conductivity data obtained from many locations in the study area as well as matrix hydraulic conductivity values measured on core samples from each of the units suggest that HGUs 11, 10, 9, 8, 7, 6, 5, and 2 are dominated by fracture flow, HGUs 4 and 1 are characterized by a combination of fracture and matrix flow, and HGU3 is dominated by matrix flow. Identification of the HGU11/HGU10 and HGU10/HGU9 interfaces The inflection in the MP-6 hydraulic head profile, labeled (A) on Fig. 3 serves as the primary evidence for the delineation of HGUs 10 and 11. Both HGU10 and HGU11 are fractured dolostones. However, several data sets show that HGU10 is likely karstic or much more densely fractured than HGU11. For example, during drilling, the rods dropped at least 1.5 m in the interval between 243 and 246 m AMSL. In addition, the MP-6 caliper log extends from a nominal 9.7 cm (3.8 in.) diameter to 64 cm (25 in.) between and m AMSL and to 56 cm (22 in.) between and m AMSL. Several other authors have noted karstic features in the Prairie du Chien (Palmquist 1969; Smith and Simo 1997; Bradbury et al. 1999). Karstic features were not observed in either the MP-5 or MP-8 corehole. There are fewer monitoring intervals in the MP-6 multilevel in the upper 17 m of bedrock because the Westbay 1 packers could not be positioned within HGU10 due to the variability of the borehole diameter. As a result, the head profile does not provide detail regarding the interface between HGU10 and HGU9. However, geologic and geophysical data sets strongly suggest an HGU contact between the karstic portion of the Prairie du Chien Group dolostones and the lower Prairie du Chien Group combined with the Jordan Formation sandstones at MP-6. Therefore, although not confirmed with sufficient resolution in the head profile, HGU10 and HGU9 are represented as distinct units using a dashed line. flow between HGU9 and HGU8. HGU9 is composed of the sandstones of the Jordan Formation. In contrast, HGU8 is composed of dolostones and siltstones of the St Lawrence Formation. Two peaks in natural gamma within HGU9 reflect the general lithology change from feldspathic sandstones in HGU9 (Odom 1975; Odom and Ostrom 1978) to the dolostones and siltstones of HGU8 (Fig. 4). Odom and Ostrom (1978) indicate that in many cases a significant unconformity exists between the St Lawrence Formation and the Jordan Formation near the Wisconsin Arch. A sediment-filled bedding parallel fracture observed in the MP-6 optical televiewer log may be indicative of this unconformity. If an unconformity exists between HGU9 and HGU8, it may function as a mechanical interface limiting the propagation of vertical fractures from one unit into the other which would only enhance the limited vertical connection between the two units (Cooke et al. 2006). Given that both HGU9 and HGU8 are dominated by fracture flow, this hydraulic resistance is likely due to poor vertical connection between the fracture networks of the two units. Identification of the HGU8/HGU7 interface The head change labeled (C) on Fig. 3 accounts for 35% of the overall head loss across the system and is indicative of Identification of the HGU9/HGU8 interface The pronounced head drop across the HGU9 and HGU8 contact labeled (B) in Fig. 3 indicates vertical resistance to Fig. 4 Comparison of borehole natural gamma logs from MP-5, MP- 6, and MP-8 123

198 Environ Geol (2008) 56: Fig. 5 Comparison of multilevel head profiles measured at MP-6, MP-15, MP- 5, and MP-8 (head profiles were not measured during the same time period and the scale for MP-5, MP-15, and MP-8 is different from the scale for MP-6) the contact between HGU8 and HGU7. The HGU8/HGU7 contact is relatively easy to interpolate across the study area because it correlates directly with a major shift in lithology easily distinguished in borehole natural gamma logs as shown in Fig. 4. Head change (C) was also observed, although not with the same resolution, at the MP- 5 and MP-8 multilevels as shown in Fig. 5. The overall head loss across the HGU8/HGU7 interface is likely, primarily due to poor connection between the fracture networks of the St Lawrence Formation dolostones, HGU8, and the upper Tunnel City Group sandstones, HGU7. Fracture logs from the continuous core at MP-6 show that oblique and vertical fractures were much more abundant in HGU8 than in HGU7 (Fig. 6). Rather, HGU7 is dominated by bedding parallel fractures with negligible dip. Identification of the HGU7/HGU6 and HGU6/HGU5 interfaces Two inflections, D and E, in the MP-6 head profile within the Tunnel City Group sandstones provide the basis for splitting this stratigraphic group into three HGUs. Inflection D occurs across the interface between HGU7 and HGU6. HGU7 is a distinct section approximately 3 m thick of the thinly laminated sandstone sublithology of the Reno Member Fig. 6 MP-6 core derived fracture frequency for HGU8 through HGU4 123

199 36 Environ Geol (2008) 56:27 44 dominated by horizontal bedding parallel fractures (Swanson 2007) (Fig. 2). The head profile within HGU7 appears to be transitional between HGU8 and HGU6 because only one monitoring zone was located in HGU7. However, other data suggest HGU7 is a distinct unit and not simply a transition from HGU8 to HGU6. HGU6 corresponds to a thick, 5 8 m, storm deposit (Swanson 2007) in which the three major sublithologies of the Reno Member cycle frequently (lithologic descriptions in Fig. 2); although the bioturbated sublithology dominates this portion of the section. Bedding parallel fractures form both within the laminated sublithology of the unit as well as at the boundary between sublithologies (Swanson 2007). These bedding-parallel fractures have been shown to be laterally extensive and hydraulically conductive (Swanson and Bahr 2004). The contact between HGU7 and HGU6 is also supported by head profiles at MP-5 and MP-8 and the resistivity logs from all three coreholes. For example, Fig. 5 shows the combined effect of head changes (C) and (D) across the interface between HGU7 and HGU6 at MP-5 and MP-8. The resistivity logs show a peak in resistivity across HGU7 followed by a decreased and consistent resistivity signal throughout HGU6. Although the exact cause for the change in the resistivity signal is not known, the change is correlatable and consistent with lithology changes observed in the continuous core at all three coreholes and described above (Fig. 7). Head inflection E occurs across the interface between HGU6 and HGU5. HGU5 corresponds to a m thick section of laminated sandstones which are coarser-grained, contain less glauconite, and have less pronounced bedding than those in HGU7. HGU5 is also coincident with a depression in the natural gamma log (Fig. 4). Odom (1975) showed with petrographic analysis that the Cambrian sandstones of the upper Mississippi Valley are primarily composed of quartz, potassium feldspar (K-feldspar), and glauconite. Both glauconite and K-feldspar contain potassium, which Swanson (2007) has shown, using spectral gamma analysis, is the primary source of gamma radiation in the Tunnel City Group in south-central and south-western Wisconsin. Odom (1975) also showed that as mean grain size increased, the percent by volume of K-feldspar decreased because most K-feldspar grains were less than mm in diameter. Therefore, the depression in the natural gamma signal is indicative of the coarser grained, glauconite poor, less feldspathic Mazomanie Formationlike lithologies. In addition, Fig.6 shows that HGU5 is characterized by an increase in vertical and oblique fractures at MP-6 compared to HGU6 and HGU7 which further supports the distinction between HGU5 and HGU6. An inflection in the MP-8 head profile shown in Fig. 5 also provides support for the contact between HGU6 and HGU5. However, the head difference between the two points is only 6.7 cm which may be encompassed by the error for manual water level measurements at MP-8. The Tunnel City Group is particularly important at the study area because the majority of the presistent DNAPL phase and an extensive dissolved phase plume are observed within the upper section of the Tunnel City Group. The contact between HGU6 and HGU5 is particularly important because previous studies have concluded that the DNAPL did not penetrate beneath HGU6 (HSI GeoTrans 1999). The apparent cessation of the DNAPL s downward movement at the HGU6/HGU5 interface is consistent with the conclusion of a poor vertical hydraulic connection between HGU6 and HGU5. Identification of the HGU5/HGU4 interface Fig. 7 Comparison of single point resistance log across HGU7 and HGU6 at MP-5, MP-6, and MP-8 Head inflection (F) in Fig. 3 appears to correlate with a brief return to the storm generated cycle of Reno Member sublithologies (Fig. 2) within the lower 3 4 m of the Tunnel City Group. Although the sublithologies cycle within the lower section of the Tunnel City Group, the frequency of the cycles is less than observed in HGU6. Therefore, the laterally continuous bedding parallel fractures characteristic of HGU6 may not be as developed or frequent in the lower section of the Tunnel City Group, top of HGU4. As a result, the MP-6 head profile suggests that the lower section of the Tunnel City Group is hydraulically similar to the upper portion of the Wonewoc Formation and 123

200 Environ Geol (2008) 56: thus included in HGU4. In addition, an inflection in the MP-5, MP-8, and MP-15 multilevels also supports the continuous presence of the HGU5/HGU4 contact as can be seen in Fig. 5. Identification of the HGU4/HGU3 interface Head change (G), which defines the interface between HGU4 and HGU3 (Fig. 3), is subtle and accounts for only 2% of the total head loss across the vertical section monitored by the MP-6 MLS. However, this head change was also observed in MLS MP-15, 1.7 km downgradient from MP-6 (Fig. 5); the inflection was reproduced in each of the MP-6 head profiles (Fig. 3); and the MP-6 head profile above and below inflection G is uniform and reproducible (Fig. 3). As was mentioned above, bulk- and matrixhydraulic conductivity data suggest HGU4 is characterized by a combination of fracture flow and matrix flow while HGU3 is dominated by matrix flow. Further support is provided by several other data sets that are consistent with a combination of fracture and matrix flow in HGU4 and dominance of matrix flow in HGU3. HGU4 is composed of the lower section of the Tunnel City Group and the Ironton Member of the Wonewoc Formation. The Ironton Member is primarily a coarse-grained quartzose sandstone that often contains interbeds of finegrained highly-feldspathic sandstone and is known to alternate between poorly and well cemented sections in outcrop (Emrich 1966; Ostrom 1978). The changes in cementation within the Ironton Member in the subsurface may create alternating zones of fracture-dominated and matrix-dominated flow. The interbedding of fine-grained highly-feldspathic sandstones within the Ironton Member also likely contributes to the slightly higher and more variable gamma signal in the upper portion of the Wonewoc Formation observed at MP-5, MP-6, and MP-8 (Fig. 4). The lower gamma signal and lack of recovery during drilling encountered at MP-5, MP-6, MP-8, and MP-15 in the lower portion of the Wonewoc Formation is indicative of the Galesville Member which is described as a well sorted, clean, friable, medium- and fine-grained quartzose sandstone (Emrich 1966; Ostrom 1978). HGU3 is composed of the Galesville Member and the upper section of the Eau Claire Formation. HGU3 is likely dominated by matrix flow with some instances of fracture flow as suggested by visual inspection of the core. This is consistant with the heat pulse flow meter logs obtained from the MP- 5, MP-6, and MP-8 coreholes. Although the sampling points are spaced relatively far apart, the flow meter log shows downward flow in the borehole adjacent to HGU3 decreases linearily with depth suggesting flow is exiting the borehole primarily through the matrix of HGU3 (Fig. 8). Fig. 8 Comparison of heat pulse flow meter log across HGU3 at MP- 5, MP-6, and MP-8 (error bars represent ±5% of each measurement) Identification of HGU2 the classic aquitard Inflection (H), shown in Fig. 3, accounts for 11% of the overall head loss across the vertical section monitored by the MLS. All of the study area data suggest that this head change is occurring across the aquitard section of the Eau Claire Formation. Similar inflections are observed at MP- 15 and MP-8 (Fig. 5). The Eau Claire Formation itself is characterized as a relatively finer-grained sandstone with abundant interbeds of argillaceous material (Aswasereelert 2005). The unit is considered heterolithic at the study area because it is composed of fine-grained sandstone, siltstone, and mudstone whose sequence and frequency are dependent on depositional environments (Aswasereelert 2005). Generally, the Eau Claire Formation correlates well with a series of gamma peaks as shown in Fig. 4. However, inflection (H), shown in Figs. 3 and 5, shows that the aquitard portion of the formation, HGU2, is thin (\3.4 m). Note, that not all gamma peaks in the Eau Claire Formation represent an aquitard unit. The gamma peak correlated with the head drop represents a section dominated by more siltstone and shale rich sublithologies rather than sandstone sublithologies (Aswasereelert 2005). The HGU2 contact at MP-5 123

201 38 Environ Geol (2008) 56:27 44 as a leaky confining unit (Krohelski et al. 2000). The aquitard is patchy and non-existent in certain areas of the county (Bradbury et al. 1999) and identification of the finer-grained facies in cuttings is problematic. The limits of the Eau Claire Formation are often delineated using natural gamma logs, and based on these logs, the Eau Claire Formation at the study area has an average thickness of 10 m. However, the detailed head profiles at MP-6, MP- 8, and MP-15 indicate that the section of the Eau Claire Formation providing vertical resistance to flow, HGU2, has an average thickness of only 2.7 m and that the natural gamma signature is misleading with respect to hydraulic characteristics. Bradbury et al. (2006) provide another example where a head profile measured using several different techniques confirms that the section of the Eau Claire Formation providing vertical resistance to flow is much thinner than the formation itself. Therefore, detailed head measurements throughout the Eau Claire Formation are essential to the appropriate delineation of the aquitard unit across Dane County. HGU1 regional aquifer Fig. 9 MP-6 Packer slug test data across HGU2 (Fig. 5) is represented by dashed lines because detailed head data across HGU2 is not available for that location. The same sublithologies observed in HGU2 at MP-6 are present at a higher elevation at MP-5 and are taken to represent HGU2 at the MP-5 location until further data can be collected. Support for the HGU2 delineation is provided by the packer hydraulic conductivity testing at MP-6. For example, the bulk horizontal hydraulic conductivity for HGU2 is lower than the units above and below as shown in Fig. 9. Note that the Eau Claire Formation above HGU2 is combined with HGU3 and the Eau Claire Formation below HGU2 is combined with HGU1. In both cases, those sections of the Eau Claire Formation are dominated by sandstones similar to the sandstones observed in HGU3 and HGU1. The detailed hydraulic head profiles served as the primary basis for identifying the Eau Claire aquitard, identified as HGU2 at this study area, which also has regional significance. For example, the finer-grained facies of the Eau Claire Formation are generally considered a regional aquitard in Dane County (Bradbury et al. 1999) and have been incorporated into regional numerical models HGU1 is composed of the lower section of the Eau Claire Formation and the Mt Simon Formation sandstones and serves as a regional aquifer in Dane County. The Mt Simon Formation is described in Fig. 2, and is estimated to be greater than 80 m thick at the study area and is as thick as 137 m in some parts of Dane County (Bradbury et al. 1999). The Mt Simon Formation overlies Precambrian rocks functioning as a base for the regional groundwater flow system (Bradbury et al. 1999; Krohelski et al. 2000). The portion of the Mt Simon Formation that has been investigated at the study area appears to be relatively uniform. However, Odom (1975), Ostrom (1978), and Runkel et al. (2003, 2006) describe sections of the Mt Simon Formation similar to the Ironton Member of the Wonewoc Formation which may alternate between finergrained, more feldspathic, and well cemented sections and coarser grained, less feldspathic, poorly cemented sandstone sections. Hence, matrix flow rather than fracture flow may dominate in the poorly cemented sections. Examination of potential causes for poor vertical connection between HGUs The entire vertical section of bedrock has appreciable horizontal conductivity with only one section within the Eau Claire Formation, HGU2, acting as an aquitard in the classic sense. Nearly all of the HGUs are dominated by fracture flow and, as such, the inflections in the MP-6 head 123

202 Environ Geol (2008) 56: Fig. 10 a Fracture distribution in vertical cross section used in FRACTRAN (Sudicky and McLaren 1992) simulation, b simulated head profiles showing inflections between each unit, c schematic of mechanical layers and interfaces (modified from Underwood et al in AAPG Bulletin, Copyright Ó2003, reprinted by permission of the AAPG whose permission is required for further use) profile are most likely caused by poor connection between the fracture networks of adjacent units. As a result, most of the HGUs are not separated by what would be considered a classic aquitard. Rather, the system would be better described as a sequence of stacked aquifers with limited vertical connection between adjacent aquifer units. The 2D numerical model FRACTRAN, developed by Sudicky and McLaren (1992), was used to explore the conceptual model discussed above. The modeled cross-sectional domain consists of four layers with uniform matrix properties, each with a discrete fracture network (Fig. 10a). The fracture networks were deliberately created with limited vertical connection between each layer (Fig. 10a) by decreasing the fracture density, length, and aperture of fractures near the interfaces between layers. High-resolution head profiles are plotted from the simulated steady-state head distribution at distances of 50, 150, and 200 m away from the inflow side of the model (Fig. 10b). The head profiles show distinct inflections in head between each of the adjacent layers. This hypothetical example demonstrates the effect of poor connection between the fracture networks of adjacent units and supports the interpretation of the inflections observed in the MP-6, MP-5, and MP-8 head profiles shown in Figs. 3 and 5, where horizontal hydraulic conductivities are large throughout the profile. Mechanisms responsible for creating vertical connectivity, or lack thereof, between fracture networks of various rock units have been studied by others from a geomechanical view. Cooke et al. (2006) provide a review of field observational evidence (Friedman et al. 1994; Rijken and Cooke 2001; Underwood et al. 2003), geomechanical experimental results (Teufel and Clark 1984; Friedman et al. 1994; Renshaw and Pollock 1995), and numerical simulations (Teufel and Clark 1984; Cooke and Underwood 2001; Rijken and Cooke 2001) showing how changes in rock material properties, preexisting fractures, and existing flaws control vertical, perpendicular to bedding, fracture propagation. According to Cooke et al. (2006), the two most common causes of vertical fracture termination are ductile layers and weak horizons. Cooke et al. (2006) show that sedimentology and stratigraphy can be used to identify these layers or horizons, thereby identifying 123

203 40 Environ Geol (2008) 56:27 44 potential mechanical interfaces. A mechanical layer is a unit of rock that behaves homogeneously in response to stress and a mechanical interface is the contact between two mechanical layers (Gross 1993). Vertical fractures do not extend between different mechanical layers as is shown schematically in Fig. 10c. Therefore, vertical connectivity between the two units relies on serendipitous alignment of the vertical fractures between the two units, flow along tortuous horizontal paths as shown by the circled sections in Fig. 10c, and/or between the relatively impermeable matrices (Cooke et al. 2006). Cooke et al. (2006) explain that, depending on the direction of the hydraulic gradient, the vertical disconnection between units separated by mechanical interfaces can cause a compartmentalization of the flow regime. This type of compartmentalization is demonstrated by the MP-6 head profile. How much detail in head profiles is enough? Detailed head profiles are required if they are to be useful for defining HGUs. The resolution of a head profile is controlled by two main factors: (1) the total vertical head differential across the section of interest and (2) the number and length of the monitoring intervals. The largest vertical head differentials, in the absence of anthropogenic effects, are most likely to occur in the recharge or discharge zones of a natural flow system. A large difference in head across the section of interest facilitates the resolution of relatively small differences in head from the overall vertical gradient. The number and length of the monitoring intervals also contributes to the resolution of the head profile. The monitoring intervals should be relatively short to avoid blending greatly different values of head and creating unnatural vertical pathways for flow, cross-connection, in the monitoring well itself. In addition, it is desirable to include as many monitoring intervals as possible in the design of the MLS because the detail necessary to produce a head profile useful for HGU delineation is not known a priori, as demonstrated herein. Though the investigators found the MP-6 head profile provided by a Westbay 1 MLS essential for defining HGUs, the Westbay 1 system is not the only available system. There are other commercially available systems, the Solinst 1 Waterloo system described by Parker et al. (2006), the Water FLUTe TM system described by Cherry et al. (2007), and the Solinst 1 CMT System described by Einarson and Cherry (2002), which are all well-suited to measuring detailed head profiles. The various systems have different advantages and disadvantages depending on factors such as borehole diameter, maximum monitoring depth, maximum head differentials between monitoring intervals, and removability. The diversity in design options and robustness of the commercially available MLSs provide hydrogeologists with a range of choices to match the site-specific needs of field investigations in many types of geologic and hydrologic circumstances. Westbay 1 systems can be installed in boreholes ranging in diameter between 7.6 and 15.2 cm (3 and 6 in.) ( The smallest monitoring interval a Westbay 1 system can accommodate is 0.6 m (2 ft) with one 0.9 m (3 ft) long packer above and below the interval. Therefore, for a corehole similar to MP-6 with 120 m of open borehole available for monitoring, a maximum of 78 head monitoring zones could be included. The MP-6 corehole included less than the maximum number of monitoring zones for several reasons. First, few packers could be positioned in the upper 17 m of the MP-6 borehole due to the highly fractured and likely karstic nature of that section. Second, many complementary data sets were used during the design of the MP-6 MLS to strategically position the monitoring intervals around likely HGU contacts which decreased the total number of monitoring intervals from the maximum possible. Third, it was desirable to add pumping ports to many of the intervals so that hydraulic tests could be performed. The addition of a pumping port increases the minimum length of the monitoring zone to 1.2 m (4 ft). Though it is important to maximize the number of monitoring intervals, it is also important to design the system so that monitoring intervals do not straddle interfaces, resulting in blending heads of different hydraulic units, which are expected to represent HGU contacts. Therefore, the design process should incorporate all available data in order to place the monitoring intervals and borehole seals appropriately; although, it is impossible to know exactly where HGU contacts occur a priori. Even though it is important to design MLSs with as much detail as possible, it is not necessary to monitor the entire study area in the same level of detail. In fact, the insight gained from a few carefully placed multilevels can provide the basis for design of less detailed systems for other locations. Flow system conceptual model The study area conceptual model for groundwater flow includes 13 bedrock HGUs dominated by horizontal flow within each HGU and vertical leakage between the HGUs, as shown in Fig. 11. The dominance of horizontal flow within each of the HGUs is supported by the nearly vertical nature of the MP-6 head profile within each HGU and the appreciable horizontal hydraulic conductivity through the vertical section. Vertical gradients are generally downward (Fig. 5) throughout the deeper parts of the system as is shown in Fig. 11. The MP-6, MP-5, and MP-8 head 123

204 Environ Geol (2008) 56: Fig. 11 a, b Cross-sectional schematics for the stratigraphy and hydrogeologic framework, c schematic conceptual model for groundwater flow profiles shown in Fig. 5 also indicate a recharge area in the western portion of the site and a discharge area in the east (Fig. 11). The conceptual model is also supported by other local studies of groundwater flow in fractured sedimentary rocks as well as regional studies of groundwater flow in Dane County. For example, other investigators (Muldoon 123

205 42 Environ Geol (2008) 56:27 44 et al. 2001; Swanson and Bahr 2004; Swanson et al. 2006; Swanson 2007) have observed high horizontal hydraulic conductivities attributed to laterally extensive bedding parallel fractures in the Silurian dolostones in northeastern Wisconsin and the Tunnel City Group Sandstones in Dane County. In addition, Bradbury et al. (1999) found that groundwater movement in the Mount Simon is primarily horizontal. McLeod (1975), Bradbury et al. (1999), and Krohelski et al. (2000) also support the expectation of vertical leakage between HGUs as these investigators found that surface water, shallow groundwater, and deep groundwater are all inextricably linked in the region. Groundwater Contamination Research and by a grant from NSERC Canada. Financial and technical support was also provided by Westbay 1 Instruments Inc. (a division of Schlumberger Water Services), FLUTe TM, and Solinst 1 Canada Ltd. In addition, expertise and time were provided by Carl Keller of FLUTe TM, Andrew Bessant, Dave Larssen and Frank Magdich of Westbay 1 Instruments Inc., and Diane Austin formerly of the University of Waterloo (UW). This paper also relied on geophysical logs collected by Peeter Pehme and Daren Mortimer (Dillon Consulting, ON) and Ken Bradbury and others at the WGNHS. In addition, the numerical modeling was performed by Steve Chapman (UW). The authors also benefited from various discussions with all of the above people as well as Sue Swanson (Beloit College), Mike Noel (GeoTrans), and Tom Miazga (Hydrite Chemical Company). References Conclusions Detailed head profiles were essential for defining HGUs in the fractured sedimentary rock at the study area. The eight inflections in the MP-6 head profile provided direct support for ten HGUs at the MP-6 location. Multilevel head profiles at the MP-5, MP-15, and MP-8 coreholes as well as geological, geophysical, and hydraulic data collected from the same coreholes provide support for the HGUs delineated at MP-6 and provide information about the lateral extent and thickness of each of the units. The detailed head profiles at several locations also provided quantitative information concerning the 3D distribution of head at the study area which provided the main basis for the conceptual groundwater flow model. The head profiles measured from MP-6, MP-5, MP-15, and MP-8 generally show little or no head differential within each HGU indicating that flow within each HGU is predominantly horizontal. The multilevel head data along with several complementary data sets suggest that the hydrogeologic framework at the study area is a system of stacked HGUs that are hydraulically distinct from adjacent units because of poor fracture connection vertically. The sharp inflections in the multilevel head profiles did not necessarily align with stratigraphic contacts and are generally not indicative of classic aquitards with appreciable thickness, but rather of interfaces where connection between fracture networks is poor. No other data set could have elucidated these distinct hydrogeologic interfaces. The detail was important to characterizing the style of these unit boundaries and accurately measuring the vertical component of the hydraulic gradient across the boundaries. Various types of commercially available MLSs can provide detailed and accurate head profiles, but they are typically designed with an insufficient number of monitoring intervals to provide the data necessary to discern hydrogeologic interfaces. Acknowledgments This research was funded by Hydrite Chemical Company through the University Consortium for Field-Focused Anderson MP (1989) Hydrogeologic facies models to delineate largescale spatial trends in glacial and glaciofluvial sediments. Geol Soc Am Bull 101(4): ASTM (2004) Standard test for permeability of rocks by flowing air, D , ASTM International, West Conshohocken Aswasereelert W (2005) Facies distribution and stacking of the Eau Claire Formation, Wisconsin: implications of thin shale-rich strata in fluid flow. Master s thesis, Department of Geology and Geophysics, University of Wisconsin, Madison Austin DC (2005) Hydrogeologic controls on contaminant distribution within a multi-component DNAPL zone in a sedimentary rock aquifer in south central Wisconsin. Master s thesis, Earth Sciences Department, University of Waterloo Black WH, Smith HR, Patton FD (1986) Multiple-level ground water monitoring with the MP system. Paper presented at proceedings of the surface and borehole geophysical methods and ground water instrumentation conference and exposition, Denver, Colorado, October 1986, NWWA, Dublin, Ohio Bradbury KR, Swanson SK, Krohelski JT, Fritz AK (1999) Hydrogeology of Dane County, Wisconsin, WGNHS open-file report Madison, Wisconsin Bradbury KR, Gotkowitz MB, Hart DJ, Eaton TT, Cherry JA, Parker BL, Borchardt MA (2006) Contaminant transport through aquitards: technical guidance for aquitard assessment. Denver, Colorado, Awwa Research Foundation Cherry JA, Johnson PE (1982) A multilevel device for monitoring in fractured rock. Ground Water Monit Rev 2(3):41 44 Cherry JA, Parker BL, Bradbury KR, Eaton TT, Gotkowitz MB, Hart DJ, Borchardt MA (2006) Contaminant transport through aquitards: a state-of-the-science review. Denver, Colorado, Awwa Research Foundation Cherry JA, Parker BL, Keller C (2007) A new depth-discrete multilevel monitoring approach for fractured rock. Ground Water Monit Remed 27(2):57 70 Clayton L, Attig JW (1997) Pleistocene geology of Dane County, Wisconsin, WGNHS Bulletin 95. Madison, Wisconsin Cline DR (1965) Geology and ground-water resources of Dane County, Wisconsin, USGS water-supply paper 1779-U. United States Government Printing Office, Washington DC Collins RE (1961) Flow of fluids through porous materials. Reinhold Publishing, New York Cooke ML, Underwood CA (2001) Fracture termination and stepover at bedding interfaces due to frictional slip and interface opening. J Struc Geol 23(2 3): Cooke ML, Simo JA, Underwood CA, Rijken P (2006) Mechanical stratigraphic controls on fracture patterns within carbonates and implications for groundwater flow. Sediment Geol 184(3 4):

206 Environ Geol (2008) 56: Davis SN, DeWiest RJM (1966) Hydrogeology. Wiley, New York Einarson MD (2006) Multilevel ground-water monitoring. In: Nielsen DM (ed) Practical handbook of environmental site characterization and ground-water monitoring, vol 11. CRC Press, Boca Raton, pp Einarson MD, Cherry JA (2002) A new multilevel ground water monitoring system using multichannel tubing. Ground Water Monit Remediat 22(4):52 65 Emrich GH (1966) Ironton and Galesville (Cambrian) sandstone in Illinois and adjacent areas, Illinois State Geological Survey Circular 403. Urbana, Illinois Freeze RA, Witherspoon PA (1966) Theoretical analysis of regional groundwater flow: 1. Analytical and numerical solutions to the mathematical model. Water Resour Res 2(4): Freeze RA, Witherspoon PA (1967) Theoretical analysis of regional groundwater flow. 2. Effect of water-table configuration and subsurface permeability variation. Water Resour Res 3(2): Friedman M, Kwon O, French VL (1994) Containment of natural fractures in brittle beds of the Austin Chalk. In: Balkema A.A Paper presented at proceedings of the 1st North American rock mechanics symposium, rock mechanics, models and measurements, challenges from industry, Austin, TX, 1 3 June 1994, Rotterdam, Netherlands Gross MR (1993) The origin and spacing of cross joints: examples from the Monterey Formation, Santa Barbara coastline, California. J Struc Geol 15(6): HSI GeoTrans (1999) DNAPL removal report, Hydrite Chemical Co., Cottage Grove, Wisconsin, project no /N152. Brookfield, Wisconsin Hydro-Search Inc. (1989) RCRA facility investigation, task 1 current conditions report, Avganic Industries, Inc. solvent reclamation facility, Cottage Grove, Wisconsin, project no. 158E Brookfield, Wisconsin Krohelski JT, Bradbury KR, Hunt RJ, Swanson SK (2000) Numerical simulation of groundwater flow in Dane county, Wisconsin, WGNHS Bulletin 98. Madison, Wisconsin Macfarlane PA, Doveton JH, Feldman HR, Butler JJ Jr, Combes JM, Collins DR (1994) Aquifer/aquitard units of the Dakota aquifer system in Kansas: methods of delineation and sedimentary architecture effects on ground-water flow and flow properties. J Sediment Res B64(4): Mai H, Dott RH Jr (1985) A subsurface study of the St Peter sandstone in southern and eastern Wisconsin, WGNHS Information Circular Number 47. Madison, Wisconsin Maxey GB (1964) Hydrostratigraphic units. J Hydrol 2(2): McLeod RS (1975) A digital-computer model for estimating drawdowns in the sandstone aquifer in Dane County, Wisconsin, WGNHS Information Circular 28. Madison, Wisconsin Meinzer OE (1923) Outline of ground-water hydrology, with definitions, US Geological Survey Water-Supply Paper 494. United States Government Printing Office, Washington, DC Meyer JR (2005) Migration of a mixed organic contaminant plume in a multilayer sedimentary rock aquifer system. Master s thesis, Earth Sciences Department, University of Waterloo Muldoon MA (1999) Data from slug tests in the Silurian dolomite using a short-interval straddle-packer assemblage, WGNHS Open-File Report Madison, Wisconsin Muldoon MA, Simo JA, Bradbury KR (2001) Correlation of hydraulic conductivity with stratigraphy in a fractured-dolomite aquifer, northeastern Wisconsin, USA. Hydrogeol J 9(6): Odom EI (1975) Feldspar grain size relations in Cambrian arenites, upper Mississippi Valley. J Sediment Petrol 45(3): Odom EI (1978) Lithostratigraphy and sedimentology of the Lone Rock and Mazomanie Formations, upper Mississippi valley. In: Lithostratigraphy, petrology, and sedimentology of Late Cambrian Early Ordovican Rocks Near Madison, Wisconsin. WGNHS Field Trip Guide Book Number 3. Madison, Wisconsin Odom EI, Ostrom ME (1978) Lithostratigraphy, petrology, sedimentology, and depositional environments of the Jordan Formation near Madison, Wisconsin. In: Lithostratigraphy, petrology, and sedimentology of Late Cambrian Early Ordovican Rocks Near Madison, Wisconsin. WGNHS field trip guide book number 3. Madison, Wisconsin Ostrom ME (1978) Stratigraphic relationships of Lower Paleozoic Rocks of Wisconsin. In: Lithostratigraphy, petrology, and sedimentology of Late Cambrian Early Ordovican Rocks Near Madison, Wisconsin. WGNHS field trip guide book number 3. Madison, Wisconsin Palmquist RC (1969) The configuration of the Prairie du Chien-St Peter contact in southwestern Wisconsin: an example of an integrated geological geophysical study. J Geol 77: Parker BL, Cherry JA, Swanson BJ (2006) A multilevel system for high-resolution monitoring in rotasonic boreholes. Ground Water Monit Remed 26(4):57 73 Poeter E, Gaylord DR (1990) Influence of aquifer heterogeneity on contaminant transport at the Hanford Site. Ground Water 28(6): Renshaw CE, Pollock DW (1995) An experimentally verified criterion for propagation across unbounded frictional interfaces in brittle, linear elastic materials. Int J Rock Mech Min Sci Geomechan Abstr 32(3): Rijken P, Cooke ML (2001) Role of shale thickness on vertical connectivity of fractures: application of crack-bridging theory to the Austin Chalk, Texas. Tectonophysics 337(1 2): Runkel AC, Tipping RG, Green JA, Mossler JH, Alexander SC, Alexander EC Jr (2003) Hydrogeology of the Paleozoic bedrock in southeastern Minnesota, Minnesota geological survey report of investigations 61. St Paul, Minnesota, University of Minnesota Runkel AC, Tipping RG, Alexander EC Jr, Alexander SC (2006) Hydrostratigraphic characterization of intergranular and secondary porosity in part of the Cambrian sandstone aquifer system of the cratonic interior of North America: improving predictability of hydrogeologic properties. Sediment Geol 184(3 4): Seaber PR (1988) Hydrostratigraphic units. In: Back W, Rosenshein JS, Seaber PR (eds) The geology of North America volume O-2. Boulder, Colorado, The Geological Society of America, vol 2, pp 9 14 Smith GL, Simo JA (1997) Carbonate diagenesis and dolomitization of the lower Ordovician Prairie du Chien Group. Geosci Wisconsin 16:1 16 Sudicky EA, McLaren RG (1992) The Laplace transform Galerkin technique for large-scale simulation of mass transport in discretely fractured porous formations. Water Resour Res 28(2): Sutherland JL (1986) Stratigraphy and sedimentology of the upper Cambrian Lone Rock Formation, western Wisconsin-focus on the Reno Member. Master s thesis, Department of Geology and Geophysics, University of Wisconsin, Madison Swanson SK (2007) Lithostratigraphic controls on bedding-plane fractures and the potential for discrete groundwater flow through a siliciclastic sandstone aquifer, southern Wisconsin. Sediment Geol 197(1 2):65 78 Swanson SK, Bahr JM (2004) Analytical and numerical models to explain steady rates of spring flow. Ground Water 42(5): Swanson SK, Bahr JM, Bradbury KR, Anderson KM (2006) Evidence for preferential flow through sandstone aquifers in southern Wisconsin. Sediment Geol 184(3 4):

207 44 Environ Geol (2008) 56:27 44 Teufel LW, Clark JA (1984) Hydraulic fracture propagation in layered rock: experimental studies of fracture containment. Soc Petrol Eng J 24(1):19 32 Toth JA (1962) A theory of ground-water motion in small drainage basins in central Alberta, Canada. J Geophys Res 67(11): Toth JA (1963) A theoretical analysis of ground-water flow in small drainage basins. J Geophys Res 68(16): Underwood CA, Cooke ML, Simo JA, Muldoon MA (2003) Stratigraphic controls on vertical fracture patterns in Silurian dolomite, northeastern Wisconsin. AAPG Bull 87(1):

208 A New Innovative Method for Continuous Hydraulic Conductivity Profiling in Fractured Rock Holes Carl E. Keller 1, John A. Cherry 2 and Beth L. Parker 2 1. Flexible Liner Underground Technologies Limited, Santa Fe, NM, School of Engineering, University of Guelph, Guelph, ON, N1G2W1 For Submission To: GROUND WATER DRAFT, October 25,

209 Abstract This paper presents a new method for investigating transmissive features in fractured rock boreholes. A sock-like borehole liner formed of a flexible, water-tight, urethanecoated nylon fabric is attached to the top of casing and filled with water in a manner to cause the liner to evert (the reverse of invert) down the borehole due to a constant excess hydraulic head inside the liner. The everting liner pushes water out of the borehole beneath it into the formation and the measured descent rate is governed by the bulk transmissivity of the features below the liner. The liner descent rate decreases each time the bottom of the everting liner passes and seals a transmissive feature. Each change in velocity identifies the depth of a permeable feature and the magnitude of each change is related to the transmissivity of the specific feature. The detailed transmissivity distribution is calculated from the liner velocity profile using the Thiem equation for radial flow. The profiling method is performed in holes 89mm (3.5 inches) to 280mm (11 inches) or larger and commonly takes less than a few hours making it much faster than conventional straddle packer testing or borehole flow metering when these are applied throughout the entire borehole. The utility of this method was demonstrated by profiling boreholes in a dolostone aquifer where straddle packer tests were done for comparison. The sealing liner is usually left in the borehole after the K profiling episode to provide a temporary continuous seal of the hole against cross connection (250 words). Keywords: fractured rock, hydraulic conductivity, packer testing, K profiling, transmissivity 2

210 INTRODUCTION Understanding flow through the fracture network in bedrock is essential for valid assessments of contaminant transport and fate, ground water resource management, and groundwater control at mining and dam sites. In most types of rock, groundwater flow is dominated by the fracture network, especially when the rock matrix blocks between fractures have low permeability. Various methods are available to acquire insight from boreholes concerning the locations and nature of fractures including borehole imaging, geophysical logging, hydrophysical logging, flow metering, hydraulic tests using single or double packers, and core logging. Neuman (2005) draws attention to the importance of assessing the numerous fractures in each hole potentially involved in groundwater flow rather than just the few fractures appearing to be the dominant ones in borehole observations. This implies that the identification of all potential flow features and the measurement of the flow capacity of each of those features are necessary for accurate characterization of flow in fractured rock. There are substantial limitations of existing methods for identifying transmissive borehole features. Borehole televiewing (optical, acoustic, or electrical) qualitatively identifies variable sized fractures but does not discern which of these are transmissive under natural or forced gradient conditions. Flow metering and profiles of fluid resistivity and temperature in open holes rely on open-hole cross connection gradients, which typically are dominated by the flow through the fractures with the largest flow capacity thereby masking the effects of smaller fractures (e.g., Pehme et al 2007). Hydraulic straddle packer tests involving water injection or withdrawal measure the transmissivity of specific intervals that are isolated from the rest of the borehole by 3

211 packers, and when these tests are done throughout the entire borehole using short test intervals, the locations of all major transmissive zones are known. However, to test an entire borehole in this manner is time intensive and therefore is typically very costly. This new, profiling method is capable of identifying and measuring the hydraulic conductivity (K) of all permeable features (fractures and fracture zones) in a borehole much more quickly than packer testing, thereby enabling the entire borehole to be tested at a much lower cost. This paper describes a new, innovative method known as hydraulic conductivity profiling (K profiling), that quickly identifies and measures the transmissive features in rock boreholes and measures the full borehole transmissivity (T) in a cumulative manner. A flexible, urethane-coated nylon tube (borehole liner), slightly larger than the borehole diameter, is attached to the top of casing (TOC), pushed into the casing creating an annulus, and driven by water pressure down the hole causing the liner to evert (Figure 1). This inflated everting liner, in effect, acts as a piston pushing the water column beneath it out into the formation. As the liner progressively fills the borehole, the applied water pressure inside the liner forces the impervious nylon fabric against the borehole wall so that the fractures are then sealed sequentially from the top down. When the end of the liner reaches the bottom of the hole, the entire hole is lined with the fabric creating a complete impervious borehole seal preventing borehole cross connection. The rate of descent of the liner going down the hole is carefully measured allowing the calculation of K with depth thereby identifying the fractures and fracture zones present. This borehole liner (continuous flexible liner) was initially developed for depth-discrete, multilevel monitoring in boreholes (e.g. Cherry et al, 2007). Later, as recognition of the 4

212 need to prevent borehole cross connection increased, liners were installed in hundreds of holes without the monitoring ports (i.e., blank liners). Price and Williams (1993) and Sterling et al (2005) provide examples of hydraulic cross connections in rock boreholes where major disturbances in the natural hydrochemistry and contaminant distribution, respectively, were observed; illustrating the need to minimize borehole cross connections between fractures. Minimization of cross connections in rock boreholes has been mandated in contaminated site investigations for some jurisdictions. For example, the State of New Jersey has established regulations requiring that boreholes not be left open more than 24 hours (NJ Reg. 7:9D-2.2 (a) 10) and that unsealed portions of the borehole must be less than 25 ft (NJ Reg. 7:9D2.4 (a)). The K profiling method described in this paper is typically conducted at the time the liner is installed in the hole to prevent cross connection and therefore the installation procedure results in two accomplishments: emplacing the borehole seal and obtaining the K profile. If a borehole is profiled more than once, there is often evidence of well development resulting from installing and removing the liner. This method has been applied since late 2003 in over 200 boreholes at forty-six sites across North America and Germany in different types of fractured rock. After profiling, the liner is typically left in the hole for several weeks or months before removal to allow time for the design and installation of a well or multi level monitoring system, thereby eliminating the effects of cross contamination on future depth discrete measurements. In this study, K profiling test results are compared with packer testing results in a fractured dolostone aquifer. 5

213 FIELD APPROACH AND TEST METHOD Each FLUTe liner is fabricated to the depth and diameter of the hole to be profiled and a tether (i.e., strong rope) is attached to the bottom interior end of the liner with roughly the same length, because after installation the bottom of the liner will be at the bottom of the hole and the tether must reach the ground surface for removal purposes. All liners have depth markings every 10 or 20 feet along the liner and tether for depth verification purposes during profiling. At the start of profiling the liner is inside-out relative to its final emplaced geometry on a reel positioned next to the hole and the open end of the liner is attached to the top of the borehole surface casing with a large hose clamp (Figure 1a). The GS marking on the liner is used to reference the liner depths to ground surface or to the TOC. The liner is then pushed down by hand into the casing to form an annular pocket into which water is added. The water column inside the liner creates a driving force that pulls the liner off the reel and drives it down the hole like a piston (Figure 1b). Initially, when the liner is lowered through the air-filled interval above the standing water level in the borehole, the air beneath the liner escapes through a perforated tube hung in the hole down to the water table. Once the liner has been everted a short distance below the water table, the liner is temporarily restrained at the wellhead to prevent its descent while the driving head inside the liner is developed by water addition. The water level inside the liner is always kept well above the water table to ensure inflating the liner against the borehole wall to create a complete borehole seal above the bottom of the advancing liner (Figure 1c). This head differential also drives the liner eversion and the greater the excess head, the faster the liner everts into the hole and the faster the water is forced out into the formation via permeable zones. 6

214 When the liner is released the descent is measured while the water level is maintained at a relatively constant head inside the liner (Figure 2a). The constant applied head in the liner throughout the profile causes the liner descent velocity to be greatest at the beginning, but as the transmissive zones are sealed successively from the top downward, the descent slows until all of the transmissive zones are sealed and the descent velocity becomes negligible (e.g ft/s) thereby completing the profile. Each time a distinctive reduction in descent rate is observed, a transmissive zone is identified. At the end of the profile the position of the liner is then fixed by tying off the tether at the surface to prevent further descent. Experience has shown that during profiling, air can become entrapped inside the liner as it descends and form a balloon in the end of the liner influencing the descent velocity. To prevent this during K profiling, a vacuum is created inside the liner through a vent tube and check valve built into the end of the liner prior to profiling. The liner can be removed by pulling on the tether that is attached to the bottom of the liner, inverting the liner as it is pulled upward while the water inside the liner is either contained or allowed to spill onto the ground. The profiling method, including the machine which performs the measurement (profiler), was developed by Carl Keller (US pat. nos. 6,910,374 and 7,281,422). Figure 2 shows the setup of equipment and identifies the components of the system used in the recording of the parameters necessary for calculating transmissivity (T) including the elapsed time, the liner tension, the driving head, and the depth of the eversion point, all of which are recorded every 1-2 seconds. The liner velocity is measured with a pair of encoders on a metering roller that can accurately measure both high and low velocities and a continuous recording of the head inside the liner is measured by a transducer and bubbler tube 7

215 system. The depth of the static water level in the open hole and the depth of the end of the bubbler tube must be measured before beginning the profile to determine the driving head throughout the profile. The tension at which the liner is held back is controlled using a monitoring roller equipped with a braking system and the tension measurement is done with a pair of load cells. The velocity of the eversion point (EP) is inferred from the velocity of liner descent as measured by the roller, because the inverted liner travels twice as fast into the hole than the EP. At the beginning of the profile, the initial rate of flow is a direct measure of the transmissivity of the entire hole. As the EP travels down the hole, the liner sequentially seals the transmissive features from the top down, and the remaining transmissivity for the unsealed hole below is diminished. Because the excess head imposed inside the liner is constant, the liner descent velocity must decrease every time it passes a transmissive feature. The reduction in velocity is directly proportional to the outward flow that was stopped by sealing the transmissive zone. Figure 3a shows a liner passing a transmissive fracture which is flowing at a rate Q. When the EP passes the fracture, the liner velocity drops by an increment equal to Q/A; where A is the horizontal cross-section of the hole (e.g. Δv). Figure 3b illustrates the ideal case for a single fracture showing how sealing the fracture reduces the transmissivity remaining in the borehole beneath the liner. The depth of the EP when the drop in velocity occurs identifies the position of a transmissive zone and the magnitude of the change is directly proportional to the flow rate into that zone before it was sealed. Therefore, the entire liner decent velocity history is governed by the transmissivity distribution in the borehole. 8

216 CONCEPTUAL FRAMEWORK FOR DATA ANALYSIS A hypothetical velocity profile illustrating the nature of some velocity changes commonly observed is shown in Figure 4. This profile starts with an interval with zero slope representing no detectable permeability because there is no flow feature being sealed by the liner. The initial abrupt step change in velocity is typical of the liner passing a thin depth discrete nearly horizontal permeable feature intersecting the hole. The sloped portions of the velocity plot are either uniform permeable beds (a smooth slope) or fracture zones (a slope composed of small steps). A steeply sloped step can be a high angle fracture or a thin very permeable bed. The radial flow out of the borehole beneath the liner is assumed to be steady state radial flow represented adequately by the Darcy equation in cylindrical coordinates. The Thiem equation (Wenzel, 1936) is used to calculate the transmissivity of the borehole as the flow paths are sealed from the top downward. T ΔQ = 2πΔH BH r ln r o w Where: ΔQ = flow reduction due to sealing an interval of the borehole T = Transmissivity of the portion of the hole measured (KΔz) ΔH BH = profile driving pressure in the borehole beneath the liner r o = radius of influence of the test r w = radius of the borehole The Thiem equation is also commonly used for obtaining transmissivity values from data acquired from constant head injection tests in boreholes using straddle packers (Lapcevic 1988, Novakowski and Bickerton 1997). Assumptions of r o ranging from 2 ft to 60 m have been used (Maini, 1971; Zeigler, 1976; Haimson & Doe 1983; Bliss & Rushton Lapcevic et al., 1999) and this study assumes a 20 m test radius of influence. 9

217 The driving head in the borehole below the liner can be measured or calculated. It is calculated from measured parameters using the following empirical equation derived from lab tests: Δ H BH = ΔH L H MIN 2 ( Θ + Θ ) w A D Where: ΔΗ L = the head in the liner H MIN = the minimum head needed to evert the liner against the resistance due to the fabric stiffness Θ w = the tension on the liner at the well head Θ D = the total drag force on the liner within the hole (friction) A = the borehole cross section. The factor 2 is an empirical coefficient determined from many eversion tests of liners made from different materials. The total drag on the liner (Θ D ) is not measured, but it is intentionally reduced to as near zero as possible and only becomes important when the water table is very deep or when profiling a borehole with extremely high transmissivity. The use of a trimmie hose to introduce the water at the water table depth without wetting the inverted liner helps minimizes the drag when deep water tables are present, while in extremely high permeability boreholes, a larger excess head in the liner will help reduce the significance of drag on the liner. An alternative method for obtaining the head beneath the liner involves installing a pressure transducer at the bottom of the hole before starting the profile using a very thin line to the surface to minimize interference with the liner seal against the borehole wall. Figure 5 is a comparison of the calculated head with the measured head beneath the liner and the excellent agreement indicates that the borehole head calculations are valid for the majority of boreholes tested. However, for deep water tables and/or very fast flowing holes the use of a transducer at the bottom of the hole is highly recommended. 10

218 Figure 6 illustrates the assumed geometry and terms for calculation of the transmissivity, showing the liner at two different positions in the hole during its descent. The bubbler pressure is converted, using the static open borehole water level (blended head) and bubbler depth, to the driving head, ΔH L, inside the liner as a function of the EP depth in the hole. This driving head must be converted to the imposed head in the borehole below the liner (ΔH BH ) by accounting for frictional effects. The velocity of the liner is then divided by the ΔH BH, to provide a velocity per unit head (v(z)/δh BH ) which is plotted versus depth to create a velocity profile of the borehole. Because the depth increments for each time step vary with the liner velocity, the hydraulic conductivity obtained from the transmissivity calculation has variable depth resolution (e.g. the largest intervals are at the top of the hole where the velocity is highest). Changes in the velocity per unit driving pressure are then calculated throughout the hole and multiplied by the borehole crosssectional area to obtain ΔQ/ΔH BH for use in the Thiem equation. Prior to calculation of the transmissivity profile the velocity data is smoothed to reduce noise in the data recording/measurement. In order to judge the data smoothing effects on the velocity profile, the velocity curves are plotted for several degrees of smoothing. In Figure 7a, the raw data (black curve), the smoothed data (pink curve) and the monotonic fit (yellow curve) are overlain on the same graph. The degree of the smoothing is judged appropriate by the lack of deviation from the raw data. The excellent match of the three curves is typical of a very good data set. The oscillation in the raw data at 48.8m (160 ft) is due to the end of the liner and associated hardware passing through the profiler. 11

219 NON-IDEAL INTERFERENCES At the start of profiling when the liner starts to propagate down the hole, there is a short period when the flow field in the hole is clearly not steady state (Figure 8) because it takes time for the induced flow out of the borehole to come into equilibrium with the imposed pressure. As the flow field develops in the formation, the gradient at the hole wall becomes much less steep and the flow rate out of the hole beneath the liner approaches a nominal steady state rate. Fortunately, the transition zone is usually only about 2 6 m long depending upon the liner velocity and the transient portion of the initial data is cropped from the final results and defined as a zone of unknown flow paths. Many boreholes have enlarged sections commonly called breakout or washout zones, and as the liner passes through one of these enlarged segments, the liner expands to a larger cross sectional area (i.e., a larger volume displacement per unit length of travel) causing a corresponding drop in the liner velocity (Figure 9). This drop in velocity is not due to flow into the formation, but instead, by a change in borehole geometry and when the liner passes out of the washout zone, the velocity then increases again to the value before the breakout zone. If the liner exit velocity from the enlargement is less than the entrance velocity, the velocity change, and the associated transmissivity, is assigned to the upper portion of the enlargement. Therefore, the effects of a washout or other enlargement of the hole can be recognized and accounted for in the data reduction process. In some cases the borehole diameter is determined independently by geophysical logging (mechanical caliper, or virtual caliper from acoustic televiewer) and can be noted before conducting the profile and included in the data reduction. 12

220 COMPARISON AND DISCUSSION OF RESULTS In this study, the K profiling method was applied in three holes in a fractured dolostone aquifer in Guelph, Ontario, Canada for comparison of profiling to packer testing results. The overburden at this site is between 3 and 5 m thick and the holes were continuously cored (HQ, 96mm nominal) from the top of rock to the bottom of the holes. The water level in the open holes varies seasonally between 3 and 5 m below the ground surface (bgs). This dolostone aquifer supplies most of the municipal water supply for the City of Guelph and has substantial transmissivity. Continuous packer testing was done using the constant head injection method at 1.5 m intervals throughout each hole. The packer equipment setup and procedure is described by Quinn (2009 PhD thesis) and is a modification of the constant head packer testing equipment introduced conceptually by Gale (1982) to include the ability to conduct all four types of hydraulic tests at high resolution in each test interval (constant head step test, slug test, injection/withdrawal constant rate pumping test, and recovery test). Borehole MW-24 was drilled in 2005 through the dolostone aquifer into the underlying shale aquitard to a total depth of 104m below ground surface. Surface casing was set at 5m below ground surface (mbgs). The continuous straddle packer testing in this hole with 1.5 m intervals was done soon after drilling, taking eight days to complete and then the hole was sealed with a liner to prevent borehole cross connection. Later in 2006, this hole was profiled twice in one day using the FLUTe profiling technique to examine reproducibility (Figure 7b). The profiles are generally similar, but the T obtained from the second run was greater than the first by about 60% at the prominent fracture at 240 ft. This difference is likely due to well development effects caused by the removal of the 13

221 liner after the first run. A specially designed linear capstan is used to remove the liners by applying strong tension to the tether causing the liner to invert back up through the borehole. This tension creates a negative pressure beneath the liner which is relieved by drawing water from the formation into the borehole. The typical tension applied produces up to 30m of head difference below the liner likely causing removal of sediment from some fractures. Figure 10a shows the transmissivity profile obtained from the monotonic fit of the velocity profile for MW-24, and illustrates some of the typical depth discrete features obtained in a fractured rock. The profile shows that the transmissive part of the hole ends at a depth of 92m (302 ft). This depth is the beginning of the underlying shale formation identified on the basis of core inspection and geophysical logs and the K profile indicates that the shale is an effective aquitard at this location. The profile also indicates that numerous hydraulically conductive features exist in this hole with a range in transmissivity over three orders of magnitude. Twenty five lab permeability tests were conducted on the rock core from nearby boreholes to determine the range of matrix hydraulic conductivities resulting in a geometric mean value of 5.3x10-9 m/s. This is near the practical limit of liner measurements and conventional packer testing, and therefore, the transmissive features identified by profiling are likely permeable fractures. This is consistent with the large number of fractures indicated by core inspection and acoustic televiewer, which identify fractures but cannot distinguish permeable ones from nonpermeable ones. Furthermore, it is overall consistent with what is known about the hydrologic nature of this dolostone aquifer in Guelph. 14

222 To compare these results to the packer test results, it was necessary to integrate the detailed profiler transmissivity data over each 1.5m packer test interval to produce an average T and K for each packer interval. The integrated profiler results for each packer interval are compared to the packer results in Figure 10b. There are many similarities between the packer testing profile and the FLUTe K profile throughout most of the hole, but there are large differences deep in the hole. The packer tests show transmissive zones below 80 m, but the FLUTe method does not. Because the large differences occurred at the bottom of the hole where more pressure is needed to overcome the ambient water pressure, it is likely the packer test conductivity values are biased high because of flow between the packers and the borehole wall. However, leakage was not monitored using pressure transducers above and below the packers during the packer testing of this hole. The need for a more definitive packer data set for comparison with liner measurements led to very detailed packer measurements using improved equipment and procedures in boreholes MW-26 and MW (Quinn, PhD thesis 2009). These holes were packer tested from the bottom upwards with a 1.5 m test interval and packer inflation tests were conducted to ensure proper interval sealing. The hole above and below the packers was monitored for leakage with pressure transducers, and care was taken to ensure all step tests were completed after pressure equilibrium has been established in the test interval and in the open hole above and below the test interval. Numerous step tests were conducted in each interval beginning at very low flow rates to ensure Darcian flow. The FLUTe T profiles for these boreholes are shown in Figures 11a and 12a and these results integrated over the packer intervals are shown in Figures 11b and 12b with the packer testing results. The significant flow zones are well identified by both the packer and the 15

223 liner measurements in both holes. However, in MW-26, in the more permeable zone at 25 mbgs, the packer K is slightly higher than the profiling K. This is probably the result of slight packer leakage and/or non-linear flow during the FLUTe liner profiling. Overall, these results suggest that the liner measurements are representative of the borehole transmissivity. However, the liner T distributions as seen in Figure 11a and 12a give a higher spatial resolution of the distribution of permeable zones in the borehole than the packer tests and are done much more quickly. LIMITATIONS OF THE FLUTe K PROFILING METHOD The ability of the liner profiling method to identify individual transmissive zones (i.e., method resolution) increases as the descent velocity decreases and therefore, resolution always increases towards the bottom of the hole. In addition, because the profile is generated by covering flow zones as the liner descends, holes with the highest transmissive zones at the bottom will cause less resolution throughout the borehole length because of the high liner velocity throughout most of the hole. However, this does not diminish the accuracy of the full hole T value. The accuracy of the T values calculated from the velocity profile depends upon the degree of validity of the Thiem model for the particular hydraulic conditions of the borehole. The Thiem equation assumes that all of the flow from the borehole is Darcian (i.e., linear with respect to the driving gradient), so if the driving pressure is high, the flow near the borehole wall in the fractures may be non-linear with respect to the gradient resulting in an underestimate of the conductivity. To minimize this effect the driving head in the liner is kept below 4.6m (15 ft) in more conductive formations. Non-linear flow 16

224 effects have also been identified in conventional packer testing in fractured rock (Maini, 1971; Louis, 1972; Gale, 1975; Elsworth and Doe, 1986). Quinn (2009) shows that flow deviates from linearity at relatively low flow rates in constant head straddle packer tests, especially in the high permeable zones. The comparison to the packer tests indicate the error in the liner measurement of K based upon the Thiem equation can under estimate the actual conductivity by a factor of 2-3 in the very fast flowing intervals, most likely because of non-linear flow. However, in boreholes where there is a large inflow from a shallow fracture set above the blended head in the hole, the cascading water has a tendency to pull the liner into the hole without any applied head, and the high flow along the borehole wall may prevent proper sealing. This can be overcome by using an extremely large flow rate (400 l/min) to keep the driving head large enough to keep the shallow zone sealed, which most likely will lead to greater flow non-linearity. Flowing artesian holes can also be difficult because their high head may not allow for a large enough driving head for liner eversion. In some cases, the well casing can be extended feet (3-4.6m) above the ground surface with scaffolding to obtain the necessary excess head inside the liner. The head distribution throughout the formation will most likely vary with depth dominated by the head in the larger fractures (fractured zones). However, the blended head that is measured in the open borehole is assumed to represent the ambient head in the formation throughout the borehole. Multi-level monitoring systems installed at this site commonly shows that the depth discrete head changes < 5% from the ambient head throughout the borehole. Therefore, the error caused by using the blended head as the ambient head instead of the actual head distribution in the formation is relatively small. 17

225 The comparison of profiling results to continuous straddle packer tests in the dolostone holes at the Guelph site show close similarities and both methods identify non-darcian flow conditions. These conditions can be eliminated in packer testing by using appropriately low injection rates, but it is generally unavoidable in liner profiling. The liner profiling method is an effective and time efficient alternative to packer testing, and may serve to help identify specific zones for more in-depth packer testing in site investigations. CONCLUSIONS Hydraulic cross connection in fractured rock boreholes is an important issue at contaminated sites, and the FLUTe blank liner is effective sealing tool stopping all vertical flow in the borehole. During the liner installation K profiling is usually done providing the full transmissivity of the borehole as well as the depth-discrete distribution of the permeable features comprising this transmissivity. The T for each permeable feature is calculated from the Thiem equation and the sum of these can be compared to the full hole T as a check of method consistency. This method can accurately identify the aquifer/aquitard contact because of the differing transmissivities of the two units. This method is commonly applied in rock holes intended for installation of monitoring wells and multi-level monitoring systems, (typically 4-6 inch diameters), but is also used in larger diameter holes more typical of water supply wells (typically 12 inch diameter). Prior to the development of the liner profiling method, straddle packer hydraulic testing was the only method available for depth-discrete T determination in boreholes. However, for high resolution T profiles, the small packer test interval required makes packer testing 18

226 very time consuming. Therefore, in conventional practice, straddle packer tests are done at only a few intervals in each hole and these intervals are usually selected for investigation of high K based on core descriptions, geophysical logs and/or flow metering. The flexible-liner profiling method offers a new, more reliable and lower cost option for identifying the location of relatively high hydraulic conductivity zones prior to straddle packer testing. At sites where estimates for cubic-law-derived hydraulic apertures are desired, straddle packer tests of priority zones identified based on liner profiles is an effective complementary use of the two methods. Liner profiling can be cost-effectively deployed at most bedrock sites around the globe providing valuable insights into the number of potentially active fractures or transmissive features and their distribution to aid in site investigations and site conceptual model development. 19

227 ACKNOWLEDGEMENTS Funding for the technical development of this borehole testing method was provided by Flexible Liner Underground Technologies Limited (FLUTe) and the comparison of profiling with packer testing conducted at the Cambridge and Guelph sites was funded by the Natural Sciences and Engineering Research Council of Canada and the University Consortium for Field Research of Groundwater Contamination. Assistance during the field work was provided by Erik Storms of FLUTe, Paul Johnson, Bob Ingleton and Pat Quinn of the University of Waterloo. Pat Quinn also performed the excellent high resolution packer measurements for our comparison. We thank the owners of the two field sites: Syngenta Crop Protection Canada, Inc. and Guelph Tool, Inc. for their cooperation and site logistics support. 20

228 REFERENCES Bliss,J.C. & Rushton, K.R "The reliability of packer tests for estimating the hydraulic conductivity of aquifers." Quarterly Journal of Engineering Geology. 17: Braester,C. and R.Thunvik "Determination of formation permeability by doublepacker tests." Journal of Hydrology. 72: Cherry, J. A., Parker, B. L., & Keller, C "A new depth-discrete multilevel monitoring approach for fractured rock", Ground Water Monitoring and Remediation, vol. 27, no. 2, pp Gale,J.E "Assessing the Permeability Characteristics of Fractured Rock." Geological Society of America. Special Paper 189: Haimson, B. C. & Doe, T. W State of stress, permeability, and fractures in the Precambrian granite of northern Illinois, Journal of Geophysical Research, vol. 88, no. B, pp Hvorslev, M. Juul Subsurface Exploration and Sampling of Soils for Civil Engineering, Purposes, U.S. Army Engineer Waterways Experiment Station, Vicksburg, MS. Hvorslev, M. Juul, and Goode, Thomas B Core Drilling in Frozen Ground, Technical Report, No , U.S. Army Engineer Waterways Experiment Station, Vicksburg, MS. Hvorslev, M.J., Time lag and soil permeability in ground-water observations U.S. Army Corps of Engineers, Waterways Experiment Station, Bulletin 36. Lapcevic, P.A Results of borehole packer tests at the Ville Mercier Groundwater Treatment Site, NWRI Report. Lapcevic,P.A., K.S.Novakowski, and E.A.Sudicky "Groundwater Flow and Solute Transport in Fractured Media." In J.W.Delleur, editor, Groundwater Engineering Handbook. CRC Press Maini, Y.N., In-situ hydraulic parameters in jointed rock their measurement and interpretation, PhD Dissertation, Imperial College, London. National Research Council (NRC) Chapter 5 Hydraulic and Tracer Testing of Fractured Rocks. Rock fractures and fluid flow: Contemporary understanding and applications. National Academy of Science. Washington, D.C Neuman, S.P.,

229 Novakowski, K. S. & Bickerton, G. S "Borehole measurement of the hydraulic properties of low-permeability rock", Water Resources Research, vol. 33, no. 11, pp Pehme, P. E., Parker, B. L., Cherry, J. A., & Greenhouse, J. P "The Potential for Compromised Interpretations When Based on Open Borehole Geophysical Data in Fractured Rock", NGWA/U.S. EPA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Portland, Maine. Le Borgne, T., Paillet, F., Bour, O., & Caudal, J. P "Cross-borehole flowmeter tests for transient heads in heterogeneous aquifers", Ground Water, vol. 44, no. 3, pp Paillet, F. L. & LeBorgne, T "Multiple-scale characterization of fractured bedrock flow paths using a borehole flowmeter", Abstracts with Programs - Geological Society of America, vol. 37, no. 1, p. 4. Paillet, F. L "Characterizing large-scale aquifer properties with local-scale borehole flowmeter measurements", Abstracts with Programs - Geological Society of America, vol. 33, no. 6, p Price, M. & Williams, A "The Influence of Unlined Boreholes on Groundwater Chemistry: A Comparative Study Using Pore-Water Extraction and Packer Sampling", The Institution of Water and Environmental Management, vol. 7, no. 6, pp Quinn, P.M High Resolution Packer Testing in Fractured Desimentary Rock, PhD Dissertation, University of Waterloo, Waterloo, Ontario. Rushton, K. R "Impact of aquitard storage on leaky aquifer pumping test analysis", Quarterly Journal of Engineering Geology and Hydrogeology, vol. 38, no. 4, pp Sterling, S. N., Parker, B. L., Cherry, J. A., Williams, J. H., Lane, J. W., Jr., & Haeni, F. P "Vertical cross contamination of trichloroethylene in a borehole in fractured sandstone", Ground Water, vol. 43, no. 4, pp Tsang, Chin-Fu; Hufschmied, Peter; Hale, Frank V, Determination of fracture inflow parameters with a borehole fluid conductivity logging method, Water Resources Research, vol. 26, no. 4, pp , Apr Wenzel, L.K The Thiem Method for Determining Permeability of Water Bearing Materials, Water Supply Paper 679-A, USGS. Williams, J. H. & Paillet, F. L "Characterization of fractures and flow zones in a contaminated shale at the Watervliet Arsenal, Albany County, New York", Open- File Report - U.S.Geological Survey. 22

230 Wilson, J. T., Mandell, W. A., Paillet, F. L., Bayless, E. R., Hanson, R. T., Kearl, P. M., Kerfoot, W. B., Newhouse, M. W., & Pedler, W. H "An evaluation of borehole flowmeters used to measure horizontal ground-water flow in limestones of Indiana, Kentucky, and Tennessee, 1999", Water-Resources Investigations - U.S.Geological Survey. Williams, et al (fracture mapping) Ziegler, T., Determination of rock mass permeability, U.S. Army Engineers Waterways Experiment Station. Tech Report S76-2. p

231 Biographical Sketches Carl Keller has Bachelors and Masters degrees in math, physics, and engineering science from Valparaiso University and the Rensselaer Polytechnic Institute. He was employed with the U.S. Department of Energy and Department of Defense from 1966 to 1984 as an underground nuclear test containment scientist, developing a variety of models for multiphase flow in the earth. In 1989 he developed the first everting flexible liner system for collection of pore water samples from the vadose zone and in 1996 he produced the first flexible liner system for monitoring in the groundwater zone. He holds 13 patents concerning vadose zone and groundwater monitoring and other flexible liner methods. He established Flexible Liner Underground Technologies in 1996 where he is owner and chief scientist/engineer. John A. Cherry has geological engineering degrees from the Universities of Saskatchewan and California, Berkeley and a Ph.D. in hydrogeology from the University of Illinois, was a professor in the Department of Earth Sciences at the University of Waterloo between 1971 and 2006 and later a Distinguished Emeritus Professor at that university and is Director of the University Consortium for Field Research of Groundwater Contamination now based at the University of Guelph. His research is focused on field studies of contaminant behavior in groundwater and groundwater monitoring methods. 24

232 Beth L. Parker has a Bachelors degree in environmental science/economics from Allegheny College, a Masters degree in environmental engineering from Duke University, and was a project manager for groundwater characterization and remediation at large industrial facilities for several years prior to receiving her Ph.D. in hydrogeology from the University of Waterloo. Between 1996 and 2006 she remained at the University of Waterloo where she was a research faculty member in Earth Sciences. She became a professor in the School of Engineering at the University of Guelph in 2007 where she holds the NSERC Industrial Chair in Groundwater Contamination in Fractured Media. Her research involves field studies of transport, fate and remediation of chlorinated solvents in diverse hydrogeologic environments including fractured rock, clayey aquitards and sandy aquifers. 25

233 Clamp 1a. Liner attachment to casing 1b. Water addition 1c. Liner descent beneath water table Tether Air vent tube Air vent tube Initial blended head in borehole ΔH L H L Eversion Point EP H BH Figure 1. The basic stages in installation of blank FLUTe liner: a) top of liner from the reel is clamped onto the top of borehole casing, b) the liner is pushed by hand down into the casing and water is added to cause the liner to descend by eversion, c) the liner descends below the static water level in the hole and water is added to maintain a positive hydraulic head differential between inside the liner and the initial static water level in the hole. 1

234 Water Addition hose Pressure measurement Surface casing Velocity Meter Tension control Liner on reel (inside out) Initial H BH H L Fractures Figure 2. Illustration of the system components for the profiling method. The system controls and measures parameters relevant to the determination of the fracture location and calculation of discrete fracture flow rates. Photograph shows the field site equipment used to conduct the profiling measurements. 2

235 (a) (b) Velocity V2 V1 V1 Depth Z 1 Z 1 V2 dz Z 2 Z 2 A Flow rate into the fracture, ΔQ = A(V1-V 2), where V1> V2 T = ΔQ ln(r /r )/( 2 πδh ) in the interval Z to Z 0 w BH 1 2 Figure 3. Schematic illustration showing the parameters involved in the measurement of transmissivity of a single permeable feature (e.g. fracture). The liner velocity changes from V 1 to V 2 as the liner passes (shuts off) a fracture over depth increment dz. 3

236 Liner Velocity ( Δz/ Δt) No flow into wall Depth in hole (z) Large fracture Permeable bed Multiple fracture zone Permeable bed A monotonic fit to data Figure 4. Hypothetical liner descent velocity profile showing changes caused by several types of borehole features. This profile assumes perfect functioning of the field measurement equipment (i.e., no noise). The monotonic fit ignores the effect of an enlargement. No flow, but hole enlargement Large fracture 4

237 Excess head in psi Comparison of measured versus calculated pressure history beneath the liner in MW 26 measured pressure beneath liner calculated pressure beneath liner Depth in borehole in meters BGS Figure 8. The comparison of the measured pressure history in the borehole beneath the liner versus the calculated pressure history using the measurements of the liner at the surface. While the agreement is very good and the transmissivity results are essentially the same using either history, fast flowing holes (>10 gpm) do not usually show such good agreement and therefore the pressure beneath the liner is best measured for those situations. 8

238 z i+1 z i Θ D Θ H L dz H BH A dq V, i ti V i+1, ti+1 r w H r 0 amb (r) C L Figure 7. Diagram illustrating the parameters and concepts used for the mathematical framework of the transmissivity measurement. 7

239 velocity/unit driving head (m/s/m) Velocity/dH (m/s/m) depth (m) depth (m) Vel/dH-tP smoothed over second run first run 90 Vel/dH-TP smoothed over N Monotonic V/dH-TP smoothed over N 100 Figure 9. a) Three velocity curves showing the stages of data reduction for MW-24: black curve is the raw data, the pink curve is the smoothed data, and the yellow curve is the monotonic fit to the smoothed data which ignores any temporary drops in velocity. b) The comparison of two profiler velocities in the same hole. The red curve is the first run. The yellow curve was measured after the liner was withdrawn and reinstalled. Some fracture flows were apparently affected by the removal of the liner, but the prominent features are evident in both curves. 9

240 0 Velocity of liner WT Depth below surface Velocity peak when liner is released Depth of transient velocity decay to steady state condition Figure 6. Upon release of the liner, the liner velocity immediately peaks and then drops to the nominal steady state flow rate out of the hole. Thereafter, the enduring velocity changes are influenced by the sealing of the borehole s transmissive zones. 6

241 a) b) Liner velocity Z1 Z1 Enlargement Z2 Z3 Z4 Z5 Depth in hole Z2 Z3 Z4 Z5 With flow path in enlargement No flow from enlargement Figure 5. a) the blank liner moving through an interval of borehole enlargement, and b) idealized velocity graph for the enlarged hole zone showing the characteristic dip in the plot produced by an enlarged-hole interval bounded above and below by nontransmissive rock. 5

242 0 Liner Transmissivity (cm 2 /s) conductivity (cm/s) Depth (m) Depth (m) packer liner Figure 10. a)this transmissivity was calculated from the second measurement in MW- 24. The transmissivity peaks occur at each change in the velocity. b) This hydraulic conductivity comparison of the straddle packer results using the Lapcevic method with the profiling results in MW-24 show some significant differences between the two methods although many intervals are relatively close. The two large packer values below 70 m were suspected due to leakage and led to more careful packer testing in hole no. MW

243 0 Transmissivity over 0.10m intervals (cm 2 /s) Hydraulic Conductivity (cm/s) packer conductivity (cm/s) FLUTe liner conductivity MW Depth (m BGS) depth (m) Figure 11. a) This transmissivity distribution was calculated from a liner velocity measurement in MW-26. b) The liner transmissivity profile of MW-26 in (a) was integrated over the packer test intervals and is compared to the packer test-derived hydraulic conductivity profile measured very carefully using the revised methodology (Quinn, 2009) in the same hole. The correlation of major flow zones is good with the lower liner peak values probably due to more turbulent flow for the liner measurement. 11

244 0 Transmissivity over 0.10m intervals (cm 2 /s) Conductivity (cm/s) Depth (m BGS) Depth below surface(ft) packer conductivity liner conductivity Figure 12. a) This transmissivity distribution was calculated from a liner velocity measurement in MW b) The liner transmissivity profile of MW in (a) was integrated over the packer intervals and is compared to the packer conductivity profile measured very carefully in the same hole. The lower values for the liner measurement in the high flow zones is probably due to non linear flow conditions for the liner measurement. 12

245 20 ELEMENTS OF THE SANTA SUSANA FIELD LABORATORY SITE CONCEPTUAL MODEL OF CONTAMINANT TRANSPORT SITE CONCEPTUAL MODEL ELEMENT 1-8 A New Method for Measurement of Volatile Organic Contaminants(VOCs) in Rock Core DRAFT Prepared for: THE BOEING COMPANY THE NATIONAL AERONAUTICS AND SPACE ADMINISTRATION UNITED STATES DEPARTMENT OF ENERGY Prepared by: Beth L. Parker 1, Sean Sterling 2 1 School of Engineering, University of Guelph, Guelph, Ontario, Canada 2 Interra Geosciences and Engineering, Ottawa, Ontario, Canada Date: December 11, 2009

246 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 TABLE OF CONTENTS 1.0 ABSTRACT INTRODUCTION FIELD METHODS COMPARISON OF ROCK CORE POREWATER RESULTS TO OPEN WELLS AND MULTILEVEL SYSTEMS REFERENCES FIGURES... 9 Figure 1. DNAPL transport through fractures in fractured sedimentary rock Figure 2. Diffusion causes VOC halos along fractures Figure 3. Diffusion halo in generic sandstone Figure 4. Typical sampling distribution. Very close sample spacing is necessary to determine mass distribution and pathways Figure 5. Rock CORE VOC method Figure 6. Cored hole in contamination source zone Figure 7. Rock core with wooden blocks indication sample positions Figure 8. Rock CORE VOC analysis showing detailed TCE profile with depth Figure 9. Rock CORE VOC analysis vs. Shallow well VOC analysis Figure 10. Rock CORE VOC analysis vs. Shallow well VOC analysis Figure 11. Shallow well and deep multilevel design Figure 12. Comparison of multilevel and rock core data (hole 35B) Figure 13. Hydraulic head in a shallow well and multilevel system (location 46) Figure 14. Open hole cross connection showing transport of TCE from one zone to another Figure 15. TCE mass distribution (location 46) Figure 16. Conceptual profiles along plume centerline (stage 4) Figure 17. Conceptual TCE profiles along plume centerline (stage 4) Figure 18. Comparison of field data and model

247 1.0 ABSTRACT This paper presents a new method for measurement of VOCs in hard sedimentary rock core. This is sufficiently efficient and effective to produce detailed high resolution concentration profiles of volatile contaminants versus distance along the core. Previous methods for VOC rock core analysis provide accurate measurement, however, they are not designed to enable sufficiently large numbers of samples to be collected and processed in a practical manner for application in numerous holes drilled for characterization of industrial sites. This paper presents the results for the application of this new method for the first time. The field site selected for the method demonstration is the Santa Susana Field Laboratory near Simi, CA, located on fractured sandstone with shale interbeds. Two VOC versus depth profiles were produced and comparisons made with nearby conventional monitoring wells. The VOC core analyses have several advantages over monitoring wells for site characterization at contaminated fractured rock sites, including providing contaminant distributions in the vadose zone and in the rock matrix where nearly all the contaminant mass resides.

248 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December INTRODUCTION Information on contaminant distributions for plumes in fractured rock is limited due to the complexities of fracture networks and the practical limitations imposed by monitoring wells or other borehole sampling systems (Sterling 1999). Dense Non-Aqueous Phase Liquid (DNAPL) releases can lead to contamination at great depths and also create highly variable contamination distributions with high concentration zones existing nearby to low concentrations (Figure 1). Standard practices with conventional monitoring wells for contaminated site characterization in fractured sedimentary rocks are often inadequate due to large monitoring intervals that obscure rather than elucidate understanding about the contaminant distribution. Advances in multilevel monitoring wells has improved understanding of the complex and highly variable contaminant distributions in groundwater, however, even multilevel monitoring wells are often unable to address individual fractures and features where groundwater flow is occurring. Both conventional and multilevel monitoring wells have another shortcoming in that they are ineffective at sampling the contamination found in the low permeability rock matrix. The rock matrix has the ability to receive and transfer contamination between the fracture water and matrix porewater via diffusion. The rock matrix properties such as high porosity relative to fracture porosity, the low permeability relative to fractures, and the presence of organic carbon allow the rock matrix to have a large contaminant storage capacity and can become a long term sink and source for contaminants for decades. The long term implications due to contaminant interaction between fractures and the matrix means that comprehensive plume characterization in fractured rock requires information on contaminant occurrence in discrete fractures as well as in the rock matrix at many locations (Sterling 1999). This has led to an alternative approach to sampling groundwater, by focusing characterization efforts on the distribution of contaminants in the rock matrix. Lawrence et al. (1990) were the first to use analysis of rock cores to examine the spatial distribution of volatile organic compounds (VOCs), primarily TCE, in fractured porous rock. They measured the VOC content of many small samples from cores from the Chalk Aquifer in England, which showed large concentration differences over short vertical intervals. The VOC concentration profiles were used to identify horizons upon which the immiscible phase may have accumulated. Parker et al. (1994, 1997), used analytical models to examine the influence of matrix diffusion of chlorinated solvent DNAPL in a generic sandstone (Figure 2). They concluded that common solvents, such as TCE, can have immiscible phase disappearance from fractures due to dissolution and then diffusion-driven mass transfer from fractures to the rock matrix. VanderKwaak and Sudicky (1996) used a numerical model to determine the combined influence of diffusion with advection on the depletion of the immiscible-phase of chlorinated solvents and on plume characteristics in ideal discrete fracture 4

249 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 networks in a generic sandstone (Figure 3). These modelling studies are an indication that the dimensions of the diffusion halo formed after decades of plume existence should be sufficiently large to be identified by rock core sampling (Figure 4). This diffusive transfer can be a particularly strong influence in sandstone because although the matrix permeability is small, the matrix porosity is commonly large (typically in the range of 2 to 20%). For chlorinated solvents, the storage capacity of the rock matrix is also generally enhanced significantly by organic carbon dominated sorption. Therefore, diffusion driven mass transfer of contaminants from the fractures to the porous rock matrix has significant implications for source zone and plume behaviour in fractured sedimentary rock. These implications include: 1) for DNAPL contamination, diffusion causes transfer of the mass distribution initially present as DNAPL phase in fractures to a dissolved/sorbed phase mass residing in the rock matrix; 2) diffusive transfer of dissolved mass in groundwater flowing in fractures to the rock matrix can greatly slow the rate of plume front advancement; and 3) remediation of source zones and plumes becomes diffusion-dominated since most of the contaminant mass resides in the rock matrix. The transfer of contaminant mass from fractures to the matrix produces a contaminant imprint or diffusion halo in the matrix which can be used to identify contaminant migration pathways and to determine the distribution of contaminant mass in source zones and plumes. 3.0 FIELD METHODS In 1997, B.L. Parker and colleagues developed a method for determination of VOC concentrations in rock core that involved crushing the core samples with extraction in methanol over several weeks (Figure 5). Through collaboration with T. Górecki, the method evolved to rapid microwave extraction and very low detection limits. The advanced suite of procedures, referred to as the CORE (Characterization of Rock Environments) approach for discrete fracture network (DFN) investigations, include methods for sampling, extracting and analyzing contaminants present within the rock matrix to assess the effects of diffusion of contaminants from fractures into the rock matrix. Sterling (1999) outlines the field methods for the rock CORE TM VOC method. The initial step is to collect continuous rock cores, typically through the zone of known or suspected contamination (Figure 6). The core is collected in a five foot long stainless steel barrel equipped with a stainless steel split sleeve and extruded into a PVC tray lined with aluminum foil. The entire length of the core is covered with aluminum foil to minimize volatilization of organics from the core. After examining the core and identifying fractures and lithologic changes, a descriptive geological log is produced. The core is then sampled based on four criteria: 1) pre-set 4.6m (15ft) spacing above the water table and 6m (20ft) spacing in the saturated zone; 2) 0.15m (0.5ft) to 0.6m (2ft) spaced intervals on both sides of apparent fractures; 3) adjacent samples at distinct changes of lithology; and 5

250 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December ) adjacent samples as duplicates representing the same depth and lithology (Figure 7). The procedure for the collection and preservation of rock core samples was specifically designed to minimize the loss of VOCs from the samples (Sterling 1999). A sample size of approximately 75mm (3 inches) can be separated from each marked sample location using a rock hammer. Approximately 0.5cm (¼ inch) of the outside diameter should be chipped off to remove the surface of the core that may have been altered during drilling. The inner portion of the sample can then be crushed using rock-crushing device which employs a hydraulic press with a piston in a cylinder that minimizes exposure of the sample to air during and after crushing. This design minimizes the amount of contaminant loss due to volatilization by enclosing the sample. The crushed sample is then added to a 120mL glass sample bottle containing approximately 60mL of laboratory grade methanol and be preserved at 4 C. The methanol extracts the sorbed phase as well as the dissolved phase contaminants from the rock and also minimizes loss of these volatile constituents from the sample container during the extraction time period. The process of chipping, crushing, adding to methanol, and storing on ice generally takes between one and two minutes (Sterling 1999). The samples are then sent to a commercial laboratory for analysis. The porewater concentrations can be plotted against the depth at which the sample originated (Figure 8). Additional intact samples representative of each of the lithologies encountered are retained for physical parameter analyses (including bulk density, porosity, organic carbon content and diffusion coefficients) relevant for assessment of contaminant partitioning (e.g. for estimating equivalent porewater concentrations from the total mass concentrations obtained from the lab VOC analysis) and matrix diffusion. 4.0 COMPARISON OF ROCK CORE POREWATER RESULTS TO OPEN WELLS AND MULTILEVEL SYSTEMS A measurement of TCE in well water is an isolated measurement for the time of sampling. In contrast, a rock porewater TCE measurement is a historical fingerprint that reflects the presence of TCE in the nearby fracture over a much longer time frame. Therefore, the two measurements are not equivalent. However, there is value in the comparison of the two if the well has a historical record of TCE monitoring. A comparison of TCE concentrations in groundwater collected from open wells and concentrations measured in rock porewater was conducted by Sterling (1999). The same depth intervals were used for comparison in both wells. Two sets of results from the study can be seen in Figures 9 and 10 which compare the open well groundwater concentrations for two wells at the Santa Susana Field Laboratory (SSFL) with rock porewater concentrations found at equivalent well depth intervals in separate, nearby (12-15m) coreholes. The TCE concentrations in the rock matrix porewater correspond closely to TCE levels in groundwater samples from comparable depth intervals 6

251 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 in the depth-blended open wells. Therefore, the VOC data obtained using the rock core sampling, crushing and processing procedures produce reliable, consistent results. TCE concentrations in groundwater sampled from multilevel monitoring systems can also be compared to rock core porewater concentrations collected from the corresponding depth intervals. Sterling (1999) compared the two concentrations using the rock CORE VOC method and a multilevel monitoring system (Figure 11). The results of the comparison show a good correlation between the groundwater concentrations and the rock matrix porewater concentrations (Figure 12). Some of the groundwater samples, however, exhibit increased groundwater concentrations that are believed to be influenced by borehole cross connection. Borehole cross connection can occur when the borehole remains open between the drilling of the hole and the installation of the multilevel system, if a change in head values exist within the bored interval (Figure 13). This can cause contaminated groundwater to move from one fracture zone to a non contaminated zone if the hydraulic heads are sufficiently different (Figure 14). The newly introduced contaminated water will then begin to diffuse into the rock matrix porewater. These cross connection effects are difficult to eradicate by purging because of slow outward diffusion and may lead to an increased groundwater contamination concentration. Thus, rock porewater VOC data are a better indication of actual concentrations than groundwater samples collected from boreholes after drilling. Also, rock core porewater VOC data allows the extent of the effects of cross connection to be recognized in a borehole. Another benefit of the rock CORE method is the fact that contaminant mass can be detected in the vadose zone and shallow groundwater zones which can contain very large amounts of the contaminant mass (Figure 15). This is not possible with shallow wells or multilevel monitoring systems. The concentration profiles of VOCs obtained through the rock CORE VOC method have also been shown to be consistent with the matrix diffusion conceptual (Figures 16 and 17) and numerical (Figure 18) models results. 7

252 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December REFERENCES Lawrence, A. R., Chilton, P. J., Barron, R. J., & Thomas, W. M. A method for determining volatile organic solvents in chalk pore waters (southern and eastern England) and its relevance to the evaluation of groundwater contamination. Journal of Contaminant Hydrology 6[4], Parker, B. L., Gillham, R. W., & Cherry, J. A. 1994, "Diffusive Disappearance of ImmisciblePhase Organic Liquids in Fractured Geologic Media", Ground Water, vol. 32, no. 5, pp Sterling, S. N. 1999, Comparison of Discrete Depth Sampling Using Rock Core and a Removable Multilevel System in a TCE Contaminated Fractured Sandstone, Master's thesis, Earth Sciences Department, University of Waterloo. VanderKwaak, 1., and E. Sudicky, Dissolution of non-aqueous-phaseliquids and aqueous-phase contaminant transport in discretely-fractured porous media, Journal ofcontaminant Hydrology, 23, 45-68,

253 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December FIGURES Figure 1. DNAPL transport through fractures in fractured sedimentary rock. 9

254 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 2. Diffusion causes VOC halos along fractures \ Figure 3. Diffusion halo in generic sandstone 10

255 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 4. Typical sampling distribution. Very close sample spacing is necessary to determine mass distribution and pathways. 11

256 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 5. Rock CORE VOC method 12

257 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 6. Cored hole in contamination source zone. Figure 7. Rock core with wooden blocks indication sample positions. 13

258 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 8. Rock CORE VOC analysis showing detailed TCE profile with depth. Figure 9. Rock CORE VOC analysis vs. Shallow well VOC analysis. 14

259 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 10. Rock CORE VOC analysis vs. Shallow well VOC analysis. Figure 11. Shallow well and deep multilevel design 15

260 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 12. Comparison of multilevel and rock core data (hole 35B) Figure 13. Hydraulic head in a shallow well and multilevel system (location 46) 16

261 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 14. Open hole cross connection showing transport of TCE from one zone to another. Figure 15. TCE mass distribution (location 46) 17

262 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 16. Conceptual profiles along plume centerline (stage 4) 18

263 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 1-8 DRAFT December 2009 Figure 17. Conceptual TCE profiles along plume centerline (stage 4) Figure 18. Comparison of field data and model 19

264 A New Downhole Passive Sampling System to Avoid Biases from Sample Handling Sanford L. Britt 1, Beth L. Parker 2 and John A. Cherry 2 1 ProHydro Inc; 1011 Fairport Road, Fairport, NY (corresponding author) 2 School of Engineering, University of Guelph, Guelph, ON, N1G 2W1 Draft, September 18, 2009 Britt Page 1

265 Abstract: This paper introduces a new downhole groundwater sampler that reduces biases due to sample handling and exposure and causes minimal disturbance on the natural flow conditions present in the formation and well. The In Situ Sealed, ISS, or Snap sampling device has three main components: (i) removable/lab-ready sample bottles; (ii) a sampler device to hold the removable bottles in an open position in the well; and (iii) a trigger line for lowering the sampler downhole and for triggering closure of the bottles. Prior to deployment, each bottle is set in an open position on both ends to allow flowthrough during installation, and during equilibration downhole. To sample, bottles are triggered to close without purging the well. As such, this is a passive or non-purge sampling method. The sample is retrieved in a sealed condition and is not exposed prior to analysis. Results of VOC, metals, and other inorganic chemical analyses are compared to those sampled by traditional methods at several sites. The ISS samples typically yielded slight to somewhat higher VOC concentrations compared to other methods. In one case study, significant chemical-specific differentials were discernable. For metals, a comparison of ISS results to filtered and unfiltered purge samples indicated negative bias due to purge sample filtration but positive bias for unfiltered purge samples. Inorganic constituents showed parity with traditional methods. This sampler enables a broad range of sampling purposes, avoids low VOC recovery bias, and enhances prospects for better reproducibility of results; while avoiding sampling complexity and purge wastewater disposal. Britt Page 2

266 Introduction Two general approaches are available for collecting water samples from wells or open boreholes: (i) uphole sampling: pump water from the well and collect the displaced sample at surface from the pumped flow or (ii) downhole sampling: lower a sampling vessel down the well to the desired depth where the water sample is (ideally) collected in a sealed container. In the 1970 s, the first downhole groundwater samplers were reported in the literature by Tate (1973) and Frost et al. (1977). These samplers were developed for investigation of inorganic groundwater chemistry and they used flow-through design where the sampling vessel is open-ended allowing water to flow through the vessel during descent in the well. At the appropriate depth the vessel ends are closed by remote control and the water isolated in the vessel is raised to surface. In the 1980 s, the need arose for investigations of volatile organic contaminants (e.g. chlorinated solvents) and other constituents with requirements for very low detection limits (i.e. micrograms per liter) and it was recognized that uphole sampling can cause the water chemistry to be unrepresentative of in situ chemistry (Pettyjohn et al. 1981; Barcelona et al. 1986). Downhole samplers were then developed to minimize escape of dissolved gasses such as methane and carbon dioxide and volatile organic contaminants. Gillham (1982) developed a technique for downhole sampling that used standard low-cost polyethylene syringes with simple modifications so that each sample is taken in a new syringe, avoiding need for sampler decontamination. The syringe, with the handle removed, is attached to a plunger connected to a tubing line used to lower the syringe to the sampling depth and a hand pump at surface applies vacuum to activate the plunger, drawing water into the syringe. At surface the syringe tip is immediately capped for transfer to the laboratory. This device has been used for studies of redox parameters, metals and radon (Gillham 1982). Other types of downhole samplers are reported in the literature. For example, Pankow et al (1984, 1985) developed two types of downhole samplers specifically for organic contaminants. These devices use small cylindrical cartridges filled with sorbent material. The device is lowered to the sampling depth where the water is drawn through the sorbent material for capture of the contaminant mass. The cartridges are transferred Britt Page 3

267 to the laboratory for thermal desorption and analysis. Like the Gillham syringe sampler, the cartridge method avoids sample exposure between sampling and arrival of the sample in the laboratory. A separate track for development of downhole samplers focused on sampling deep boreholes (>300m) for studies concerning nuclear waste repositories and deep hydrochemistry. For example, Nurmi and Kukkonen (1986) and Sherwood Lollar et al. (1994) describe different versions of such samplers. These and other deep samplers constructed of stainless steel or other metals are relatively expensive and therefore not intended for dedicated sampling. Hence, sample transfer to the other containers occurs in the field site so samplers can be reused. Concurrent with the development of various downhole samplers in the early 1980s, regulatory guidance in the United States pushed the groundwater industry towards vigorous well purging prior to sampling (e.g. removal of three-to-five well volumes) (US EPA 1986) and this caused uphole sampling to become standard practice. The well purging guidance was based on the premise that purging removes the stagnant water from the well prior to sampling. However, Robin and Gillham (1987) and Powell and Puls (1993), among others, have argued (and shown) that the aquifer water in the screened interval of wells is not stagnant. These investigators illustrated that water does indeed flow naturally through the screen zone of the well under background gradients, given hydraulic communication between the well and aquifer. The screened interval itself functions as a very high permeability zone, as indicated by borehole dilution tests (Freeze and Cherry 1979). Low flow purging and sampling (Puls and Barcelona, 1996) in many cases also relies on flow-through in wells because the method typically removes only a portion of the water within the screen interval (ASTM 2002). Passive sampling generally relies on this phenomenon as well (ITRC 2004; 2007). Well purging has three negative aspects: it generates contaminated water needing disposal; it creates hydrologic disturbance in the aquifer prior to sampling; and it requires labor. The perception that monitoring wells must be purged prior to sampling is losing its appeal as the non-purge method (i.e. passive sampling) regularly provides similar data sets (Church 2000; Vroblesky and Peters 2002; Parsons 2003; Parsons 2005, Parker and Mulherin, 2007). Britt Page 4

268 One non-purge sampling method, which has gained wide use primarily because of low cost, is the polyethylene diffusion bag (PDB) sampler. This sampler is a flexible polyethylene bag, filled with non-contaminated water that is positioned at the sampling depth for the time period necessary for equilibration of the water chemistry inside the bag with the surrounding well water. (Vroblesky 2001; Vroblesky and Campbell 2001). This method is described as passive sampling because there is no active transport of water to the sampling device water passively flows to and by the device under ambient/natural aquifer flow. The PDB method is limited to volatile organic compounds that readily diffuse across the polyethylene membrane. To collect a sample, the PDB is retrieved from the well and the water is drained into the sample container by pouring or using a straw drainage device. Since the introduction and acceptance of the PDB sampler, newer passive sampling approaches have been developed and this paper describes features and results of several field investigations for one of these devices: the Snap Sampler or In Situ Sealed (ISS) downhole sampler. This device is unique among the passive devices in that it seals samples in the containers that are used to transport sample to the laboratory. In many cases, samples are unexposed from the in situ condition in the formation screen interval to the laboratory equipment that analyzes the samples. The overarching goal of groundwater sampling is to collect samples that are the closest feasible representation of the in situ condition in the aquifer (Yeskis and Zavala 2002) and therefore it is desirable to avoid sample exposure at any time prior to the chemical analysis. The sorptive samplers developed by Pankow et al. (1984, 1985) avoid sample exposure because the sample cartridge is entered directly into the analytical procedure in the laboratory. However, the sorptive cartridge method is limited to organic contaminants and it requires a thermal desorption step in the analytical procedure that is not standard in commercial laboratories. In contrast to the cartridge method, the PDB method has gained substantial acceptance because of low cost and its development followed the change in standard practice, away from vigorous purging. However, the PDB method has limitations: it does not avoid transfer of the sample in the field to a second vessel (i.e. a VOA vial), and it is only suited for hydrophobic volatile organic Britt Page 5

269 contaminants such as benzene and PCE; but cannot be used for chemicals such as MTBE, MEK, or 1,4-dioxane (Vroblesky 2001; ITRC 2004). The ISS/Snap sampler is designed to avoid these disadvantages. By keeping the advantages and eliminating the disadvantages of diffusion-based sampling, the ISS sampler is aimed at low long-term cost and improved data quality for many analyte types. The system minimizes possibilities for biases in groundwater monitoring results due to procedural steps and exposure at all stages between the moment the sample is sealed in the sample vessel to the moment the sample enters the analytical system in the laboratory. Methods Description and Operation of the ISS/Snap Sampler The ISS sampler is designed to hold specialized double end-opening sample bottles. The bottles have an internal Teflon-coated stainless steel spring that retains two Teflon end caps onto each end of the bottle. A release pin system on the sampler holds the bottles open during a user-defined downhole equilibration period. The sampler is lowered downhole with a triggering line that supports the sampler at a specific position downhole. A triggering mechanism allows the user to trip the sample bottles to close from the well head without the disruption of a messenger system or movement of the sampler downhole. The triggering line consists of a polyethylene tube with an internal Teflon-coated stainless steel wireline. The trigger is hung from the well head with a support ring. Activation of this simple mechanical system consists of maintaining the static position of the outer tube which is attached to the support ring at the well head and pulling the internal cable to trigger closure of the bottles. There are pneumatic and electric triggering systems for applications deeper than ~50ft/15m. The trigger system(s) retract the release pins on the sampler(s) downhole freeing the Teflon caps to close onto each bottle and sealing the sample in situ. Figure 1 illustrates the deployment configuration of a string of ISS/Snap samplers. One sampler is required for each bottle that is needed for laboratory analysis. Samplers can be attached in a series for multiple bottle collection. There are three bottle sizes that the ISS/Snap sampler can accommodate: a 40ml glass VOA bottle that is compatible with standard laboratory autosampler equipment; a 125ml high density polyethylene (HDPE) bottle; and a 350ml Britt Page 6

270 HDPE bottle. The 40ml and 125ml bottles and samplers will fit into 2-inch (5cm) and larger monitoring wells. The 350ml bottle will fit into 4-inch (10cm) and larger monitoring wells. Preparation for deployment and sampling consists of inserting sample bottles in the samplers, setting the bottle caps open onto the release pins, attaching the trigger line, and lowering downhole (Figure 2a). The samplers are left in the well in an open position for the well to restabilize to undisturbed ambient flow after insertion of the device. This period can be short if sampling objectives allow it (hours or days) or for the entire period between scheduled sampling events (3-6 months or more). Once the samplers are triggered to close, the trigger line and samplers are then retrieved and bottles removed from the samplers. The bottles normally have no headspace air and in many cases can be sent to the laboratory without exposing sample to atmospheric air. Surface preparation of the sample bottles consists of clipping the release pin tabs on the vial caps and securing a septa cap on each end of the bottle. If acid preservation is required, there is a cavity on the vial cap sized to accept 0.5 ml of 1:1 HCl (Figure 2b). To add acid to the sample, a small amount of acid is added to the cavity, and the membrane is pierced with a piercing tool provided with the samplers. The small hole allows acid to mix with the sample. Adding acid to the cavity before piercing the cap membrane prevents exposing sample to air or adding an air bubble during the piercing process. Once the membrane is pierced, more acid is added to form a positive meniscus on the vial cap cavity. The septa screw cap is then applied, similar to standard VOA bottle filling procedures. After the sample is collected, the sample bottles remain closed at all times in either preserved or unpreserved preparation. The method avoids the open air transfer step, resulting in the sample never being exposed from the in situ condition in the well to the laboratory analytical equipment. Standard laboratory autosampler equipment can be used with no special considerations. Sampling with this method is relatively rapid compared to purge sampling and can be accomplished with little equipment. In dedicated applications, the only items brought to field site are replacement bottles, a water level meter, a cooler, and documentation forms. There is no wastewater generated from purging or extra sample waste all water collected is submitted with the bottles to the laboratory. Typical time to Britt Page 7

271 trigger, retrieve and redeploy the ISS samplers in a well is usually in the range of 10 to 20 minutes. Figure 3 depicts the process of deployment and sampling using this downhole sampler. Field Investigations Several field-based studies of the ISS sampler are presented. These studies include side-by-side comparisons to traditional uphole pump sampling approaches, and data reductions of those comparisons designed to identify differences from the subject downhole sealed approach. Data sets from multiple sites were selected to present results for a wide range of chemical classes and hydrogeologic conditions. Table 1 highlights site conditions and sampling approaches. Chatsworth Site, California Comparison field testing of the Snap Sampler and low flow purging and sampling was conducted at a fractured and jointed sandstone bedrock site in southern California. The Santa Susana Field Laboratory (SSFL) is a former rocket engine test facility where chlorinated solvents were released during test activities. Three wells were used to compare Snap Sampling and low flow purging. Depth of bedrock completions ranged from 80ft (24m) to 170ft (52m). Saturated open intervals ranged from 30ft (9m) to 70ft (21m). Primary constituents of concern included TCE and its degradation by-products. The low flow sampling method comparator consisted of deployment of a bladder pump alongside ISS/Snap Samplers at one depth in each of the three wells. The ISS samplers and bladder pump were deployed 6 days in advance of sampling. At the time of sample collection, the ISS samplers were triggered to close and were left in place while low flow purge sampling was conducted. Parameter stabilization protocol using an above-ground flow through cell was used to establish readiness for sample collection (Puls and Barcelona 1996; ASTM 2002). Following collection of pumped samples, ISS sampler were retrieved and prepared for laboratory submittal without exposure. All samples were tested at a commercial laboratory by EPA method 8260B for volatile organic compounds. Britt Page 8

272 Guelph Site, Ontario, Canada A second field site was located in Guelph, Ontario, Canada, at a chlorinated solvents release site with thin overburden over fractured and jointed dolostone bedrock. Five wells in the study were each completed in the shallow bedrock with relatively short open intervals ranging from meters. Depth to completions ranged from 3-6 meters. Depth to water was 2-4 meters below grade. Primary constituents of concern are PCE, TCE and their degradation by-products. The sampling methods selected for comparison to the ISS sampler included two different approaches to sampling: low flow purge sampling and passive PDB sampling. The PDB allowed a direct, depth-specific, comparison. Low flow allowed a comparison of an active sampling method collected from the same depth interval in the bore. Peristaltic pumping was used for the low flow sampling approach. Care was taken to minimize the shortcomings of negative pressure pumping with polymeric tubing (Parker and Ranney, 1998). This included deployment of the siphon tube for two weeks along with the ISS samplers; collecting the sample from tubing backflow; and careful bottle filling. Using the backflow method, the degree of negative pressure and exposure to discharge tubing experienced by the water sample is minimized through collection of sample from the lower portion of the discharge tube after purging is complete. This is accomplished by stopping the peristaltic pump after purge stability criteria are met, removing the siphon tube from the well, and either reversing the pump direction or venting the upper end of the tube to drain sample into the sample bottle at the well head (Chapman et al., 2007). ISS samplers, PDBs and low flow pump tubing were deployed in each well for subsequent sampling. Two sets of paired ISS/PDB were deployed in each well (paired shallow and paired deep). The PDB was always attached just below the ISS sampler. Low flow pump tubing attached to shallow ISS/PDB pair. Downhole equipment was left to equilibrate with the natural flow regime of the aquifer and boreholes for 15 days prior to sampling. Sampling was conducted at each well by first triggering the equilibrated ISS samplers. After triggering, purge sampling was commenced without removing the ISS samplers or PDB samplers from the well. The ISS samples were sealed, so flow around the samplers during purging would have no effect on the samples. PDB samplers require Britt Page 9

273 many hours or days to re-equilibrate, therefore concentrations in the PDB was not expected to change substantially for the brief time the well was purged (less than 35 minutes in all cases). After purging, ISS samplers, PDB samplers and pump tubing was removed from each well. Purge samples were collected by draining or reverse pumping the lower end of the pump tubing into a VOA vial; ISS samples were prepared by capping the vials; and PDB samples were prepared by clipping a corner of the bag and carefully pouring into a VOA vial. All samples were analyzed for VOCs by GC/MS methods at the University of Waterloo. Morgan Hill Site, California Comparison field sampling was conducted in semi-consolidated overburden deposits among foothills south of the San Francisco Bay area. This private site was sampled for VOCs and perchlorate at 14 shallow wells using the ISS/Snap Sampler and volume-based purge methods. Depth of sampling intervals ranges from 15ft (5m) to 70ft (21m). Well screen intervals were all less than 10ft (3m) long. ISS samplers were deployed two weeks in advance of sampling. When sampled, the ISS samples were retained in the in situ deployed containers for submittal to the analytical laboratory. During the same day or the next, each of the test wells was purged with a bottomemptying bailer, or submersible pump. Sampling was conducted using a bailer and pouring sample into laboratory containers in all cases. All samples were tested at a commercial laboratory by EPA method 8260B for volatile organic compounds and perchlorate by EPA Method Hillside Site, New Jersey Comparison field testing at this facility consisted of deploying ISS samplers in 17 wells approximately 2 weeks in advance of scheduled volume-based purge sampling. The site geology consists of fluvial silty sand and sand deposits. Water occurs approximately 30-40ft (9-12m) below grade and wells are screened up to 60ft (18m) below surface. Well screen range from 10 to 20 feet (3-6m) in length and typically have saturated screen lengths under 20 feet (6m). Some wells were sampled at multiple depths during the tests. Volume-based purge sampling consisted of bailing three well volumes Britt Page 10

274 using a bottom emptying bailer and sampling with the same bailer. Purge samples were tested by two methods. One method placed whole water samples into acidified laboratory containers for arsenic analysis. The other method included a filtration step where the bailed sample was first filtered with a 0.45-micron filter prior to placement in the acidified laboratory container. ISS/Snap samples were transferred to acidified laboratory containers directly without filtration. All samples were tested at a commercial laboratory by EPA 6000 series methods for metals. McClellan Site, California The former McClellan Air Force Base is located in an area of thick unconsolidated silt, sand and gravel fluvial deposits within the San Joaquin Valley of California. Twenty wells were tested using several passive sampling methods, including the ISS sampler and two purge methods and reported by the US Army Corps of Engineers and Parsons Engineering Science (Parsons 2005). Ten ISS test wells ranged from about 110ft (33m) to 170ft (52m) below grade. Water levels were relatively consistent across the site, with depth to water ranging from about 95ft (29m) to 110ft (33m). Well screens are generally short, with open intervals of 10ft (3m), with a few of 15 or 20-foot length (4.5-6m). Passive methods included equilibration in the test wells for 1 to 3 weeks. Following retrieval of all of the passive devices, purge sampling was conducted. Purge sampling consisted of low purge rate sampling followed by a volume-based purge sampling approach. Purge sampling was conducted using new polyethylene tubing (not Teflon-lined tubing). A submersible electric pump was used to purge the wells at low flow rates while monitoring stability parameters using an uphole flow-through cell (Puls and Barcelona 1996; ASTM 2002). Samples were collected from the discharge stream after purge stability was achieved. Additional volume-based purge sampling was also conducted during the field work but analysis of those results was not part of this evaluation. Samples were tested for VOCs, 1,4-dioxane, metals and anions. VOC recoveries using the different methods are the focus of the present assessment. All samples were tested at a commercial laboratory by EPA method 8260B for volatile organic compounds. Britt Page 11

275 Los Angeles Site, California A private site in Los Angeles County was used to conduct a comparison study of sampling variability. The test site geology consists of fluvial silts, sands and gravels with well depths ranging from approximately 40 to 60 ft (12-18m). Water occurs about 30-35ft (9-11m) below grade and saturated well screens are 10 feet or less. Three wells were sampled with the ISS/Snap method and volume-based purge sampling for nine consecutive quarterly-sampling events. The ISS samplers were set at mid-saturated screen and sampled each quarter immediately in advance of purge sampling. Purge sampling consisted of pumping three to five well volumes with a submersible pump and collecting samples using a new bottom emptying polyethylene bailer. ISS samplers were redeployed after purge sampling in preparation for the subsequent sampling event the next quarter. VOC analysis results for the two methods were compared over time to assess how individual contaminant concentrations changed from event to event for each sampling method. All samples were tested at a commercial laboratory by EPA method 8260B for volatile organic compounds. Results Each field site yielded results that illustrate one or more data-quality improvements of the sealed-in-situ sampling approach. For sites where VOC differentials were assessed, VOC concentrations were usually somewhat higher using the in situ approach (Guelph, Morgan Hill, McClellan). Chatsworth, on the other hand, did not show substantial difference between the purge and ISS methods. In the comparisons among non-volatile constituents, concentration equivalence between the purge samples and ISS samples was very good (Hillside, Morgan Hill, McClellan). Apparent among all of the comparisons of the ISS method to purge methods or alternate passive methods was the strong correlation among the them. Correlation coefficients were very high despite the wide variety of sampling approaches. Britt (2005) suggested that in-well mixing effects may tend to homogenize stratification in wells, creating a flow-weighted averaging effect, limiting the influence of inflow heterogeneity. This may partly explain why different sampling approaches often yield similar results. However, there were findings in these six field Britt Page 12

276 investigations that point to potential improvements in sampling methodology that can be achieved through in situ collection and sealing of samples. Figure 4 illustrates where the sample comes from in different field sampling scenarios. During purge sampling, early time discharge incorporates water closer to the pump intake, while later time discharge incorporates larger volume and distance from the pump. Figure 4 shows these pumping-time envelopes in contrast to the passive approaches included in this study. The passive approaches collect either a time-weighted sample in the case of the PDB or an instantaneous sample in the case of the ISS sampler. As such, the passive samples collect water at (or very near) the deployment position in the well at the time of collection. Early time purge samples (i.e. low flow/low volume purging) collect water from nearly the same position in the well, assuming the pump is placed at the same position. Longer purge times and larger purge volumes interrogate larger portions of the well and eventually the formation adjacent to the well and beyond. However, water delivered to the well under ambient flow-through conditions may be effectively the same in many cases, yielding similar results for this sampling approach. Data from these field studies illustrate these phenomena. Figures 5 and 6 contain data from the Guelph site, Chatsworth site and Morgan Hill site. The Guelph data show both low flow and PDB results compared to ISS sampler results. Trendline slope, correlation coefficient and median relative percent differences are shown on the plots. All three sites show VOC results, while the Morgan Hill site also includes perchlorate data. For the Guelph and Morgan Hill sites, PDB and purge sample data have, on average, slightly lower in VOC concentration that than the ISS sampler. This is illustrated with the Y slope of the trendline greater than 1 and a positive RPD. In the case of the Chatsworth site, the y-slope and RPD are very close to neutral for VOCs, as is perchlorate at Morgan Hill. These comparisons highlight differentials among methods and analyte type. For constituents that are not substantially affected by air or polymer exposure (e.g. perchlorate), or are collected using low-bias methods (e.g. pre-deployed bladder pump using Teflon bladder and tubing), results are remarkably similar. For those comparisons where there was a differential, the purge method (e.g. peristaltic pump, bailer), and/or Britt Page 13

277 collection method (bailer, pouring), and/or exposure to unequilibrated plastics (tubing, bailer) contributed to the low bias. For the McClellan study site in Sacramento, California, data were reanalyzed to find additional clues about VOC recovery apparent in the other field data. The McClellan data is extensive and allows comparisons among the variety of the sampling methods explored. One particular aspect of the results was the strong low bias produced by low flow sampling method compared to the other methods at this site. There were multiple causes of low bias in this particular study. An electric submersible pump was employed using new polyethylene tubing in each well. The tubing was not Teflon lined and was deployed immediately in advance of purging. Depth to sampling positions were often well over 100ft (30m), so extensive exposure to the tubing occurred. Parker and Ranney (1998) showed exposure to polymer tubing, especially at low flow rates tends to promote sorptive VOC loss. Sample handling of the low flow and most passive samples also added to the bias. Samples were poured into VOA vials in the field, then re-transferred into different vials in the laboratory to accommodate autosampling equipment at the contract lab. The ISS sampler did not experience exposure to tubing or either of these transfer steps. The ISS sampler consistently yielded the highest VOC concentrations. The low flow method consistently yielded the lowest VOC concentrations. Table 2 tallies the relative recovery of all VOC by the different methods. Best-fit trendline slopes and RPDs those reported by Parsons (2005) in their data analysis. Beyond the obvious VOC recovery differentials among the methods are chemical specific differentials that support the proposition that bias is due to exposure to air and exposure to sorptive polymer materials. Figure 7 plots recoveries of several volatile organic chemicals as a percentage of the presumed full recovery of the ISS sampler, which in all cases yielded higher results. Figure 7a shows recoveries with chemicals identified by octanol-water partitioning coefficient (Kow). Chemicals tended to cluster among sampling methods, with poorer recovery percent associated with higher Kow. Similarly, Figure 7b plots recoveries separated by Henry s vapor partitioning coefficient. Again, chemicals tended to cluster among sampling methods, with poorer recovery percent associated with higher Henry s constant. This is the only dataset to date the authors are aware of that allows such analysis. Differentials of this magnitude are rare these are by far the largest Britt Page 14

278 identified. The high differential recoveries and large size of the overall database is likely the reason individual associations could be developed. Differences are statistically significant in many if not most cases. Britt (2006) spoke at length about individual constituent differences and the statistical tests employed. The Hillside New Jersey investigation was directed at a different aspect of chemical constituent recovery. At this site arsenic is a primary constituent of concern. Sampling for metals in groundwater has been historically problematic due to entrainment of a non-mobile fraction during purging (Puls et al. 1990). When purging wells elevates turbidity, samples are commonly filtered to remove the artifact solids from the sample. Unfortunately, filtration removes the mobile colloidal fraction as well. Multiple filtration methods have been employed to delineate the mobile fraction from artifact (e.g. Bailey et al. 2005). Ideally, however, a whole water sample capturing the mobile fraction but not adding artifact particulates is the best approach. The ISS sampler does not mobilize particulates, but it does allow collection of a whole water sample with naturally-mobile colloids included. During Hillside field sample collection effort, purge samples were collected and prepared for analysis by two methods: 1) direct transfer to acidified laboratory bottles, and 2) filtration with 0.45micron filter prior to transfer to acidified bottles. ISS/Snap samples were simply acidified after collection without filtration. Figure 8 illustrates results of the comparison of ISS samples to unfiltered purge samples and ISS samples to filtered purge samples. Filtered purge samples tended to be lower on average than unfiltered ISS samples, while unfiltered purge samples tended to be higher than unfiltered ISS samples. This is expected as the ISS sampler captures colloidal material, while unfiltered purge sample incorporates colloidal material plus artifact particulates. A filtered purge sample removes all particulates, including the mobile colloid fraction. The Los Angeles site data analysis compared long-term consistency of the sealed in situ method versus a traditional purge and bail-sample approach. ISS samples were consistently collected at the same fixed depth in the well and contained in sealed vials when retrieved. Purge samples were collected using a portable pump deployed above the screen to purge multiple well volumes from the well and sampled with a disposable bailer. Over two years, ISS sampling and traditional purge sampling yielded Britt Page 15

279 approximately 181 comparators of concentration from one calendar quarter to the next calendar quarter. Overall, purge results changed from one event to the next approximately 40% more than the ISS samples changed. This includes all factors influencing concentration, including the actual change in ambient contaminant concentration at the well, and any artifacts of either of the sampling procedures. The actual ambient concentration is incorporated into this 40% differential. Therefore, 40% reflects the low end of the error. Take, for example, a concentration of 100 that changes to 120 from one event to the next for the ISS, and 100 changes to 128 for purge. That differential is a 40% greater change for the purge result (28/20=1.40). However, the actual groundwater concentration underlying the change likely changed as well. The unknown actual groundwater concentration may have changed from 100 to 115, for example. That example translates to a corrected change of 5 for the ISS vs. 13 for purge ( =5; =13), a 260% difference. Since the actual groundwater concentration cannot be isolated, the difference in concentration change is the closest proxy, with the understanding that it is a minimum error estimate. Figure 9 illustrates concentration changes for TCE over the course of the comparison. Both sampling approaches result in event-to-event changes that are in the same direction, but the magnitude of the changes is reduced with the downhole in situ approach. The overall range was smaller for the Snap ( mg/l to mg/l) compared to purging ( mg/l to mg/l). The median RPD from event to event for purge sampling was 94%, while it was 35% for the ISS sampler, a substantial reduction that is indicative of an approach that eliminates often unknown, uncontrolled, or uncontrollable sources of data variability. Discussion Data quality is a controlling factor in selecting groundwater sampling methods. The change from strong-purge sampling to low flow sampling was prompted by the desire improve data quality by reducing mobilization of normally-immobile particles into the pumped sample (Puls, et al. 1990, Puls and Barcelona, 1996). Passive, non-purge, sampling is an alternative to low flow sampling that avoids mobilization of particles, but also avoids handling and disposal of contaminated waters and the time and costs Britt Page 16

280 associated with pumping and field parameter measurements (ITRC, 2004, 2007). The ISS sampler is suited to take advantage of these benefits for nearly all categories of analytes, including volatiles, gases, metals, and dissolved inorganics. The ISS sampler also, importantly, improves data quality by avoiding exposure of sample to air at the collection point. Indication of the importance of atmospheric exposure and exposure to plastics is evident in the field studies, and shows VOC losses may be chemical-specific, and dependent on Henry s vapor partitioning coefficient or octanol-water partitioning coefficient (Baerg et al. 1992; Gibs et al. 1994; Britt 2006). The ISS approach will likely improve data quality by reducing sample exposure and field errors. It removes reliance on the operator for pump placement, purge times, purge parameter measurement, and bottle fill technique. The ISS avoids data quality problems with tubing sorption loss, inwell mixing effects during purging, and limits the effect of uncontrollable aspects of field sampling such as ambient temperature, humidity or precipitation. Millions of monitoring wells are used around the world to track temporal trends in natural groundwater quality and at contaminated sites; however, the results are influenced by many factors associated with the sampling procedures themselves. These influential factors cause error in individual sample results and diminish the value of such long term trend monitoring. The in situ sealed sampler described in this paper provides opportunity for minimization of decision-making by the operator on these influential factors and limits uncontrollable factors such as partitioning coefficients and weather. The ISS approach imparts the least degree of differential influence of any of these factors from one sampling event to the next through elimination of procedures and sample handling. Interpretation of time-series data can then be focused most reliably on the influences caused by the hydrologic system rather than the sampling process. Acknowledgements (needs completion) The field trials reported on this paper were supported financially by the University Consortium for Field Focused Groundwater Contamination Research, Boeing, NSERC (Beth please spell this out), and the site owners themselves. Clare Stewart and Matthew Whitney assisted with field work at the Guelph site and Maria Gorecka supervised the laboratory analysis at the University of Waterloo. Access to the monitoring wells at the Britt Page 17

281 Guelph site was provided by Guelph Tool, Inc. Laura Zimmerman assisted with field work at the Chatsworth site. Access to the Chatsworth site was provided by Boeing. Data from the private sites in Morgan Hill, CA and Hillside, NJ were provided by ARCADIS. Data from the McClellan study was provided by PARSONS. Data from the Los Angeles site was provided directly by the private site owner. These contributors are acknowledged with sincerest appreciation. References American Society of Testing Materials (ASTM) Standard Practice for Low-Flow Purging and Sampling for Wells and Devices Used for Ground-Water Quality Investigations. ASTM Subcommittee D18.21: Designation D Baerg, D.F., R.C. Starr, J.A. Cherry, and D.A. Smyth Performance Testing of Conventional and Downhole Samplers and Pumps for VOCs in a Laboratory Monitoring Well. Proceedings of the National Groundwater Sampling Symposium, November 1992, Washington D.C. Bailey, R., J. Marchesani, A. C. Marinucci, J. Reynard, P. Sanders Use of Sequential Filtration for Determining Transportable Lead in Ground Water. Ground Water Monitoring and Remediation, v25:3, p Barcelona, M.J. and J.A. Helfrich, 1986, Well Construction and purging effects on ground-water samplers. Environmental Science and Technology, v. 20: Britt, S.L Testing the In-Well Horizontal Laminar Flow Assumption with a Sand- Tank Well Model. Ground Water Monitoring and Remediation, v25:3, p Britt, S.L Differential Recoveries During VOC Sampling: Low Flow, the Snap Sampler, and Remedial Decision-Making. Proceedings of the Fifth International Conference on Remediation of Chlorinated and Recalcitrant Compounds, Monterey, California, May 22-25, Church, P.E Evaluation of a diffusion sampling method for Determining Concentrations of Volatile Organic Compounds in Ground Water, Hanscom Air Force Base, Bedford, Massachusetts. U.S. Geological Survey Water Resources Investigation Report Freeze, R.A. and J.A. Cherry, 1979, Groundwater, Prentice-Hall, Englewood Cliffs, N.J. 604 p. Britt Page 18

282 Frost, R.C., T.F. Bernascone and T. Cairney, A light-weight and cheap depth sampler. Journal of Hydrology, 33: Gibs, J., T.E. Imbrigiotta, J.H. Ficken, J.A. Pankow, and M.E Rosen Effects of Sample Isolation and Handling on the Recovery of Purgeable Organic Compounds. Ground Water Monitoring Review 14, no 2: Gillham, R.W., Syringe devices for ground-water sampling. Ground Water Monitoring Review, 2 (Spring 1982): Interstate Technology and Regulatory Council (ITRC) Technical and Regulatory Guidance for Using Polyethylene Diffusion Bag Samplers to Monitor Volatile Organic Compounds in Groundwater. ITRC Technical/Regulatory Guidelines, Document DSP3. ITRC, Protocol for Use of Five Passive Samplers to Sample for a Variety of Contaminants in GroundWater. Interstate Technology and Regulatory Council, Document DSP5. Nurmi, P.A., and I.T. Kukkonon, A new technique for sampling water and gas from deep drill holes. Canadian Journal of Earth Sciences 23,: Pankow, J.F., Isabelle, L.M., Hewetson, J.P., and Cherry, J.A., A syringe and cartridge method for downhole sampling for trace organic compounds in groundwater. Ground Water 22, no. 3: Pankow, J.F., Isabelle, L.M., Hewetson, J.P. and Cherry, J.A., A tube and cartridge method for down-hole sampling for trace organics in groundwater. Ground Water 23, no.6: Parker, L.V. and T.A. Ranney Sampling Trace-Level Organics with Polymeric Tubing: Part 2. Dynamic Studies. Ground Water Monitoring and Remediation 18, no. 1: Parker, L.V. and N. Mulherin Evaluation of the Snap Sampler for Sampling Groundwater for VOCs and Explosives. US Army Corps of Engineers, ERDC/CRREL TR-07-14, 68p. Parsons Engineering Science Final Technical Report for the Evaluation of Groundwater Diffusion Samplers. Air Force Center for Environmental Excellence. Parsons, 2005, Final Results Report for the Demonstration of No-Purge Groundwater Sampling Devices at Former McClellan Air Force Base, California. US Army Corps of Engineers/Air Force Center for Environmental Excellence/Air Force Real Property Agency, October Britt Page 19

283 Pettijohn, W.A., W.J. Dunlap, R. Cosby, and J.W. Keeley, 1981, Sampling Groundwater for Organic Contaminants, Groundwater, 19, Powell, R.M., and R.W. Puls Passive Sampling of Groundwater Monitoring Wells Without Purging: Multilevel Well Chemistry and Tracer Disappearance. Journal of Contaminant Hydrology 12: Puls, R.W., J.H. Eychaner, and R.M. Powell, 1990, Colloidal-Facilitated Transport of Inorganic Contaminants in Ground Water, Part I, Sampling Considerations, USEPA Environmental Research Brief, EPA/600/M-90/023, 12p. Puls, R.W. and Barcelona, M.J., 1996, Low-Flow (Minimal Drawdown) Ground-Water Sampling Procedures, USEPA Ground Water Issue Paper, EPA/540/S-95/504, 12p. Robin, M.J.L. and R.W. Gillham Field Evaluation of Well Purging Procedures. Ground Water Monitoring Review 7, no. 4: Sherwood Lollar, B., S. K. Frape, and S. M. Weiss, New sampling devices for environmental characterization of groundwater and dissolved gas chemistry (CH 4, N 2, He). Environmental Science and Technology 28, No. 13: Tate, T.K., Variations in the design of depth samplers for use in groundwater studies. Water and Water Engineering, 77:223. United States Environmental Protection Agency (USEPA), RCRA Ground-Water Monitoring Technical Enforcement Guidance Document. USEPA Office of Solid Waste and Emergency Response, EPA/530/SW-86/055, 329p. Vroblesky, D.A User s Guide for Polyethylene-Based Passive Diffusion Bag Samplers to Obtain Volatile Organic Compound Concentrations in Wells, Parts 1 and 2, U.S. Geological Survey Water Resources Investigations Reports and Vroblesky, D.A. and T.R. Campbell Equilibration Times, Stability, and Compound Selectivity of Diffusion Samplers for Collection of Groundwater VOC concentrations. Advanced Environmental Restoration 5, no 1: Vroblesky, D.A. and B. C. Peters Diffusion Sampler Testing at Naval Air Station North Island, San Diego County, California, November 1999 to January U.S. Geological Survey Water Resources Investigation Report Yeskis, D. and B. Zavala Ground Water Sampling Guidelines for Superfund and RCRA Project Managers. U.S. Environmental Protection Agency Report 542-S Britt Page 20

284 Single column figure 40ml VOA Overburden 40ml VOA Siltstone/ sandstone bedrock Not to scale Snap Sampler 125ml Poly; 350ml Poly not shown Figure 1. In situ sealed (ISS) samplers deployed downhole, secured and locked at well head; inset shows loaded ISS/Snap samplers with caps set into open position; full bottles after collection and removal from samplers.

285 Double column figure Add 1-2 drops preservative Pierce Snap Cap with driver Tool F Top off preservative G Screw septa caps to seal sample without opening H I Figure 2. A-E: Illustration of steps to prepare p ISS/Snap samplers for deployment. LOAD-SET-ATTACH TRIGGER-LOWER DOWNHOLE F-I Steps for Snap VOA vial preparation TRIGGER-RETRIEVE-UNLOAD-PREPARE.

286 Set & Deploy Downhole Initial deployment: - Fast, low manpower Double column figure Reload Snap Samplers for redeployment Prepare samples for laboratory submittal; No sample exposure Lock and leave downhole No well purge or wastewater Trigger & Retrieve Uphole Deployment period: Days to months Figure 3. Initial sampler deployment includes loading sampler bottles, setting caps, and lowering downhole. Locked well cap secures samplers and well until sample collection. Sample collection takes place days, weeks or months after initial deployment. Collection includes tripping bottles to close with manual, electric or pneumatic trigger system and retrieving the sample bottles from the well. No purging is required. Sealed sample bottles are not transferred to laboratory bottles, removing a handing step and exposure to atmospheric air. Redeployment with new bottles starts the subsequent sample cycle.

287 Double column figure Site Location Geology Sampling Depth (m) Depth to Water (m) Number of Wells Sample Intervals Analytes Compared to: Notes Chatsworth Fractured Sandstone 24 to 52 6 to VOC, gases Low Flow Electric trigger used Guelph Fractured Dolostone 3 to 8 2 to VOC Low Flow PDB Pull trigger Morgan Hill Silt/sand overburden 5 to 21 3 to VOC perchlorate Volume purge Pull trigger Hillside Silt/sand overburden 10 to 18 9 to VOC, As Volume purge Pull trigger McClellan Silt/sand/ gravel overburden 33 to VOC, anions 1,4-dioxane Low Flow, Vol. purge Multiple comparisons Los Angeles Silt/sand/ gravel overburden 12 to 18 9 to VOC Vol. purge Repeated long term test Table 1. Site information for field deployments.

288 Double column figure Pumped / Fractured Rock Pumped/ Granular Passive/ Diffusion-based Passive ISS/Snap Sample Overburden Bedrock t1 t2 t1 t2 t3 Not to scale t3 Figure 4. Water sampled by the devices is represented by shaded areas. Purge sampling reaches further from the pump with longer pumping duration; diffusion samples equilibrate over several days, thus represent water that has been in contact with the sampler during that time; ISS/Snap samples are (equilibrated) grab samples which collect water in the well at the time of bottle closure. Granular media is depicted as homogeneous for illustration.

289 Double column figure Figure 5. ISS/Snap sampler, diffusion sampler, and low flow comparison at Guelph site; ISS/Snap sampler and low flow purge comparison at Santa Susanna Field Laboratory (SSFL) site. Slight positive offset of trendline (y>1) indicates y-axis comparator is slightly higher on average. Very good correlation coefficients relate tight correspondence among the methods. SSFL shows tighter correspondence (y=1.02) with use of pre-deployed bladder pump rather than peristaltic pump.

290 Single column figure Figure 6. VOC and perchlorate comparative data plots. These illustrate recovery may differ for volatile/sorptive chemicals while a non-volatile/sorptive constituent from the same well(s) show little difference. Volume purge-based method used for purge samples (private site in northern California)

291 Single column figure Sample method Y-slope Med. RPD Low Flow Baseline comparator reference PSMS RPPS Hydrasleeve RCS PDBS ISS/ Snap Sampler Table 2. VOC recovery data comparison among sampling methods in McClellan Study. All methods had greater VOC recovery than low flow; it is used as the reference comparator. Trendline y-slope >1 indicates the comparator method is on average higher than the reference. Positive median RPD indicates comparator is higher than reference.

292 Single column figure Figure 7. Sorption and volatilization among passive and active sampling methods. Data from the McClellan Study of passive groundwater sampling devices conducted at the former McClellan Air Force Base in Sacramento, California. This study showed relatively large VOC recovery differences among chemicals. ISS/Snap sampler always had highest recovery and is used as the baseline comparator. Illustration shows device and chemical-specific recoveries relative to (full) Snap Sampler recovery. There is a strong association with Kow and recovery percent, while vapor partitioning shows weaker recovery association.

293 Figure 8. Arsenic concentration comparison. Unfiltered trendline slope (y<1) indicates x-axis comparator is higher concentration on average than y-axis comparator, while the filtered example shows the opposite. Comparison sample differences suggest purge sample filtration eliminates a high bias, yet introduces a low bias. Single column figure

294 Single column figure Figure 9. Illustration of variation reduction. Example of reduced event-to-event concentration change. Top figure illustrates direction of change is consistent between the methods while the bottom figure shows that the magnitude of event-to-event concentration change is lower with the ISS/Snap method. On average, the RPD from event to event is reduced by 2/3. This includes the actual change in concentration contributed from the aquifer, which implies that the reduction in change attributed to the sampling error may be substantially larger than 2/3.

295 A versatile packer system for high resolution hydraulic testing in fractured rock boreholes Patryk Quinn, Department of Earth and Environmental Sciences, University of Waterloo 200 University Avenue West, Waterloo, Ontario, Canada N2L3G1 Phone/Fax (519) Ext 32933, John A. Cherry Department of Earth and Environmental Sciences, University of Waterloo 200 University Avenue West, Waterloo, Ontario, Canada N2L3G1 Phone/Fax (519) Ext:84516, Beth L. Parker, School of Engineering, University of Guelph 50 Stone Road East, Guelph, Ontario N1G2W1 phone: (519) Ext , Fax: (519) , For Submittal to: Canadian Geotechnical Journal Draft: October 4, 2009

296 Abstract: Packer testing equipment for use in fractured rock boreholes was designed to allow hydraulic tests to be completed by four methods, constant head step tests, slug tests, pumping tests and recovery tests in the same test interval without removing the equipment from the hole while acquiring high precision data for transmissivity (T) calculation. This equipment records pressure above, within, and below the test interval. Flow rates are measured with a series of turbine flow meters enabling a measurement range of 13 ml/min 15 L/min. Flow rates outside this range, as low as 6 ml/min up to 20 L/min, can be measured manually with sight gauges on the injection tanks. Temperature in the test interval and at the ground surface is measured with high resolution RTD sensors to account for density and viscosity variations. The equipment is designed such that each type of test may be conducted repeatedly over a wide range of imposed applied pressures and flow rates. Repetitions of the same type of test reveal useful insights regarding the variability of the test results as the driving force is gradually increased. Conducting all four tests in the same interval provides useful insights concerning transmissivity not otherwise available. This paper will describe each of the components of this equipment indicating the purpose and accuracy associated with each component and performance of this equipment is demonstrated using representative test results from a number of field sites. (235 words) Keywords: Darcian flow, constant head test, fracture dilation, fractured rock, hydraulic conductivity, nonlinear flow, turbulent flow, packer testing, pumping test, recovery test, slug test, transmissivity 1

297 2.1 INTRODUCTION Hydraulic tests (injection or withdrawal) in rock boreholes for measurement of transmissivity (T) are commonly conducted using inflatable packers to isolate specific intervals (e.g. NRC, 1996; Sara, 2005). Such T measurements are used in many types of fractured rock investigations including radionuclide waste isolation, mine site water control, groundwater resource assessments, and contaminated site characterization. Most commonly, two packers are used (straddle packers) to isolate and test the interval between the packers, but single packer tests can also be conducted to isolate and test the bottom portion of the borehole. Four very different categories of hydraulic tests are reported in the fractured rock literature: constant head step tests (e.g. Maini, 1971; Gale, 1975; Doe & Remer, 1980, Mackie, 1982; Novakowski et al., 1997), pumping tests carried out to near steady state (e.g. Rushton et al., 1985; Huntley et al., 1992; Gemand et al., 1997; Andrews et al., 2002), the subsequent recovery test (e.g. Pollard, 1959; Gringarten et al., 1975; Shapiro et al., 1998;), and slug tests (e.g. Schwartz, 1975; Barker & Black, 1983; Shapiro & Hsieh, 1998; Lee, 1999; Svenson et al., 2007; Schweisinger et al, 2009). All of these hydraulic tests are summarized in Figure 1. In nearly all applications of straddle packer tests reported in the literature, the investigators have opted to conduct only one type of hydraulic test. However, because of the complexities inherent in fractured rock and the large differences between the four categories regarding the hydraulic conditions imposed on the formation, new possibilities for acquiring useful additional insights appear when more than one category of test is conducted in selected test intervals. In one study, Schweisinger et al. (2009), conducted rising/falling head slug tests and constant rate pumping tests in the same test interval and report that the falling head slug test produced a larger value for T than the rising head tests and the pumping tests resulted in the 2

298 smallest value for T. They postulate that the T values determined from these tests were sensitive to the changes of effective stress in the fracture causing fracture dilation and contraction. This paper describes versatile straddle packer hydraulic test equipment developed to conveniently conduct all four types of tests with high precision in each test interval using a wide range of head differentials and flow rates into and out of the formation. An impetus for the development of this equipment is the desire to use the T values to calculate hydraulic apertures using the Cubic Law. Results from simulations of contaminant or heat transport in discrete fracture network models (DFN) are strongly sensitive to aperture values (Smith & Schwartz, 1984). Therefore, it is necessary to conduct assessments concerning biases or uncertainties in the T values derived from these different methods and the resultant hydraulic apertures calculated from these values. Figure 1 indicates that a basic assumption inherent in the analysis of all categories of tests for T is Darcian (linear) flow and this assumption is accepted uncritically in the literature providing guidance on test procedures (U.S. Bureau of Reclamation, 1974 & 1977; Sara, 2005; Neilson, 2006). Therefore, the equipment was designed to conduct tests beginning with minimal driving force where Darcian flow is most likely, and then gradually increasing the driving forces causing non-darcian flow to examine the influence of the test conditions on the T values obtained. This equipment development for examination of the influence of flow conditions was prompted by concerns of other researchers of the phenomena of non-linear flow (Maini, 1972; Mackie, 1982; Elsworth & Doe, 1986; Atkinson et al., 1994), and fracture dilation/formation compressibility (Svenson et al., 2007; Schweisinger et al., 2009). 3

299 The equipment was developed in stages, starting with conventional equipment used for constant head injection step tests, and then modifying over time to enable the conduction of the other three tests. During this evolutionary development of the equipment, performance was assessed by applications in boreholes in a 100m thick, fractured dolostone aquifer in Guelph, Ontario, overlain by Quaternary deposits and bounded below by a shale aquitard. Tests were conducted in two distinctly different portions of this heterogeneous aquifer, a relatively competent dolostone with a low permeable rock matrix and flow dominated by the fractures, and a less competent dolostone in which reef mounds had been identified based on core examination (Brunton, 2008) indicating much higher matrix permeability. 2.2 TEST EQUIPMENT AND MEASUREMENT RESOLUTION Development of this packer testing system began with use of an adaptation of the system for constant head step injection tests first presented conceptually by Gale (1982). Figure 2 is schematic of the modified system. Based on descriptions of other packer systems (Maini, 1971; Louis, 1972; Hsieh, 1983; Shapiro, 2001), the equipment was modified to improve flow control and measurement in the constant head tests and to allow for pneumatic slug tests and injection/withdrawal pumping tests while allowing monitoring of the pressure in the open borehole above and below the test interval. This modified test equipment will be described in this section. Downhole Equipment The system was first modified by using 2 inch diameter Solinst well casing (5 foot lengths) extending from the top packer to the ground surface. This creates a temporary 2 inch well in each test interval in which all four tests can be conducted. All of the 2 inch well casing was marked at 4

300 0.5 m intervals relative to the top of the test interval and depths were referenced to either the top of casing when the casing was above the ground surface, and to the ground surface when testing a flush mounted borehole in a vault. Durable sliding head P packers made by RST Instruments are used (7.1 cm deflated, 14.7 cm max confined inflated diameter) to isolate test interval. A high pressure regulator (1500 psi) is used on the nitrogen cylinder used for packer inflation to enable testing at greater depths. (i.e. greater packer pressures are needed as the open borehole water pressure increases with depth). The packers are separated by 1 ¼ diameter perforated steel pipe and the through pipe in the packers is 1 ¼ in diameter. Because the maximum working pressure of the packers is dependant on the water column height and the borehole diameter, a computer program was developed to calculate the maximum safe working pressure of the packers throughout the borehole before testing using a virtual caliper log and the operational curve of the P packer. The uncertainty arising from using sliding head packers arises from the fact that the bottom of the top packer moves upward as the packer expands when it is inflated. This sliding movement allows the use of thicker material for the packer gland making the packers more resilient in rough borehole environments. However, the amount of this movement changes depending on the size of the hole because the packer will slide less in smaller diameter holes than it will in larger holes. This distance was measured while inflating the packer inside a 4 inch pipe (4 cm movement). Therefore the depth measurements are highly accurate when conducting tests in a 4 inch borehole. Most of the data collected as part of this study was obtained from 4 inch diameter cored holes and therefore depth measurements can be considered to be correct to within ± 1 cm. 5

301 Previous researchers (Maini, 1971; Louis, 1972; Shapiro, 2001) identified the need for measuring above and below the test interval in their packer systems, and therefore, three pressure transducers are used to measure pressure in the test interval, the free standing water level in the annulus between the 2 inch casing and the borehole wall above and the pressure in the open hole below the packed off interval. The transducer for the pressure measurement in the test interval is attached with an elbow compression fitting to the riser pipe, measuring the pressure in the riser pipe just above the packed off interval. Measurement below the packed off interval is done through a ¼ inch flexible tubing that is run through the system and is fixed with a bored through compression fitting on the top of the top packer and on the end cap at the bottom of the bottom packer. The transducer connects to the tubing at the same depth as the transducer measuring the interval pressure, above the top packer. Data resolution was very poor measuring the pressure in this fashion because it was inevitable that air would become entrapped in the ¼ inch throughput tube as the equipment is lowered into the borehole. The transducer measuring the pressure in the open borehole above the test interval was fixed in the same location as the other transducers making total head calculations simpler because all transducers are located at the same depth. A new system is being developed that will use underwater plugs to allow each transducer to be situated at the measurement point which should allow all pressure measurement outside of the equipment in the test interval, thereby eliminating the effects of equipment non-linear flow and below the test interval eliminating the need for the throughput tubing resulting in better data resolution. Three types of transducers were used in the data collection in this study including vented Druck PDCR 1830 (0-100 mv output), and vented, current output PMC VL400 series (4-20 ma output) and a set of Schlumberger Mini-Divers (20 m, 50 m, and 100 m full scale) that measure absolute 6

302 pressure. Transducers vented to the atmosphere alleviate the need to correct for atmospheric changes. The mv output transducer had a consistently higher resolution than the current output transducer and was therefore used for test interval measurements for all tests. Both transducers are rated to have an accuracy of 0.1% of full scale (± 20 cm). However, accuracy is only needed for the head profiles constructed at the end of the borehole testing. For the hydraulic tests, resolution is most important because all tests are recording changes in the pressure. At constant pressure throughout the entire range of pressures encountered during testing, the mv output transducer consistently would measure the pressure with a resolution of ± 2 cm and therefore can be considered the pressure resolution of all constant head tests conducted. These transducers were periodically calibrated using a Druck DPI 603 portable pressure calibrator. Calibration curves consistently had a very good regression factor (R 2 =1) when calibrated over the transducer full range. Recovery tests and slug tests used Mini Divers for pressure measurements because these transducers have a slightly higher resolution (± 1 cm). When the Divers were used to measure pressure a barologger was hung in the trailer to correct for barometric fluctuations. However, this study indicates that because it is unknown how connected the test interval is to the atmosphere, vented transducer data may be misleading. The revised system will therefore use gauge pressure transducers and the data will be selectively corrected for atmospheric pressure changes based upon evidence of unconfined conditions in the test interval. Aboveground Equipment A trailer is outfitted with a series of tanks of different diameter with sight gauges used to measure flow rates by timing the rate of water level drop and knowing the tank inside diameter. Tank diameters range from 2.5 to 40 cm with the smaller tanks used for less permeable test 7

303 intervals. All tanks are connected through a manifold system to a nitrogen cylinder used to pressurize the void space above the water in the tanks. This is the driving force for all constant head injection tests. A pump (March TE5KMD) with a maximum flow rate of 60 L/min (6 m Head) is used for constant rate injection/recovery tests. In addition to the manual flow measurement using the sight gauges, flow is also measured through a series of flow meters including a McMillan G111 ( ml/min, 0-5 mv output), an Omega FTB 601B (0.1-2 L/min), and an Omega FTB 603B ( L/min). The redundancy of flow measurement with the flow meters and the sight gauges increases reliability in the flow measurement. Both Omega flow meters have a square wave pulse output. The McMillan flow meter has a rated accuracy of 0.5% of full scale (± 0.5 ml/min). The Omega flow meters are rated as 1% of the reading with a repeatability of 0.1% of the reading. Therefore, the middle range flow meter has an accuracy of ± 1 ml/min to ± 20 ml/min, and the high range flow meter has an accuracy of ± 5 ml/min to ± 150 ml/min. However, routine calibration of all flow resulted in very good linear regressions (R 2 = 1) for the entire flow range and the values measured by volume output were very similar to values measured electronically (± 3%). Because transmissivity is slightly dependent on water viscosity and density, temperature was measured in the test interval and in the trailer before injection with high resolution RTD sensors (± 0.03 C) obtained from Waage Electric (NJ). The test interval temperature was measured inside the bottom of the 2 inch pipe, (Figure 2) and the surface measurement was made at the main manifold for the injection tanks outlet. Test interval temperature can be affected by the injection process when the injection water is much warmer than the ground water temperature such as on hot days when the water tank is in full sunlight or when the injection water is colder 8

304 on cloudy cold days in winter. Data from all of the electronic measurement devices was collected using a Campbell Scientific data logger (CR 10X). The chemistry of the injected water can influence carbonate rock flow systems as reported by Polak et al. (2004) when DI water was used in the injection tests causing the etching of solution channels over short periods of time. The water used for the injection tests in this study was obtained from an on-site hydrant drawing from the same aquifer that was being tested, and therefore the injection water is assumed to have the same chemistry as the borehole water, with minimal alteration to the rock system. 2.3 HYDRAULIC TESTS AND PROCEDURE Each different category of test requires use of specific features of the test equipment. Both the equipment design and test procedures were developed to achieve consistency in data collection. For constant head tests, injection is done through a mini-packer that is lowered to below the water table in the 2 inch pipe. Once the mini-packer is inflated, the interval is isolated from the atmosphere and becomes a closed system in which the induced pressure rapidly achieves steady state. The injection lines include 1/8 inch OD and 1/4 inch OD flexible tubing for low flow rates, 3/8 inch OD and 1/2 inch OD for middle flow rates, and 5/8 inch OD tubing for very high flow rates. For the lowest flow rate, the 1/8 and ¼ inch lines along with the 1/8 inch mini-packer inflation line were pulled through a section of 5/8 inch OD flexible tubing for ease of manipulation and a series of valves are used to control flows from the injection tanks. In low permeable test intervals the mini-packer inflation creates a pressure pulse that can be analyzed on its own as a slug test. Another advantage of this injection system is that the shorter, variable diameter injection lines allow for accurate flow control at all injection rates. Test injection times 9

305 varied between 5 and 15 minutes for all constant head steps, with the test ending after the flow and pressure record clearly showed that neither was changing with time. A 2 inch submersible Grundfos pump (RediFlow-2) is used for withdrawal recovery tests and a ¼ inch injection line is lowered into the 2 inch pipe for injection/recovery tests. A check valve fitting was required on the pump for rising head recovery tests to prevent water in the outflow line from falling back into the test interval when the pump was turned off. Flexible tubing is connected to this fitting to allow for the water in the line to be purges with compressed nitrogen before pump removal to lighten the hose weight and minimize leakage when transporting the pump. The outflow from withdrawal tests was routed through the largest flow meter ( L/min) for accurate flow measurements. Higher flow rates (15-20 L/min) must be measured manually. This equipment is also designed to conduct pneumatic slug tests using a special fitting that locks on top of the 2 inch casing. Pressure from a nitrogen tank is used to push the water table down and a 2 inch valve on the fitting is used to release the pressure and begin the test. Two pressure gauges are used to monitor the pressure in the 2 inch pipe prior to beginning the test, a 0-5 psi gauge and a 0-30 psi gauge, which can be easily interchanged. This setup makes it easy to conduct multiple slug tests over a large range of initial head displacements (5 cm to 20 m). 2.4 EQUIPMENT PERFORMANCE AND DISCUSSION The equipment test range is governed by the range of flow and pressure measurements and because the sight gauge on the one inch diameter tank allows measurement of flow rates as low as 5 ml/min, extremely low T values can be measured with this equipment. It should also be possible to measure higher T values if non-linear flow and fracture dilation can be accounted for 10

306 because it is difficult to achieve any change in head in extremely permeable zones without causing nonlinear flow or fracture dilation near the test well as discussed further below. In all constant head step testing done using this equipment to date the range of T measured was in the range of 1x10-8 to 5x10-4 m 2 /s (K=7x10-9 to 3x10-4 m/s) when using 30 m as the assumed radius of influence. Response to Isolating the Test Interval from the Rest of the Borehole All boreholes act as short circuits between the fractures it intersects so that the blended head in the open hole is the resultant of the individual head in each fracture zone. The variation of head with depth causes vertical flow fields in the hole from the locations of higher head to lower head and interrupting this flow field causes changes that can be observed during packer inflation. Figure 3 shows the response to isolating the packed off interval in which a downward gradient is identified. This is the result of plugging the flow system at the depth of the top packer stopping the downward flow causing the test interval pressure to decrease, while the pressure above the top packer increases. Based on subsequent multilevel head profiles this hole has a large downward gradient. Response to Isolating the Test Interval from the Atmosphere The mini-packer is lowered into the 2 inch riser pipe to the same depth below top of casing (TOC) in all intervals prior to any packer inflation. Once the main packers are inflated and the isolated test interval has reached equilibrium, the mini-packer is inflated to isolate the test interval from the atmosphere. The test interval response to the mini-packer inflation may supply additional information as illustrated in Figure 4. In Figure 4A when the main packers hit the borehole wall the pressure below the test interval drops while the pressure in the test interval 11

307 increases without any change in the water level in the open borehole above. The decrease in pressure below the test interval is indicative of a downward gradient as described above, but the lack of response above suggests that there is a very permeable zone above the test interval capable of dissipating the additional flow without any rise in water level in the annulus. The increase in interval pressure can be interpreted as a lower permeable zone under a confining pressure. Upon mini-packer inflation the interval pressure shows an initial slug followed by a large drop, approximately 1 m. When the mini-packer is deflated the pressure returns to the value before inflation. This behavior can be explained by understanding that the transducer is measuring a pressurized system when it is shut in and when the 2 inch vent is opened to the atmosphere the water level rises in the amount of the overpressure. This behavior is indicative of a confined unit with the pressure increase equal to the confining pressure. This zone was determined to have low permeability based on the constant head tests (T = 4.2e-7 m 2 /s) which supports this interpretation. On the other extreme, in very highly permeable intervals there may or may not even be an initial slug, but the pressure after inflation does not change from pre-inflation values as seen in the single packer test conducted at the bottom of this hole illustrated in Figure 4B. This behavior indicates that the packed off interval is well connected to the atmosphere, probably through the fracture network. The entire hole was tested in this fashion and it was observed that there was a gradation from low permeable zones (confined) towards the zones of high permeability (unconfined) with progressively less confinement. Table 1 summarizes the data for this borehole. The data indicates that this fractured dolostone aquifer acts as an unconfined aquifer at three depths (14m 28m and 40m), and the level of confinement is maximum in between these depths. This is important information when determining the type of data analysis. 12

308 Identification of Connection to the Open Borehole When water is injected into the test interval, there is a possibility for the injected water to enter the open hole above or below the test interval through either leakage between the packers and the rough borehole wall or short circuiting through the formation. This phenomenon is identified by a pressure change above/below the test interval during a hydraulic test. However, sensitivity to pressure changes is greater below the test interval than it is in the open hole above because the open hole below is a closed system which will instantly react to pressure changes. Above the test interval, the water level in the hole must rise to reflect the increased pressure. However, in fractured sedimentary rock, it is conceivable that even though the pressure is increased just above the test interval the water level in the open hole does not change because the pressure is relieved by highly conductive zone(s) above the test interval but below the water level in the open borehole. For this reason the pressure in the open hole was measured just above the top packer to ensure the detection of any pressure changes caused by leakage and/or short circuiting. One type of connection to the open hole involves leakage between the packers and the borehole wall caused by an imperfect seal (Figure 5A). In portions of the borehole the walls are not smooth because of the presence of extensive vugs or fossils, or in highly fractured zones, where pieces of the rock have broken out and fallen to the bottom of the hole. When the borehole walls are not completely smooth, there is a propensity for leakage caused by incomplete packer seal along the borehole wall. There is no delay in the response above or below the test interval when leakage occurs between the packers and the borehole wall. Formation short circuiting is defined as flow around the packers through the formation to the open borehole above or below the test interval (Figure 5B). In some zones the fracture network is 13

309 so dense that part of the network is connected to the open borehole above or below the test interval. In other zones when the rock matrix permeability is large enough, the connection to the open hole may be through the rock matrix. In both of these instances, the injected water travels through the formation to the open borehole causing a delayed pressure response above or below the test interval. Constant Head Step Tests Over 150 constant head injection step tests were conducted at the Guelph field site in four different boreholes and Figure 6 illustrates typical data collected. Pressure measurement above and below the test interval identifies any leakage or short circuiting occurring during the test, and the data indicates that 10 minutes is an adequate time for the system to achieve equilibrium. Most often it is assumed that flow during the test is Darcian and therefore, a plot of flow (Q) vs applied pressure (dp) should result in a straight line passing through zero. However, non-linear flow through fractures has been identified in both lab experiments (Sharp, 1970; Louis, 1972; Elsworth, 1984; Zimmerman et al., 2004) and field studies (Maini, 1971; Gale 1975; Mackie, 1980; Atkinson et al., 1994) at relatively low flow rates. The straight line Q vs dp relationship can change because of fracture dilation or convective acceleration as illustrated in Figure 7. In this study most of the data collected (> 95%) show deviation from linearity at low flow rates consistent with convective acceleration, and deviation from linearity caused by dilating fractures was seen much less frequently. Slug Tests Over 200 rising head and 30 falling head slug tests were conducted in five boreholes in the same dolostone aquifer. Figure 8 shows representative raw slug test data collected with this system. In 14

310 this case, the lack of a response above and below the test interval indicates no leakage or short circuiting. The slug response curves all show a pressure spike prior to the slug recovery. This is caused by the pressurization of the 2 inch pipe that pushes the water level down replacing the water column with pressurized nitrogen and can be considered a falling head slug test. When the 2 inch valve is opened to atmosphere, the water column instantly drops and begins to recover to the static water level (rising head slug test). Previous studies have observed fracture aperture changes in fractured igneous and sedimentary rock during slug tests (Rutqvist et al., 1992; Svenson et al., 2007; Schweisinger et al., 2009). They report that the fracture dilates initially during the falling head slug test and gradually returns to the original geometry as the pressure is relieved. Because of the short term stresses in slug tests, changes in fracture geometry may be more readily apparent because the dilation of fractures appears to lag behind the applied pressure. Non-linear flow has been identified during slug tests in unconsolidated deposits (McElwee & Zenner, 1998), but no studies have reported evidence of non-linear flow during slug tests in fractured rock. Based on the results of this study, fracture dilation effects cause the falling head T to be greater than the rising head T, which is consistent with these previous studies. In addition, nonlinear flow causes the calculated T to decrease with increasing initial displacement. Pumping and Recovery Tests Over 50 pumping/recovery tests were conducted at different flow rates in five boreholes in the same dolostone aquifer and typical raw data are shown in Figure 9. The withdrawal pumping and recovery tests conducted in the rural holes commonly indicate a connection to the open hole above the test interval and the connection appears to be greater as the pumping rate increases. 15

311 This violates a radial flow model for the test and illustrates the effect of a long term displacement on the large scale fracture network and rock matrix. Various double porosity models have been developed for pumping/recovery tests in fractured rock (Pollard, 1959; Warren & Root, 1962; Hazemi, 1969; Boulton & Stretslova, 1977). However, most analysis rely strictly on recovery data because at the typical high flow rates used in entire borehole testing causes the pumping portion of the data to be of low quality (e.g. variable pumping rate, well bore storage effects, etc.). Because of the low pumping rates used in this study, both the pumping portion and the recovery portion of the test resulted in high quality data allowing a comparison of these two tests. Most of the tests (both pumping and recovery) resulted in a s-curve on a semi log plot (Figure 10) that is consistent with the above double porosity models reflecting the permeability of the large fractures near the test interval at early time, followed by the permeability of the entire system at late times. 2.5 CONCLUSIONS AND IMPLICATIONS The packer testing equipment has been shown to achieve high resolution measurements over a broad range of test conditions with operational ease. Monitoring pressure above, below, and within the test interval provides improved understanding of the test conditions prior to and during each test including the degree of connection to the open hole and the atmosphere for each test interval and recording the pressures as the packers are inflated can give insight into the flow environment in the open hole. Application of the equipment in six boreholes at the Guelph field sites provided results showing the value of conducting different types of tests over a wide range of test conditions. When multiple step constant head tests are conducted in the same test interval at increasing injection flow rates, it is relatively easy to identify non-linear flow. Slug tests show 16

312 non-linearity at higher displacements and evidence of fracture dilation/contraction, while pumping and recovery tests give insight into the surrounding fracture network and/or matrix. 17

313 ACKNOWLEDGEMENTS Bob Ingleton, UW, Paul Johnson, UW and Jay Sitts of Hydrite Chemical Company, Cottage Grove, WI were instrumental in aiding in the development of the packer testing equipment. Campbell Scientific personnel were very helpful aiding in programming the datalogger. Pumps used in the packer testing system were provided by March Pumps, Glenview, Illinois, and temperature RTD sensors were provided by Waage Electric, NJ. Funding for this investigation was provided by NSERC, the Guelph Tool Company (Guelph, ON), the Hydrite Chemical Company (Cottage Grove, WI) and the City of Guelph. 18

314 REFERENCES Atkinson,L.C., J.E.Gale, and C.R.Dudgeon "New insight into the step-drawdown test in fractured-rock aquifers." Applied Hydrogeology. 1:9-18. Boulton, N. S. & Streltsova, T. D. 1977, "Unsteady flow to a pumped well in a fissured waterbearing formation", Journal of Hydrology, vol. 35, no. 3-4, pp Brunton F.R Preliminary Revisions to the Early Silurian Stratigraphy of Niagara Escarpment: Integration of Sequence Stratigraphy, Sedimentology and Hydrogeology to Delineate Hydrogeologic Units Ontario Geological Survey Open File Report Choi, H., Nguyen, T.-B., & Lee, C "Slug test analysis to evaluate permeability of compressible materials", Ground Water, vol. 46, no. 4, pp Doe, T. W., Remer, J., Schwarz, W. J "Analysis of constant-head well tests in nonporous fractured rock", Berkeley, CA, United States, March 26-28, Elsworth, D Laminar and Turbulent Flow in Rock Fissures and Fissure Networks, PhD, University of California Berkeley. Elsworth,D. and T.W.Doe "Application of Non-linear Flow Laws in Determining Rock Fissure Geometry From Single Borehole Pumping Tests." International Journal Rock Mechanics, Mineral Science & Geomechanics. 23: Gale, J. E A Numerical Field and Laboratory Study of Flow in Rocks with Deformable Fractures, PhD, University of California Berkeley. Gale,J.E "Assessing the Permeability Characteristics of Fractured Rock." Geological Society of America. Special Paper 189: Haimson, B. C. & Doe, T. W State of stress, permeability, and fractures in the Precambrian granite of northern Illinois, Journal of Geophysical Research, vol. 88, no. B, pp Hollett, K. J., Wilbourn, S. L., & Latkovich, V. J "Proceedings of a U.S. Geological Survey workshop on the Application and needs of submersible pressure sensors, Denver, Colorado, June 7-10, 1994", Denver, CO, United States, June 7-10, Hsieh, P. A., Neuman, S. P., & Simpson, E. S "Pressure testing of fractured rocks; a methodology employing three-dimensional cross-hole tests; topical report", NUREG/CR (United States.Nuclear Regulatory Commission), vol Lapcevic, P. A Results of Borehole Packer Tests at the Ville Mercier Groundwater Treatment Site, National Water Research Institute, Burlington, Ontario. 19

315 Louis C "Rock Hydraulics," in Rock Mechanics, notes and courses, L. Mueller, ed., Springer Verlag. Mackie, C. D "Multi-rate testing in fractured formations", Australian Water Resources Council Conference Series, vol. 5, pp Maini, Y.N., In-situ hydraulic parameters in jointed rock their measurement and interpretation, PhD Dissertation, Imperial College, London. Maini, Y. N., Noorishad, J., & Sharp, J. C "Theoretical and Field Considerations on the Determination of In-Situ Hydraulic Parameters in Fractured Rock", International Society for Rock Mechanics: Percolation through Fissured Rock, Stuttgart, Germany. McElwee, C. D. & Zenner, M. A "A nonlinear model for analysis of slug-test data", Water Resources Research, vol. 34, no. 1, pp McLane, G. A., Harrity, D. A., & Thomsen, K. O "Slug testing in highly permeable aquifers using a pneumatic method", Washington, DC, United States, Nov Molson, J. W., Pehme, P. E., Cherry, J. A., & Parker, B. L "Numerical Analysis of Heat Transport Within Fractured Sedimentary Rock: Implications for Temperature Probes", NGWA/U.S. EPA Fractured Rock Conference: State of the Science and Measuring Success in Remediation, Portland, Maine, pp National Research Council (NRC) "Hydraulic and Tracer Testing of Fractured Rocks." Rock fractures and fluid flow: Contemporary understanding and applications. National Academy of Science. Washington, D.C Nielsen, David M., Practical handbook of environmental site characterization and ground water monitoring, CRC/Taylor & Francis, Boca Raton, FL. Novakowski, K. S "Comparison of Fracture Aperture Widths Determined from Hydraulic Measurements and Tracer Experiments", Proceedings - Canadian/American Conference on Hydrogeology, vol. 4, pp Novakowski, K.S., Research into the processes of flow and solute transport in fractured rock; experiences in equipment development by staff from Environment Canada, Openfile report [ ], pg: Patchett, R. G "Pneumatic well insert; performing pneumatic rising head tests in wells with screens straddling the water table", Outdoor Action Conference, Las Vegas, NV, United States, May 25-27,1993. Pearson,R. and M.S.Money "Improvements in the Lugeon or packer permeability test." Quarterly Journal of Engineering Geology. 10:

316 Polak, A., Elsworth, D., Liu, E., & Grader, A. S "Spontaneous switching of permeability changes in a limestone fracture with net dissolution", Water Resources Research, vol. 40, no. 3. Pollard, P Evaluation of Acid Treatments from Pressure Build up Analysis. Petroleum Transactions American Institute of Mining, Metallurgical, and Petroleum Engineers 216, Prosser, D. W "A method of performing response tests on highly permeable aquifers", Ground Water, vol. 19, no. 6, pp Rutqvist, J., Noorishad, J., Stephansson, O., & Tsang, C. F "Theoretical and field studies of coupled hydromechanical behaviour of fractured rocks; 2, Field experiment and modelling", International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, vol. 29, no. 4, pp Sara, M. N "Fractured-Rock Assessments," in Site Assessment and Remediation Handbook, 2nd edn, pp Schincariol, R. A., Markle, J. M., & Molson, J. W "Heat transport and thermal pollution in porous and fractured aquifers", Abstracts with Programs - Geological Society of America, vol. 39, no. 6, p Schweisinger, T., Svenson, E. J., & Murdoch, L. C "Introduction to Hydromechanical Well Tests in Fractured Rock Aquifers", Ground Water, vol. 47, no. 1, pp Shapiro, A. M., Hsieh, P. A., & Winter, T. C The Mirror Lake Fractured-Rock Research Site--A Multidisciplinary Research Effort in Characterizing Ground-Water Flow and Chemical Transport in Fractured Rock, USGS Fact Sheet Shapiro, A. M. & Hsieh, P. A "How Good are Estimates of Transmissivity from Slug Tests in Fractured Rock?", Ground Water, vol. 36, no. 1, pp Shapiro, A. M "Fractured-rock aquifers; understanding an increasingly important source of water", Fact Sheet - U.S.Geological Survey. Shapiro, A. M "Characterizing ground-water chemistry and hydraulic properties of fractured-rock aquifers using the multifunction Bedrock-Aquifer Transportable Testing Tool (BAT (super 3) )", Fact Sheet - U.S.Geological Survey. Shapiro, A. M "Characterizing hydraulic properties and ground-water chemistry in fractured-rock aquifers; a user's manual for the multifunction Bedrock-Aquifer Transportable Testing Tool (BAT (super 3) )", Open-File Report - U.S.Geological Survey. Sharp, J.C., Fluid flow through fissured media, PhD Dissertation, Imperial College, London. 21

317 Smith, L., Schwartz, F.W An Analysis of the Influence of Fracture Geometry on Mass Transport in Fractured Media, Water Resources Research, 20(9), Sudicky, E. A. & McLaren, R. G "The Laplace transform Galerkin technique for largescale simulation of mass transport in discretely fractured porous formations", Water Resources Research, vol. 28, no. 2, pp Sisavath, S "A Simple Model for Deviations from the Cubic Law for a Fracture Undergoing Dilation or Closure", Pure and Applied Geophysics, vol. 160, pp Svenson, E., Schweisinger, T., & Murdoch, L. C "Analysis of the hydromechanical behavior of a flat-lying fracture during a slug test", Journal of Hydrology, vol. 126, p. 45. U.S.Bureau of Reclamation "Field Permeability tests in boreholes." Earth Manual US Dept of the Interior Bureau of Reclamation "Permeability Tests in Individual Drill Holes and Wells," in Ground Water Manual,, pp USGS. Characterizing Ground-Water Chemistry and Hydraulic Properties of Fractured-Rock Aquifers Using the Multifunction Bedrock-Aquifer Transportable Testing Tool (BAT). FS , USGS. van Dyke, N. V. R., Rhodes, J. A., Richardson, D. W., & McTigue, W. H "Evaluating confined aquifer properties using the pneumatic displacement method and the repeated pressure pulse technique", Outdoor Action Conference, Las Vegas, NV, United States, May 25-27,1993. Warren, J. E. & Root, P. J "The Behavior of Naturally Fractured Reservoirs". Fall Meeting of the Society of Petroleum Engineers in Los Angeles on Oct.7-10,1962. Witherspoon, P. A., Wang, J. S. Y., Iwai, K., & Gale, J. E "Validity of Cubic Law for Fluid Flow in a Deformable Rock Fractures", Water Resources Research, vol. 16, no. 6, pp Zimmerman,R.W., A.AL-Yaarubi, C.C.Pain, and C.A.Grattoni "Non-Linear Regimes of Fluid Flow in Rock Fractures." International Journal Rock Mechanics, Mineral Science & Geomechanics. 41:

318 Tables Table 1 Confining pressures based upon mini-packer inflation. The highest permeable zones are highlighted. Year Well Zone Top Depth (mbtoc) Bot Depth (mbtoc) Confining Pressure (m) 2008 MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW MW

319 Figures Figure 1 Four types of hydraulic tests 24

320 RTD Sensor at bottom of 2 inch riser pipe Packer Trailer Injection Equipment 3 Mini- Packer Flow Meters Valves for greater flow control Pump for filling tanks and injecting 1/4 tubing runs through system to measure pressure below the bottom packer Constant Head InjectionTanks 2 Test (with sight gauges) Compressed Interval Nitrogen 1 Transducers measure pressure at three locations, 1) below the test interval, 2) in the test interval, 3) above the test interval Pressure Release Valve Slug Test Equipment Pressure Pumping Test Equipment Grundfos 2 pump Check Valve Outflow Figure 2 Schematic of Packer Testing System 25

321 Blended Head Above 6 Interval Pressure (m) Main Packers form seal at the borehole wall Test Interval Pressure in Open Hole (m) Time (minutes) Figure 3 Equilibrium pressures in the test interval and above the test interval in the open hole and evidence of a downward gradient 0 26

322 A High Level of Confinement) mini-packer inflates Interval Below Above vented pressure Pressure (m) in-situ pressure mini-packer deflates 30 Main Packers hit wall Time (min) B No Confinement Packer hits wall Pressure (m) mini-packer inflates Above Interval mini-packer deflates Time (sec) Figure 4 (A) Once the zone is isolated the pressure below the test interval drops indicating a downward gradient. The pressure in the test interval rises due to the presence of confining pressure and when the 2 inch conduit to the surface is opened, the actual confining pressure is revealed. (B) The response of seating the packers in high permeable test intervals does not exhibit this phenomenon. 27

323 A 28 Pressure (m) Above Interval Below Packers Hit Wall Time (sec) Figure 5 (A) Leakage occurs when injected water reaches the open hole above or below the test interval because of an imperfect seal between the packer and the borehole wall. Packer leakage is identified by an immediate pressure response above or below the test interval and increasing the packer pressure does not lessen the leakage. 28

324 B Interval Pressure (m) Test Interval Above Open Hole Pressure (m) Time (min) Figure 5 (B) Short circuiting occurs when injected water reaches the open hole above or below the test interval through the formation and this phenomenon is indicated by a delayed response above or below the test interval

325 A Above Interval 0.05 Pressure (m) Flow Below Q (L/min) Time (min) Figure 6 Example of Constant Head injection test raw data. 30

326 A Pressure Differential (dp) Darcian Flow Non-Darcian flow due to convective acceleration Fracture dilation or hydrofracing 0 0 Flow (Q) B dp (m) Q (L/min) Figure 7 (A) Possible deviations from linearity. (B) Example of deviation due to convective acceleration. 31

327 Above 22 Pressure (m) Below 20.5 Test Interval Tim e (m in) Falling Head Test Rising Head Test Pressure (m) Above Below Time (min) 20 Figure 8 Typical Slug Data showing the falling head and rising head test pairs. 32

328 A Injection and Recovery Pressure (m) Above Below Test Interval Flow Q (L/min) Tim e (m in) B Withdrawal and Recovery 13 Above Pressure (m) Test Interval Flow (L/min) Flow Time (min) 0 Figure 9 (A) Injection data from the City Site (minimal connection to the open hole) and (B) Withdrawal data from the Rural Site (high connection to the open hole). 33

329 (A) Injection Pumping Test s (m) log t (B) Recovery Test s' (m) log t/t' Figure 10 Semi log plots of (A) Injection pumping test and (B) Recovery test conducted in the same test interval. 34

330 Other possible figures Temperature profile Elevation (masl) Temperature (Deg C) Head profile Interval vented Interval closed 325 Elevation (masl) Total Head (m) The red points are when the vented head equals the closed head. 35

331 340 Elevation of center of test interval (masl) vented-open hole (seat main packers) vented-closed (seat minipacker) Pressure changes inside test interval (m) Elevation of center of bottom of hole (masl) Pressure difference below test interval after setting packers (m) 36

332 20 ELEMENTS OF THE SANTA SUSANA FIELD LABORATORY SITE CONCEPTUAL MODEL OF CONTAMINANT TRANSPORT SITE CONCEPTUAL MODEL ELEMENT DOCUMENT 1-11 Development of a Storage System for Data from the Discrete Fracture Network (DFN) Approach DRAFT Prepared for: THE BOEING COMPANY THE NATIONAL AERONAUTICS AND SPACE ADMINISTRATION UNITED STATES DEPARTMENT OF ENERGY Prepared by: Jonathan Kennel 1, Jessica Meyer 2, Beth L. Parker 1, and John A. Cherry 1 1 School of Engineering, University of Guelph, Guelph, Ontario, Canada 2 School of Environmental Sciences, University of Guelph, Guelph, Ontario, Canada Date: December 11, 2009

333 Development of a Storage System for Data from the Discrete Fracture Network (DFN) Approach SCM Document 1-11 DRAFT December 11, 2009 ABSTRACT Application of the DFN method results in production of large amounts of diverse types of high resolution data from each drillhole, both from measurements conducted in the hole and on the rock core. From the drill hole, the measurements typically include various geophysical logs, various temperature logs, flow metering, FLUTe hydraulic conductivity profiling, and packer testing, each done with a different resolution and different errors in depth measurement. The geological descriptions of the rock core from continuously cored holes include many types of information, including geology and identifications of various types of fractures. For many holes, multilevel monitoring systems such as Westbay or FLUTe systems with exceptionally large numbers of ports are installed resulting in several more types of data, including hydrochemistry, hydraulic head and hydraulic test data. The DFN approach is based on comprehensive interpretation of many complementary data sets and, for this, it is necessary to efficiently and effectively manipulate and display the data in many combinations to discern relations between data. Early on in the development of the DFN approach, experience showed that existing data storage systems were very inadequate. This inadequacy derives from the fact that the existing commercial systems for data storage in the groundwater industry were not designed to accommodate such exceptionally large, constantly evolving, and most importantly, diverse data sets produced from the DFN method. Detailed studies of fractured rock sites are rare, and those that have been done use mostly conventional field methods that produce many orders-of-magnitude less data per hole than the DFN approach. Two members of the University of Guelph team (Jonathan Kennel and Jessica Meyer) began in 2007 to develop a relational data storage system referred to as the DFN data storage system, (DSS-UoG) aimed at effective and efficient storage and manipulation of DFN data from the four contaminated sedimentary rock sites that the University of Guelph is investigating intensely, including the SSFL. The goal of the development of the system is to end up with a relational database or set of databases that efficiently and effectively store all of the types of data previously obtained from the fractured rock study sites. The system is set up to accommodate new types of data anticipated to arise in the future as the DFN method continues to improve and evolve. Furthermore, the systems should facilitate and insure the integrity of the data during the data entry process. For example, data entry forms have been developed that identify and preventas much as possible, inconsistencies and errors in data entry. 1

334 Development of a Storage System for Data from the Discrete Fracture Network (DFN) Approach SCM Document 1-11 DRAFT December 11, 2009 This DSS-UoG comprises a set of databases: one to store data collected from the rock core, one to store hydraulic head and hydraulic conductivity data, and another to store analytical results from the rock core and groundwater samples. The rock core database requires geologic information obtained by visual inspection of cores (i.e. geologic logs) to be represented in digital form (i.e. as data rather than written description). Therefore, it was necessary to design a field logging sheet that facilitated the entry of the geological visual observations into the database. This geologic logging sheet is then used by the field geologist producing the geologic core logs. For this DSS-UoG to serve the ultimate intended purpose for data display/interpretation/modeling there must be suitable software for the next stage up. The UoG team has experimented with various types of software for display of data from the DFN-DSS, but all have major limitations. For example, Viewlog allows efficient importation of data but does not allow plotting of acoustic televiewer data. On the other hand, WellCad plots acoustic televiewer data but is inefficient for importation of many other types of data. The ultimate goal is to be able to display all forms of data to detect/display relationships and statistically analyze the data sets. Existing software is poorly suited for this goal because they were developed with much more limited categories of data in mind. No software systems were designed with detailed rock core chemical data or data from comprehensive multilevel systems (i.e. Westbay) in mind. This DSS-UoG is now being used as the principal data storage system for the four intensive study sites that comprise the field portion of the Parker NSERC Industrial Chair at the University of Guelph (SSFL CA, Hydrite WI, Guelph and Cambridge ON). However, improvements are continually being made as experience is gained with use of the system including potential interfacing with external software such as WellCad. It is estimated that the development of the preliminary DSS-UoG is 60-70% complete, however is already serving into present form as a very useful tool in the SSFL bedrock groundwater study. For example, the DSS-UoG is essential to relating the data collected at the drill rig to the data collected at the crushing station during a rock core VOC sampling event. It also facilitates the calculations involved in converting rock core data in mass concentration units to equivalent pore water concentrations. The DSS- UoG also helps to ensure that all data collected are stored in the same format which makes interpretations of the data more efficient and reliable. 2

335 SITE CONCEPTUAL MODEL ELEMENT 2: The rock matrix porosity provided by interconnected pores is large and the bulk fracture porosity is extremely small. Overview The rock matrix porosity and permeability for the Chatsworth Formation has been rigorously determined by laboratory measurements using several methods including both conventional and advanced microscopic imaging methods. This SCM element includes three documents pertaining to the laboratory measurements. One (2-1) has very recently been submitted to a refereed scientific journal. The other two are undergoing major revisions. Values of porosity and permeability were and are also determined from numerous geophysical logs (e.g., FMI and ACTV calibrated to the laboratory measurements). Although these geophysical assessments are not yet completed, sufficient analyses have been completed to determine the rock matrix porosity and permeability across the SSFL. At the SSFL the rock matrix total porosity and the effective porosity relevant to diffusion of contaminants into the rock matrix are nearly identical. The relative uniformity of the rock matrix porosity across the SSFL is consistent with the fact that all strata in the Chatsworth Formation were deposited by turbidity currents. All lines of evidence indicate that the rock matrix porosity is relatively large nearly everywhere. This supports the major conclusion of the SCM that the rock matrix is strongly conducive to diffusion driven chemical mass transfer between the fractures and the rock matrix, providing a large storage reservoir for all types of contaminants.

336 Statistical Synthesis of Imaging and Porosimetry Data for the Characterization of Microstructure and Transport Properties of Sandstones E.S. Amirtharaj, M.A. Ioannidis * Department of Chemical Engineering, University of Waterloo, Ontario, Canada, B. Parker Department of Earth Sciences, University of Waterloo, Ontario, Canada, and C.D. Tsakiroglou Foundation for Research and Technology Hellas, Institute of Chemical Engineering and High Temperature Chemical Processes, Patras, Greece Submitted for publication in Transport in Porous Media * Corresponding author. Tel.: ext ; fax: E- mail address: mioannid@cape.uwaterloo.ca (M.A. Ioannidis). Now at the School of Engineering, University of Guelph, Ontario Canada

337 Abstract The microstructure of a suite of sandstone samples is quantitatively analyzed using a method which combines information from thin section micrographs of the pore space with mercury injection porosimetry in a statistical framework. This method enables the determination of a continuous distribution of pore sizes ranging from few nm to several hundred m. The data obtained unify fractal and Euclidean aspects of the void space geometry, yield estimates of the pore-to-throat aspect ratio and challenge the ability of commonly used network models to describe fluid percolation in multiscale porous media. Application of critical path analysis to the prediction of flow permeability and electrical conductivity of sandstone core samples using the new information produces results comparable to those obtained by the classical approach a fact attributed to the presence of macroscopic heterogeneity at the scale of several millimeters. Keywords: fractal, percolation, porous media, imaging, permeability, conductivity, correlation, scattering, magnetic resonance.

338 1. Introduction Understanding how the internal architecture of porous solids affects their function as hydrocarbon or groundwater reservoirs, building materials, filters, electrodes, or catalyst supports, is central to the optimization of a broad range of technological applications involving fluid transport in porous materials. To this end, quantitative pore structure characterization, and particularly determination of the pore size distribution, has been actively pursued for over sixty years (Dullien 1992). Unlike many man-made porous solids, which are designed to have relatively narrow pore size distributions, natural sedimentary rocks comprise a random network of interconnected pores, the sizes of which can range from nanometers to millimeters. A key feature of rock pore structure is the presence of local constrictions (pore throats) in the void continuum, through which pores of different size communicate. This feature is suggested by observation of 2D thin sections of rock samples (Wardlaw and Cassan 1979) and confirmed by analysis of 3D volume data obtained by serial sectioning (Ioannidis et al. 1997) and X-ray computed microtomography (Lindquist et al. 2000). Additionally, the presence of pore length scales ranging over several orders of magnitude challenges conceptual descriptions of the pore space in terms of lattices possessing a fixed mesh size (hereafter referred to as ordinary lattices). Instead, a description of sedimentary rock pore space as a multi-scale percolation system has been proposed (Xu et al. 1997). In this context, the fractal nature of the solid-void interface in sedimentary rocks is well established (Katz and Thompson 1985, Wong and Howard 1986, Radlinski et al. 1999). This feature has important implications for determining the distribution of pore length scales in rocks, but fractal scaling laws cannot describe the microstructure over all length scales. In fact, a considerable fraction of the pore volume exhibits Euclidean features and is adequately described by models of grain packing and compaction (Bakke and Oren 1997, Thovert et al. 2001) or by 3D stochastic reconstruction from limited 2D morphological information (Adler et al. 1990, Liang et al. 2000, Talukdar et al. 2002). To fully understand the capillary properties of sedimentary rock, it is

339 necessary to quantify the entire spectrum of pore length scales actually present (Tsakiroglou and Payatakes 1993, Chang and Ioannidis 2002). For example, failure to model the electrical resistivity and wetting-phase relative permeability at low values of water saturation has been attributed to inadequate resolution of pore geometry over sub-micrometric length scales (Bekri et al. 2003; Han et al. 2009), that is well within the size range where fractal scaling laws apply. An experimentally validated picture of sedimentary rock, honoring length scales of the order of the grain size as well as length scales associated with microporosity, is very difficult to obtain and, to the best of our knowledge, has yet to be achieved. Numerous experimental probes of pore geometry are available (e.g., gas adsorption/condensation, small-angle scattering, mercury porosimetry, petrographic image analysis, NMR relaxometry and imaging, X-ray microtomography, etc.), yet no single method can probe five or more orders of magnitude of the pore length scale, as required. For example, direct imaging methods, such as backscatter scanning electron microscopy (BSEM) (Ioannidis et al. 1996, Blair et al. 1996) and X-ray microtomography (Spanne et al. 1994, Dong and Blunt, 2009) become unwieldy for providing statistically significant microstructure data at length scales smaller than a few m. These methods are not well suited for the study of the surface fractal characteristics of sedimentary rocks. Indirect imaging methods, like small-angle neutron and X-ray scattering (SANS and SAXS), yield the volume-averaged Fourier transform of the density correlation function on length scales ranging from 1 nm to about 10 m (Radlinski et al. 1999). Greater length scales, accounting for much of the pore volume in sedimentary rock, cannot be probed by these techniques. Other methods of pore structure characterization require the introduction of fluids into the pore space, seeking to exploit capillary properties (as in mercury intrusion porosimetry and N 2 gas or water vapor adsorption/condensation) or magnetic properties (as in NMR relaxometry) of the solid-fluid system. Mercury intrusion porosimetry (MIP) is widely used to probe invasively the pore space in the range 20 nm to 100 m. Unfortunately, this method does not provide the

340 pore size distribution, but instead gives the distribution of pore volume that is accessible to mercury through pore throats of different size (Ioannidis and Chatzis, 1993). Deconvolution of MIP data is complex and requires independent information on the size distribution, spatial order and interconnectedness of the pores (Tsakiroglou and Payatakes 2000; Tsakiroglou et al. 2009). Gas or vapour adsorption/condensation methods are unable to probe the entire range of pore length scales within sedimentary rocks, but have been used to establish the fractal characteristics of the solid-void interface in a manner consistent with SANS/SAXS measurements (Broseta et al. 2001). Finally, attempts to obtain the pore size distribution by analyzing the NMR relaxation dynamics of fluidsaturated samples have been met with variable success, depending on the approach taken (Dunn et al. 2002, Song et al. 2000). A promising approach capitalizes on the magnetic field inhomogeneity induced by an applied field due to the variation in magnetic susceptibility between solid and void (Song et al. 2000). Application of this method to Berea sandstone, for example, has revealed pores with radii in the range m (Song 2001, Chen et al. 2003). An alternative method of pore structure characterization based on the statistical fusion of SANS and BSEM data and their subsequent interpretation in terms of a polydispersed spherical pore (PDSP) model has been proposed by Radlinski et al. (2004). Application of this method to a sample of reservoir sandstone has provided the pore size distribution in the range 1 nm to 1 mm, probing both fractal and Euclidean aspects of the microstructure. The pore size information thus obtained was shown to be consistent with and complementary to MIP and NMR transverse relaxation data. Padhy et al. (2007) have reported a modification of this method, whereby information normally provided by SANS measurements is substituted approximately by a fractal scaling law using estimates of the surface fractal dimension obtained by MIP. This modification renders the method widely applicable by circumventing the acquisition of experimental SANS data. Applications of the method to a small number of soil and rock samples has been reported elsewhere (Tsakiroglou and Ioannidis 2008; Tsakiroglou et al. 2009).

341 In this paper, the approach of Radlinski et al. (2004) and Padhy et al. (2007) is followed to characterize systematically the microstructure of ten rather diverse sandstone samples. We then investigate whether this more comprehensive pore structure characterization can improve predictions of flow permeability and electrical conductivity by critical path analysis (CPA) (Katz and Thompson 1986; 1987). The paper is structured as follows. The theory behind the fusion and subsequent interpretation of BSEM and SANS data, as well as the approximation of SANS data, are described in Section 2. Section 3 recalls the critical path analysis (CPA) arguments used to estimate the permeability and electrical conductivity of porous rock from MIP measurements (Katz and Thompson, 1986, 1987) and presents a modification of the classical approach appropriate for prediction of the permeability and electrical conductivity using the new pore size information. This modification takes into account the presence of both pores and throats by modeling individual pore space channels as constricted circular cylinders. The sandstone samples studied and experimental methods employed are described in Section 4. Results are presented and discussed in Section 5, where the pore size information obtained is contrasted to MIP data, providing hitherto unavailable information on an apparent pore-to-throat size aspect ratio. Predictions of permeability and electrical conductivity, obtained by using the new pore structure information, are also compared to those provided by the classical theory of Katz and Thompson (1986, 1987), which assumes that individual pore space channels are cylinders of uniform cross-section. Key conclusions are summarized in Section Pore size distribution from imaging and scattering data A porous medium may be generally described in terms of a binary phase function Z (x), taking the value of unity if x points to void and zero otherwise (Adler et al. 1990). The moments of the phase function Z (x) constitute the basis for the statistical description of a microstructure with random disorder (Torquato

342 2002). The first two moments are readily accessible from binary BSEM images (Blair et al. 1996; Ioannidis et al. 1996) and correspond to the porosity and twopoint correlation function: Z(x), (1) S ( r) Z( x) Z( x ), (2) 2 r where r is a lag vector and angular brackets denote averaging. For isotropic media the function S ( ) depends only on the modulus of the lag vector, i.e., 2 r S r ) S ( ). Determination of S ( ) from binary micrographs of the pore space 2( 2 r is typically limited to length scales greater than about 1 m. 2 r In small-angle neutron scattering experiments (SANS and USANS), the measured scattering intensity I (Q) is the Fourier transform of the density-density correlation function (r ) (Glatter and Kratky 1982): Qr I Q 2 sin( ) ( ) 4 r ( r ) dr, (3) Qr 0 2 where ( r ) ( ) (1 ) R ( r ) and R z ( r ) ( S ( r ) )/( ) is the void-void z autocorrelation function. In Eq. (3), 2 ( ) is the scattering length density contrast, a material constant depending on grain density and chemical composition, and Q is the magnitude of the scattering vector. The latter depends on the scattering angle and beam wavelength as Q 4 sin( / 2) /. For periodic structures, the magnitude of the scattering vector is related to the characteristic size of the scattering object as Q 2 / r. Since rocks scatter neutrons as a quasi two-phase system (Radlinski et al. 1999), the function I (Q) measured by SANS and the function S ( ) calculated from BSEM images are a 2 r

343 Fourier transform pair. I (Q) data in the small-q (large-r) range, which cannot be obtained from SANS/USANS measurements, can be computed via Eq. (3) from S ( ) measured on binary BSEM images. It is on this basis that SANS and 2 r BSEM imaging information may be combined to obtain the scattering crosssection I (Q) in the range 10-7 < Q < 10-1 Å -1 (Radlinski et al. 2004). Inverse Fourier transform then yields the autocorrelation R z (r ) in the size range 10 Å < r < 1 mm: R z ( r ) ( ) 2 ( 1 ) 0 Q 2 I( Q) sin( Qr ) dq. (4) Qr The fusion of structural information from BSEM and SANS in the form of I (Q) data spanning the range 10-7 < Q < 10-1 Å -1, enables the determination of the pore size distribution in the range 10 Å < r < 1 mm. This has been recently done for a sample of reservoir sandstone, for which both SANS/USANS and BSEM data were available (Radlinski et al. 2004). Assuming that the pore space can be represented by an assembly of independent spherical pores with an arbitrary distribution f (r ) of radii r, the scattering intensity per unit volume is given by: R max 2 2 I( Q) ( ) Vr f ( r ) Fs ( Qr ) dr. (5) V r R min In Eq. (5), R max and R min are the maximum and minimum pore radii, respectively, V r 3 V( r ) (4 / 3) r is the volume of a sphere of radius r, R max V V f ( r ) dr is the average pore volume, f (r ) is the probability density of r R min r the pore size distribution, and F s (Qr ) is the structure factor for a sphere of radius r (Radlinski et al. 2000) :

344 sin( Qr ) Qr cos( Qr ) Fs ( Qr ) 3 (6) 3 Qr 2 The function f (r ) is determined by inversion of the extended I (Q) data using Eq. (5). This inversion is implemented in the computer program PRINSAS (Hinde 2004). In the absence of SANS information, the missing I (Q) data in the large-q range may be approximately reconstructed (Padhy et al. 2007). For a surface fractal object of dimension D, the scattering intensity follows the power law 6 Q D I ( Q) with 2 < D < 3, whereas the pore size distribution follows the power law ( D 1) f ( r ) r (Pfeifer and Avnir 1983). This scaling holds in the large-q (small-r) range, but breaks down for length scales of the order of tens of micrometers, that is for length scales of the order of the grain size. The range of pore length scales over which a fractal scaling law applies (large-q range) may also be accessed by MIP. Indeed, this technique has been used to determine surface fractal dimensions (Ehrburger-Dolle et al. 1994), yielding results in reasonable agreement with SAXS measurements (Blacher et al. 2000). Fractal analysis of MIP data is based on the scaling law: ds Hg dr r 2 D, (7) where S Hg (r ) is the sample saturation to mercury at capillary pressure P c 1 r. Eq. (7) is consistent with a scaling of the number-based pore size distribution according to the power law f ( r ) r ( D 1). Over a limited range of pore length scales, I (Q) data computed from S ( ) via Eq. (3) may also follow the scaling I ( Q) Q D 6 2 r, furnishing a rough estimate of D that can be compared to the one

345 obtained by analysis of MIP data using Eq. (7). Provided that correspondence between the two values is established, one may extrapolate I (Q) in the large Q- range according to 6 Q D I ( Q). In this manner it is possible to obtain f (r ) in the range 10 Å < r < 1 mm in the absence of SANS data. The systematic determination, interpretation and application of the complete distribution of pore sizes in sedimentary sandstones are a key objective of this paper. The cumulative pore volume distribution, F (R), is readily calculated from the number density f (r ) as follows: 1 R max F ( R) V V f ( r ) dr. (8) r R r If the pore space were an assembly of disconnected pores (e.g., spheres or cylinders with length equal to their diameter), then we should expect r R for S Hg ( r ) F( R), i.e., the two functions should be identical. In reality, the pore space is interconnected. Not only are large pores likely to be surrounded by smaller ones, but each pore is accessible through constrictions in its immediate vicinity (pore throats). Both situations cause pores of size R to be invaded by mercury at a capillary pressure higher than corresponds to their size. If the number fraction of pores allowed (open) to mercury during drainage is the number fraction of pores actually accessible by mercury is cumulative distribution of pore volume accessible by mercury is given by: q s and Y s, then the Ys ( qs ) FA( R) F( R), (9) q s where q s is related to the number density f (r ) as follows:

346 R max q f ( r ) dr. (10) s R In percolation studies, the function Y ) is known as the site accessibility s ( q s function. It has been determined for different kinds of pore networks represented by ordinary lattices and has been shown to depend on network topology, poresize spatial correlations and network size (Chatzis and Dullien 1977, Stauffer and Aharony 1992, Ioannidis et al. 1993). Clearly, the functions F A (R) and (r ) both measure the cumulative distribution of accessible pore volume, the former by pore size and the latter by throat size. The ratio R r for which SHg ( r ) FA( R) therefore determines an effective pore-to-throat size aspect ratio for pores of size R - a geometric parameter known to control the efficiency of immiscible displacement in porous media (Chatzis et al. 1983, Seth and Morrow 2007). It should be kept in mind, however, that determination of the function F A (R) requires knowledge of the pore accessibility function Y ), which in s ( q s turn depends on the percolation lattice that represents the pore structure. This is discussed further in Section 5.1. S Hg 3. Prediction of permeability and electrical formation factor A broad distribution of pore sizes implies that transport through the pore space must be understood in terms of a broad distribution of local conductances (Katz and Thompson 1986, 1987). On the basis of percolation arguments, it has been proposed that transport in such systems is dominated by those conductances with magnitude greater than some critical (threshold) value associated with a threshold pore diameter c. The critical value g c is defined as the largest conductance such that the set of conductances g g still forms g c g c,

347 an infinite, connected cluster. If all local conductances with values g gc are assigned the value g c and the rest are set to zero, then the sample conductance is found by maximizing: t c( ) p( p c g( ) g ), (11) where g is a function of the length parameter, p ( ) is the probability for a c given conductance to be greater than or equal to g c, t is the percolation exponent ( t 1. 9 for 3D random lattices with no long-range correlations) and p c p( c ) is the so-called percolation threshold. As long as the conductance has a maximum at max and assuming that the local pore geometry is cylindrical, the following results are obtained for the sample permeability and formation factor, respectively (Katz and Thompson 1986, 1987): k h 2 h ( 1 32) ( max ) p( max t c ) p, (12) F 1 e t c p( max ) p, (13) where h max for the hydraulic conductance is generally different from electrical conductance. e max for the Notably, Katz and Thompson (1987) did not use Eqs. (11)-(13), but instead postulated the following trial functions for the hydraulic conductance and the electrical conductance g e : g h 3 g h SHg ( ) SHg, c, (14) g e SHg ( ) SHg, c, (15)

348 where S Hg ( ) is the volume fraction of connected pore space composed of pore diameters of size and larger and S S ( ). The function S ) is Hg, c Hg c determined by an MIP experiment. The threshold saturation S Hg Hg (, c corresponds to the point at which the mercury first forms a sample-spanning cluster and may be identified with the inflection point of the mercury intrusion curve. appearance of The 3 and in Eq. (14) and Eq. (15), respectively, follows directly from the basic assumption that the pores are cylinders of length equal to their diameter. Using Eq. (14) and Eq. (15), these authors have derived the following expressions for the sample permeability and formation factor, respectively: h h 2 h k KT ( 1 89)( max c ) ( max) SHg ( max) SHg, c, (16) 1 e e F KT ( max c ) SHg ( max) SHg, c. (17) A computation of local conductance, consistent with the new pore structure information provided in this work, is based on an idealization of individual pore space channels as constricted cylinders (see Fig. 1). Accordingly, the trial functions, Eq. (14) and Eq. (15), are replaced by: R 3 g h FA ( R) S 3 2 Hg, c 16(1 ) 3 ( 1) 2 R (1 ) g e FA ( R) SHg, c, (18), (19) where the aspect ratio is a function of pore size R. The term 3 ( 2 1) / 2 in Eq. (18) arises from taking into account the additional resistance to fluid flow due

349 to the presence of the sudden contraction/expansion (Sisavath et al. 2002). The main differences between Eq. (18) and Eq. (19) and the trial functions postulated by Katz and Thompson (1987) are the appearance of the pore-to-throat size aspect ratio,, and the replacement of S Hg ( ) by (R). For = 1 (straight cylindrical pores), the new equations are appropriately reduced to Eq. (14) and Eq. (15). To derive expressions for the permeability and formation factor we consider the hydraulic and electrical conductance of a unit cell in the critical path (CP) (see Fig. 1). The cell hydraulic and electrical conductances are given by: F A g ( CP ) h h 16[1 ( max ) ( R 3 h 3 max ) h max ] 3 [( h max 2 ) 2, (20) 1] g ( CP ) e e cwrmax, (21) e (1 ) 2 max where and c w are the viscosity and electrical conductivity of the saturating fluid. Application of Darcy s law to the unit cell leads to the following expression for the permeability: ( CP ) x k g h, (22) A where x and A, are the length and cross-sectional area of the unit cell, respectively. The latter quantities are calculated from the following expressions: h h max max 1) h max x 2R (, (23)

350 A [ F A ( R ( R h max h max ) 2 ) S Hg, c. (24) ] Substituting Eq. (20), Eq. (23) and Eq. (24) into Eq. (22), we derive the following expression for the sample permeability: k h max 2 ( 16[1 ( h max h 3 max 1)( R ) ] 3 ) h 2 max h max [( h max 2 ) 2 1] F ( R A h max ) S Hg, c. (25) Similarly, application of Ohm s law to the unit cell leads to the following expression for the formation factor: c A F w ( CP ) g x, (26) e where e e max ( max 1) e max e 2 e x 2R and A R ) { [ F ( R ) S ]}. ( max A max Hg, c Therefore, the sample formation factor is given by the equation: F e max e F ( R ) S 1 A max Hg, c. (27) Eq. (25) and Eq. (27) highlight the effect of local constrictions in the void continuum (pore throats) on k and F. 4. Description of rock samples and experimental methods A total of ten sandstone samples were investigated. Six of the sandstone samples (C-series) are Late Cretaceous deep-sea turbidite sandstones (Ventura County, Southern California) inter-bedded with siltstone and shale. All of these

351 samples are poorly sorted and exhibit a rather heterogeneous microstructure. Sample MP5-6 is also well-cemented, cross-bedded sandstone (Mazomanie formation, Wisconsin). Three relatively clay-free sandstones were also studied. These include a Berea sandstone sample, a reservoir sandstone sample from the Toolachee formation in central Australia (sample M4), which has been extensively investigated elsewhere (Radlinski et al. 2004) and sample MP5-6, which is a hard, well-cemented sandstone (Wonewoc formation, Wisconsin). Sample petrophysical properties were measured on cylindrical core plugs (3.85 cm in diameter and 3 to 6 cm in length) and are reported in Table I. These include the porosity, measured gravimetrically by a saturation method, the Klinkenberg-corrected permeability (Dullien 1992) to nitrogen gas, measured by a steady-state method, and the electrical formation factor, measured with 20,000 ppm NaCl brine using a two-electrode method. Mercury porosimetry tests were performed on 1-cm 3 cubic samples, cut from the same plugs used in physical measurements and lightly coated with epoxy on all but one faces to minimize surface penetration effects (Larson and Morrow 1981). These tests were carried out to a maximum capillary pressure of 30,000 psia, stepping pressure to increasing values only after equilibration of the mercury volume. The threshold saturation S Hg, c and the corresponding capillary pressure o P c were assumed to correspond to the inflection point of the mercury intrusion curve (Katz and Thompson 1986, 1987). The characteristic length scale c was computed from o Pc 4 cos using a mercury-air surface tension = 480 mn/m and a c contact angle = 140 o. The S, and c values are listed in Table I. The Hg c reported accuracy of the length scale c is ±15% of its value and is due to errors in the determination of the inflection point (Katz and Thompson 1986). End-pieces from the core plugs were impregnated with a low-viscosity epoxy resin, first under vacuum and then under pressure. Flat, polished sections were used to obtain SEM images in the back-scatter mode with resolutions of m/pixel. Fifteen to twenty images were obtained from each sample and

352 processed using previously reported methods (Ioannidis et al. 1996) to determine the average and S ( ). Representative micrographs of the samples studied are shown in Fig r 5. Results and Discussion 5.1. Pore size distributions and apparent pore-to-throat aspect ratio In this section we illustrate in detail the computation of I (Q) data from binary image and MIP information, and the interpretation of these data in terms of Eq. (5), using as an example the Berea sandstone sample. The quality of the proposed approximation is also checked on sandstone sample M4, for which BSEM imaging, SANS/USANS and MIP data are all available. This is followed by comparisons of the inferred pore volume distributions F (R) and F A (R) to S Hg (r ) obtained from MIP for several of the samples investigated. The estimation of surface fractal dimension from MIP data using Eq. (7) is shown in Fig. 3(a) for the Berea sandstone sample. This analysis yields D = The data seem to suggest an upper cut-off of the fractal scaling near r 2.5 m, implying that ca. 25% of the pore volume may be attributed to a surface fractal. Yet, deviations from linearity do not necessarily mean departure from the scaling law ( D 1) f ( r ) r. This is because S Hg (r ) is sensitive to pore accessibility effects which are predominant in the vicinity of the percolation threshold, as evidenced by the peak of the ds Hg dr data near r 15 m (note in Table I that c = 32 m for this sample). As shown in Fig. 3(b), the trend 6 Q D I ( Q) with D = 2.58 agrees with the I (Q) data computed via Eq. (3) from the average image statistical properties. The fitting of the extended I (Q) data by Eq. (5) produces the pore size distribution f (r ) shown in Fig. 3(c). This distribution obeys ( D 1) f ( r ) r with D = 2.58 for pore length scales up to r 50 m, indicating that the majority of the pore volume (ca. 80%) in Berea sandstone

353 may be attributed to a surface fractal. The cumulative pore volume distribution F (R) of Berea sandstone is compared to MIP data in Fig. 3(d). As expected, the function F (R) reveals the presence of pores with radii as large as 100 m, consistent with the pore space micrograph shown in Fig. 2(a). About half of the pore volume of this sample is owed to pores with radius larger than ca. 30 m. This information cannot be obtained from MIP, which instead shows that about half of the pore volume is accessible through pore throats of radius larger than ca. 10 m. It is important to note that the cumulative number fraction of pores open at the percolation threshold R max q f ( r ) dr, that is when F A( Rc ) SHg, c, is sc R c vanishingly small. This is contrary to what is known about percolation in ordinary lattices (i.e., lattices with a fixed mesh size). For example, q for bondcorrelated site percolation in 3D cubic networks with no spatial correlations among the sites (e.g., Ioannidis et al. 1993). In this sense, pore accessibility functions for ordinary lattices, such as the one shown in Fig. 4, are inconsistent with the pore size distribution data shown in Fig. 3(c). Further interpretation requires assumptions regarding the organization of multiple pore length scales. While more sophisticated approaches are possible (Xu et al. 1997; Tsakiroglou et al. 2009), we shall simply assume here that at least some fraction of the pore volume can be represented by an ordinary lattice and explore the consequences of this assumption. If it can be supposed that only the accessibility of pores with sizes in the range ( R 0, R sc max ) is described by Y s ( q s ) data for a 3D cubic lattice (see Fig. 4), then the following equations apply:

354 q sc R R max R max R f ( r ) dr c f ( r ) dr 0, (28) Ys ( qsc ) SHg, c F( Rc ). (29) q sc Given the pore accessibility function Y ), simultaneous solution of Eq. (28) and Eq. (29) yields R 0 and the range ( R 0, R max c s ( q s R. The function F A (R) may then be determined in ) using Eq. (9). In the absence of accessibility information for pores with sizes in the range ( R, R 0 ), we assume that F A ( R) F( R) (i.e., min all open pores in this size range are accessible). For the Berea sandstone sample S, = 0.1 (Table I), and using q Hg c SC and the data of Fig. 4, we obtain R O 16.5 m and R C 25.8 m. The cumulative distribution of accessible pore volume by pore size, F A (R), is calculated from Eq. (9) using the pore accessibility function shown in Fig. 4 and is plotted alongside F (R) and S Hg (r ) in Fig. 5(a). A comparison of the (R) and S Hg (r ) data gives information on the pore-to-throat size aspect ratio. For the Berea sandstone sample, Fig. 5(b) shows that the aspect ratio varies in the narrow range 2 < < 2.8 for 0.01 < R < 30 m. This estimate is rather a lower bound on, since pores with radius greater than c F A R also contribute to F ). A( Rc Using the DDIF-NMR method (Song et al. 2000), Song (2001) has determined the distribution of pore volume by pore size in a Berea sandstone sample of similar porosity and permeability (sample Berea 300 in Song 2001) as the one reported here. Comparing the mode of this distribution to the mode of the distribution of accessible pore volume by pore throat size (obtained by MIP),

355 Song (2001) reported = 3.2, in fair agreement with the data shown in Fig. 5(b). It is instructive to compare in more detail the pore volume distribution F (R) plotted in Fig. 3(d) to the results of Song (2001). The volume-weighted pore size distribution obtained by DDIF-NMR (see Fig. 2 of Song 2001) is bimodal, with a predominant peak at pore radius of about 35 m and a second peak at pore radius of about 0.8 m. Pores of radius smaller than 0.5 m were not detected. The inset of Fig. 3(d) shows the Berea F (R) data as a histogram. A single peak at about 45 m and a long tail extending to pores as small as m in radius are observed. The predominant peak agrees with the results of Song (2001), but their observed secondary peak around 0.8 m is not reproduced by our analysis. The appearance of this secondary peak in DDIF- NMR data is likely an artifact, due to the averaging of magnetization between micropores and macropores caused by molecular diffusion (pore coupling effect; see Zielinski et al. 2002). Chen et al. (2003) have also obtained a bimodal volume-weighted pore size distribution from DDIF-NMR measurements on a different sample of Berea sandstone at 100% water saturation. Pores of radius smaller than about 1.5 m were not detected. Upon centrifugation to 31% water saturation, however, the secondary peak at small pore sizes disappeared and pore sizes as small as 0.3 m were measured. This behavior is predicted by pore network simulations of NMR magnetization decay (Chang and Ioannidis 2002), under conditions of diffusion coupling between micropores and macropores. Estimation of the surface fractal dimension from the MIP data of sample M4 yields D = 2.60 ± 0.03, as shown in Fig. 6(a). This value is to be compared to D = 2.49 ± 0.03, obtained from the SANS/USANS data (Radlinski et al. 2004) and with D = 2.53 ± 0.07 from analysis of the BSEM data. Using I ( Q) Q D with D = 2.60 in place of the actual SANS/USANS data we obtain the extended I (Q) data shown in Fig. 6(b). Upon inversion via Eq. (5), these data yield a pore 6

356 size distribution in very good agreement with the one reported by Radlinski et al. (2004) for this sample, as shown in Fig. 6(c). Figure 7 compares the F A (R) and S Hg (r ) data for some of the samples, with relevant parameters listed in Table I. The comparisons quantify a general trend of increasing pore-to-throat aspect ratio with increasing degree of cementation and presence of silt and clay. For example, much of the porosity of poorly-sorted sandstone sample C6-174 is due to pores larger than 20 m. These pores are, however, accessible through pore throats that are at least five times smaller Prediction of permeability and electrical formation factor Table II and Table III summarize the predictions of absolute permeability and formation factor, respectively. In these tables, the predictions of the classical theory, Eq. (16) and Eq. (17), are contrasted to the predictions obtained using Eq. (25) and Eq. (27). Characteristic pore length scales (pore size and aspect ratio) for electrical conductivity and permeability are also given in the tables. Figure 8(a) shows that Eq. (25) yields predictions of the permeability of the samples studied which are not greatly different from those obtained from Eq. (16), whereas Fig. 8(b) shows that Eq. (27) predicts higher values of formation factor by comparison to Eq. (17). These observations may be understood as follows. Comparing h h max h e e max Rmax to max 2 and Rmax to max 2 using the data of Table II and Table III, respectively, it is found that e R h max h max h max e max e max e max 2 and R 2, which imply that F h h A( Rmax Hg max e e A( max Hg max ) S ( ) and F R ) S ( ). Therefore, the new trial functions, Eq. (18) and Eq. (19), associate the same characteristic pore throat sizes and total pore volumes to the critical paths for viscous flow and electrical conduction, as do the trial functions chosen by Katz & Thompson (1986, 1987). Pore geometrical details other than the pore throat size characterizing the critical path have a greater influence on electrical conductivity than permeability, the

357 latter being predominantly determined by pore throat size. For both permeability and formation factor, predictions by either the formulas of Katz & Thompson (1986, 1987) or the ones developed here are within a factor of two of the experimental values. The simplifying geometric assumptions made in the analysis and uncertainties in the determination of S Hg, c, are factors contributing to the deviation between the experimental and predicted values. Equally important, however, is unaccounted sample inhomogeneity (Katz & Thompson, 1987). As with other approaches based on MIP data (Katz & Thompson, 1986, 1987) or image analysis (Blair et al. 1996; Ioannidis et al. 1996; Lock et al. 2002), which also claim agreement between measured and predicted permeability or formation factor to within a factor of two, it is assumed here that the samples used for imaging and mercury porosimetry are representative of the core sample on which permeability and formation factor were measured. This will not be the case if correlated heterogeneity is present over length scales greater than the size of the samples used for MIP (ca. 10 mm) or the size of binary images (ca. 2-5 mm). Methods for quantifying the length scales of heterogeneity in natural porous media (Pomerantz et al. 2008) will likely play an important role in elucidating the difference between prediction and measurement. 6. Summary and conclusions Using a method first proposed by Radlinski et al. (2004) and modified by Padhy et al. (2007), we determined the continuous distribution of pore sizes over the range m for a suite of sandstone samples, accessing quantitatively both surface fractal and Euclidean aspects of the microstructure. The data obtained show that a significant amount of pore volume is due to pores which are orders of magnitude smaller than the pores associated with breakthrough of a non-wetting phase during drainage. This picture is not accommodated by non-hierarchical (single-scale) lattices commonly used in invasion percolation modeling of capillarity and transport in porous media. Pore networks extracted from micro-computerized tomography images (e.g., Han et al.

358 2009; Dong & Blunt, 2009) are likely to miss significant amounts of sandstone pore volume contributed by pores below the resolution limit (a few micron). Critical path analysis arguments were also pursued to predict the permeability and formation factor of the samples studied using the new pore size information. The new predictions consider the effect of local constrictions in the pore space through estimates of an apparent pore-throat aspect ratio and are consistent with the predictions of the classical theory of Katz & Thompson (1986, 1987). Agreement between theory and experiment for both permeability and formation factor was within a factor of two, and therefore not significantly improved. In light of a large body of literature on the prediction of core-scale properties from mm-scale image and/or mercury porosimetry data, the present results also suggest the influence of unaccounted spatial heterogeneity and highlight the need for development of large-scale, high-resolution models of the pore space. Acknowledgments Financial support for this work, provided by the University Consortium for Field- Focused Groundwater Contamination Research, directed through the University of Guelph, Ontario, and by the Natural Sciences and Engineering Research Council of Canada (NSERC) is gratefully acknowledged. The authors thank Boeing Co. and Hydrite Chemical Co. for providing sandstone samples used in this work. References 1. Adler, P.M., Jacquin, C.G. & Quiblier, J.A. (1990). Flow in simulated porous media. International Journal of Multiphase Flow, 16, 691.

359 2. Bakke, S. & Oren, P. (1997). 3D pore-scale modeling of sandstones and flow simulations in pore networks. SPE Journal, 2(2), Bekri, S., Howard, J., Muller, J. & Adler, P.M. (2003). Electrical resistivity index in multiphase flow through porous media. Transport in Porous Media, 51, Blacher, S., Heinrichs, B., Sahouli, B., Pirard, R. & Pirard, J.-P. (2000). Fractal characterization of wide pore range catalysts: Application to Pd Ag/SiO 2 xerogels. Journal of Colloid and Interface Science, 226, Blair, S.C., Berge, P.A. & Berryman, J.G. (1996). Using two-point correlation functions to characterize microgeometry and estimate permeabilities of sandstones and porous glass. Journal of Geophysical Research, 101(B9), Broseta, D., Barre, L., Vizika, O., Shahidzadeh, N., Guilbaud, J.P. & Lyonnard, S. (2001). Capillary condensation in a fractal porous medium. Physical Review Letters, 86, Chang, D. & Ioannidis, M.A. (2002). Magnetization evolution in network models of porous rock under conditions of drainage and imbibition. Journal of Colloid and Interface Science, 253, Chatzis, I. & Dullien, F.A.L. (1977). Modeling pore structure by 2D and 3D networks with application to sandstones. Journal of Canadian Petroleum Technology, 16, Chatzis, I., Morrow, N.R. & Lim, H.T. (1983). Magnitude and detailed structure of residual oil saturation. SPE Journal, 23, Chen, Q., Gingras, M.K. & Balcom, B.J. (2003). A magnetic resonance study of pore filling processes during spontaneous imbibition in Berea sandstone. Journal of Chemical Physics, 119, Dong, H. & Blunt, M.J. (2009). Pore-network extraction from microcomputerized-tomography images. Physical Review E, 80, article no Dullien, F.A.L. (1992). Porous media: Fluid transport and pore structure. San Diego: Academic Press.

360 13. Dunn, K.J., Bergman, D.J. & LaTorraca, G.A. (2002). Nuclear magnetic resonance: Petrophysical and logging applications. In: K. Helbig & and S. Treitel (eds), Handbook of geophysical exploration, vol. 32, Oxford, Pergamon Press. 14. Ehrburger-Dolle, F., Lavanchy, A., & Stoeckli, F. (1994). Determination of the surface fractal dimension of active carbons by mercury porosimetry. Journal of Colloid and Interface Science, 166, Glatter, O. & Kratky, O. (1982). Small-angle X-ray scattering. London: Academic Press. 16. Han, M., Youssef, S., Rosenberg, E., Fleury, M. & Levitz, P. (2009). Deviation from Archie s law in partially saturated porous media: Wetting film versus disconnectedness of the conducting phase, Physical Review E, 79, Hinde, A.L. (2004). PRINSAS - a Windows-based computer program for the processing and interpretation of small-angle scattering data tailored to the analysis of sedimentary rocks, Journal of Applied Crystallography, 37, Ioannidis, M.A. & Chatzis, I. (1993). Network modeling of pore structure and transport properties of porous media. Chemical Engineering Science, 48, Ioannidis, M.A., Chatzis, I. & Sudicky, E.A. (1993). The effect of spatial correlation on the accessibility characteristics of 3-dimensional cubic pore network as related to drainage displacements in porous media. Water Resources Research., 29, Ioannidis, M.A., Kwiecien, M.J. & Chatzis, I. (1996). Statistical analysis of the porous microstructure as a method for estimating reservoir permeability. Journal of Petroleum Science and Engineering, 16, Ioannidis, M.A., Kwiecien, M.J., Chatzis, I., Macdonald, I.F. & Dullien, F.A.L. (1997). Comprehensive pore structure characterization using 3D computer reconstruction and stochastic modeling. SPE Preprint 38713,

361 presented at the 1997 SPE annual technical conference and exhibition. San Antonio, Texas. 22. Katz, A.J. & Thompson, A.H. (1985). Fractal sandstone pores: Implications for conductivity and pore formation. Physical Review Letters, 54, Katz, A.J. & Thompson, A.H. (1986). Quantitative prediction of permeability in porous rock. Physical Review B, 34, Katz, A.J. & Thompson, A.H. (1987). Prediction of rock electrical conductivity from mercury injection measurements. Journal of Geophysical Research Solid Earth and Planets, 92(B1), Larson, R.G. & Morrow, N.R. (1981). Effects of sample size on capillary pressures in porous media. Powder Technology, 30, Liang, Z., Ioannidis, M.A. & Chatzis, I. (2000). Permeability and electrical conductivity of porous media from 3D stochastic replicas of the microstructure. Chemical Engineering Science, 55, Lindquist, W.B., Venkatarangan, A., Dunsmuir, J. & Wong, T.F. (2000). Pore and throat size distributions measured from synchrotron X-ray tomographic images of Fontainebleau sandstones. Journal of Geophysical Research - Solid Earth, 105(B9), Lock, P.A., Jing, X.-D., Zimmerman, R.W. & Schlueter, E.M. (2002). Predicting the permeability of sandstone from image analysis of pore structure. Journal of Applied Physics, 92, Padhy, G.S., Lemaire, C., Amirtharaj, E.S. & Ioannidis, M.A. (2007). Pore size distribution in mulitscale porous media as revealed by DDIF-NMR, mercury porosimetry and statistical image analysis. Colloids and Surfaces A Physicochemical and Engineering Aspects, 300, Pfeifer, P. & Avnir, D. (1983). Chemistry in non-integer dimensions between two and three. 1. Fractal theory of heterogeneous surfaces. Journal of Chemical Physics, 79,

362 31. Pomerantz, A.E., Tilke, P. & Song, Y.-Q. (2008). Inverting MRI measurements to heterogeneity spectra. Journal of Magnetic Resonance, 193, Radlinski, A.P., Radlinska, E.Z., Agamalian, M., Wignall, G.D., Lindner, P., & Randl, O.G. (1999). Fractal geometry of rocks. Physical Review Letters, 82, Radlinski, A.P., Boreham, C.J., Lindner, P., Randl, O., Wignall, G.D., Hinde, A.L., & Hope, J.M. (2000). Small-angle neutron scattering signature of oil generation in artificially and naturally matured hydrocarbon source rocks. Organic Geochemistry, 31, Radlinski, A.P., Ioannidis, M.A., Hinde, A.L., Hainbuchner, M., Baron, M., Rauch, H. & Kline, S.R. (2004). Angstrom to millimeter characterization of sedimentary rock microstructure. Journal of Colloid and Interface Science, 274, Seth, S. & Morrow, N.R. (2007). Efficiency of the conversion of work of drainage to surface energy for sandstone and carbonate. SPE Reservoir Evaluation & Engineering, 10(4), Sisavath, S., Jing, X.-D., Pain, C.C. & Zimmerman, R.W. (2002). Creeping flow through an axisymmetric sudden contraction or expansion. Journal of Fluids Engineering Transactions of the ASME, 124, Song, Y.Q. (2001). Pore sizes and pore connectivity in rocks using the effect of internal field. Magnetic Resonance Imaging, 19, Song, Y.Q., Ryu, S.G. & Sen, P.N. (2000). Determining multiple length scales in rocks. Nature, 406, Spanne, P., Thovert, J.-F., Jacquin, C.G., Lindquist, W.B., Jones, K. & Adler, P.M. (1994). Synchrotron computed microtomography of porous media: Topology and transports. Physical Review Letters, 73, Stauffer, D. & Aharony, A. (1992). Introduction to percolation theory, London: Taylor & Francis.

363 41. Talukdar, M.S., Torsaeter O., Ioannidis, M.A. & Howard, J.J. (2002). Stochastic reconstruction of chalk from 2D images. Transport in Porous Media, 48, Thovert, J.-F., Yousefian, F., Spanne, P., Jacquin, C.G. & Adler, P.M. (2001). Grain reconstruction of porous media: Application to a lowporosity Fontainebleau sandstone. Physical Review E, 63, Torquato, S. (2002). Random heterogeneous materials: Microstructure and macroscopic properties. New York: Springer-Verlag. 44. Tsakiroglou, C.D. & Ioannidis, M.A. (2008). Dual-porosity modelling of the pore structure and transport properties of a contaminated soil, European Journal of Soil Science, 59, Tsakiroglou, C.D., Ioannidis, M.A., Amirtharaj, E. & Vizika, O. (2009). "A new approach for the characterization of the pore structure of dual porosity rocks", Chemical Engineering Science, 64, Tsakiroglou, C.D. & Payatakes, A.C. (1993). Pore wall roughness as a fractal surface and theoretical simulation of mercury intrusion-retraction in porous media, Journal of Colloid and Interface Science, 159, Tsakiroglou, C.D. & Payatakes, A.C. (2000). Characterization of the pore structure of reservoir rocks with the aid of serial sectioning analysis, mercury porosimetry and network simulation. Advances in Water Research, 23, Wardlaw, N.C. & Cassan, J.P. (1979). Oil recovery efficiency and the rock-pore properties of some sandstone reservoirs. Bulletin of Canadian Petroleum Geology, 27, Wong, P.-Z. & Howard, J. (1986). Surface roughening and the fractal nature of rocks. Physical Review Letters, 57, Xu, K., Daian, J.-F. & Quenard, D. (1997). Multiscale structures to describe porous media. 1. Theoretical background and invasion by fluids. Transport in Porous Media, 26,

364 51. Zielinski, L.J., Song, Y.-Q., Ryu, S. & Sen, P.N. (2002). Characterization of coupled pore systems from the diffusion eigenspectrum. Journal of Chemical Physics, 117,

365 Table I. Sample petrophysical properties from core and image analysis. Sample k (md) F S Hg, c c ( m) D R o ( m) R c ( m) C C C C C C MP MP M Berea

366 Table II. Prediction of permeability by Eq. (23) ( k new ) and Eq. (14) ( k KT ). Sample k core (md) k new (md) h R max ( m) h max k KT (md) h max ( m) C C C C C C MP MP M Berea

367 Table III. Prediction of formation factor by Eq. (25) ( F new ) and Eq. (15) ( F KT ). Sample F core F new e R max ( m) e max F KT e max ( m) C C C C C C MP MP M Berea

368 Figure captions 1. Geometric representation of an individual pore space channel as a constricted circular cylinder. Each channel is characterized by a pore and a throat size. 2. Typical BSEM images of the pore space of sandstone samples: (a) Berea sample, (b) sample M4, (c) sample MP5-7, (d) sample MP5-6, (e) sample C6-174, (f) sample C4-137, (g) sample C6-495, (h) sample C4-207, (i) sample C3-408, and (j) sample C (a) Estimation of surface fractal dimension of Berea sandstone sample from MIP data. Straight line corresponds to D = (b) Synthetic scattering intensity for Berea sandstone sample from BSEM and MIP data. The fit of these data by the polydispersed spherical pore (PDSP) model is shown as a solid line. (c) Number-based pore size distribution of Berea sandstone sample inferred from PDSP fit. (d) Cumulative pore volume distribution of Berea sandstone sample by pore size ( F (R)) and by accessible throat size ( S Hg (r ) ). Inset is a histogram representation of F (R). 4. Cumulative number fraction of accessible pores as a function of the cumulative number fraction of open pores (solid line) for simple cubic lattice network with no spatial correlations. 5. (a) Cumulative distributions of total pore volume ( ), F (R), and accessible pore volume (---), F A (R), by pore size compared to the distribution of accessible pore volume by throat size ( ), S Hg (r ), for Berea sandstone sample. (b) Apparent pore-to-throat size aspect ratio for Berea sandstone sample. 6. (a) Estimation of surface fractal dimension of M4 sandstone sample from MIP data. Straight line corresponds to D = 2.60 ± (b) Comparison of synthetic scattering intensity for M4 sample (circles this work) to

369 scattering intensity from BSEM/SANS/USANS (triangles Ref. [34]). (c) Number-based pore size distribution of sample M4 from inversion of the data of Fig. 6(b) using Eq. (5). 7. Comparison of pore volume distributions, F (R)( ), (R) (---) and S Hg (r ) ( ) for representative samples: (a) M4, (b) MP5-6, (c) MP5-7, (d) C6-174, (e) C4-137, (f) C Critical path analysis predictions of (a) absolute permeability and (b) formation factor of the samples studied. Dashed lines represent a factor of two deviation between prediction and measurement. F A

370 Solid 2R 2r R 2r R Fig. 1

371 (a) (b) (c) (d) (e) (f) Fig. 2

372 (g) (h) (i) (k) Fig. 2 (continued)

373 ds Hg /dr ( m -1 ) (a) r ( m) Fig. 3(a)

374 I (Q ) (cm -1 ) 1.E+15 1.E+14 1.E+13 1.E+12 1.E+11 1.E+10 1.E+09 1.E+08 1.E+07 1.E+06 1.E+05 1.E+04 1.E+03 1.E+02 1.E+01 1.E+00 1.E-01 (b) BSEM - Eq. (3) Extrapolated data Fit - Eq. (5) 1.E-08 1.E-07 1.E-06 1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 Q (Angstrom -1 ) Fig. 3(b)

375 f(r ) (Angstrom -1 ) 1.E+00 1.E-04 1.E-08 1.E-12 1.E-16 1.E-20 1.E-24 1.E-28 (c) 1.E-32 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08 r (Angstrom) Fig. 3(c)

376 Cumulative pore volume F(R) SHg(r) (a) (d) R ( m) Pore (R) or throat (r) radius ( m) Fig. 3(d)

377 Fig. 4

378 Fig. 5(a) (a)

379 6 Apparent aspect ratio, (b) Pore radius, R ( m) Fig. 5(b)

380 ds Hg /dr ( m -1 ) (a) r ( m) Fig. 6(a)

381 I (Q ) (cm -1 ) 1.E+15 1.E+14 1.E+13 1.E+12 1.E+11 1.E+10 1.E+09 1.E+08 1.E+07 1.E+06 1.E+05 1.E+04 BSEM + extrapolated data 1.E+03 1.E+02 BSEM/SANS/USANS data 1.E+01 1.E+00 1.E-01 1.E-02 (b) 1.E-03 1.E-04 1.E-08 1.E-07 1.E-06 1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 Q (Angstrom -1 ) Fig. 6(b)

382 f(r ) (Angstrom -1 ) 1.E+01 1.E-01 1.E-03 1.E-05 1.E-07 1.E-09 1.E-11 1.E-13 1.E-15 1.E-17 1.E-19 BSEM/SANS/USANS data BSEM + extrapolated data 1.E-21 1.E-23 (c) (b) 1.E-25 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 r (Angstrom -1 ) Fig. 6(c)

383 Fig. 7(a) (a)

384 Fig. 7(b) (b)

385 Fig. 7(c) (c)

386 Fig. 7(d) (d)

387 Fig. 7(e) (e)

388 Fig. 7(f) (f)

389 (a) Fig. 8(a)

390 (b) Figure 8(b).

391 20 ELEMENTS OF THE SANTA SUSANA FIELD LABORATORY SITE CONCEPTUAL MODEL OF CONTAMINANT TRANSPORT SITE CONCEPTUAL MODEL ELEMENT 2-2 DRAFT The Chatsworth Formation underlies most of the SSFL and its matrix porosity provided by interconnected pores is large. Prepared for: THE BOEING COMPANY THE NATIONAL AERONAUTICS AND SPACE ADMINISTRATION UNITED STATES DEPARTMENT OF ENERGY Prepared by: Jennifer C. Hurley 1, John A. Cherry 2, Beth L. Parker 2, 1 Schlumberger Water Services, Waterloo, Ontario, Canada 2 School of Engineering, University of Guelph, Guelph, Ontario, Canada Date: December 11, 2009

392 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Table of Contents 1 Abstract SSFL Stratigraphy and Lithology Rock Matrix Porosity Measurements Rock Matrix Physical Property Methods Matrix Porosity Matrix Porosity Values Thin Section Porosity Values BSEM Porosity Values Specific Gravity Results Measurement Method Comparison Gravimetric versus Specific Gravity Calculations Gravimetric versus Mercury Intrusion Influence of Depth on Rock Matrix Properties Conclusions References List of Tables Table 1. Summary of Matrix Porosity Measurements and Methods Table 2. Duplicate Analyses for Matrix Porosity Data Table 3. Statistical Summary for Porosity Values for Lithology Types at the SSFL Table 4. Summary of Selected SSFL and Literature Values for Rock Matrix Properties Table 5. Matrix Porosity Analytical Results for 17 Samples from Corehole C Table 6. University of New Brunswick Thin Section Porosity Values Table 7. Summary of Specific Gravity Data Table 8. Comparison of Vacuum Imbibition and Golder 2004 Specific Gravity Matrix Porosity Data Table 9. Comparison of Mercury Intrusion and Gravimetric Methods for Rock Matrix Porosity

393 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 List of Figures Figure 1. Stratigraphic Units at the SSFL... 6 Figure 2. Photographs of the lithology types identified in the field Figure 3. Enlarged photographs of the lithology types identified at the SSFL... 8 Figure 4. Frequency Distribution Histograms of Rock Matrix Porosity Data for each Lithology Type Figure 5. Mineral Composition of the Chatsworth Formation Figure 6. Mineral Composition of Chatsworth Formation Sandstone Figure 7. University of Waterloo Images of Sandstone Collected from Corehole C Figure 8. UW BSEM Image (left) and Binary Image (right) from 408'4 409 ft bgs in Corehole C Figure 9. UW BSEM Image (left) and Binary Image (right) from '11 ft bgs in Corehole C Figure 10. BSEM Image (left) and Binary Image (right) from 207'9-208'1ft bgs in Corehole C Figure 11. UW BSEM Image (left) and Binary Image (right) from 174'2 175 ft bgs in Corehole C Figure 12. BSEM Image (left) and Binary Image (right) from '8 ft bgs in Corehole C Figure 13. UW BSEM Image (left) and Binary Image (right) from 495' ft bgs in Corehole C Figure 14. UNB SEM Images showing Contrast between High and Low Porosity Samples Figure 15. UNB SEM Images showing Scalar Quality of Porosity Measurements Figure 16. Hurley (2003) Vacuum Imbibition Effective Matrix Porosity versus Golder (2004) Specific Gravity Total Matrix Porosity Figure 17. Gravimetric Fresh Water versus Mercury Intrusion Matrix Porosity Values Figure 18. Matrix Porosity Correlation with Depth for Samples from the SSFL

394 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December ABSTRACT The matrix porosity of the Chatsworth Formation is important concerning the distribution, transport, and fate of contaminants and therefore much effort was directed at measurement of this porosity using five different laboratory methods to determine the nature of the porosity and the spatial distribution at the Santa Susana Field Laboratory (SSFL). Previously, the various lithologies that constitute the Chatsworth Formation were grouped into nine categories encompassing various types of sandstone, siltstone, and breccia. Each measured value of porosity was assigned to one of these lithology types. Matrix porosity values were measured on 259 rock core samples from 13 coreholes drilled at the SSFL within both the Chatsworth Formation Operable Unit, and the Surficial Media Operable Unit. These samples are distributed across a wide range of lithology types so that a strong representation of matrix porosity variations was obtained. The values from 202 of these samples were used to conduct an assessment based on nine different lithological groupings. Overall, the porosity varies from a minimum of 0.73% by volume to a maximum of 23% by volume, with finer-grained rock and well-cemented rock exhibiting the lowest values, and poorly cemented clean sandstone generally showing the highest values. A comparison between methods that measure total porosity (ASTM D (specific gravity calculation method), thin section analysis, and BSEM imaging with those that measure effective porosity (vacuum weight difference, and mercury porosimetry) showed nearly identical values. This indicates that essentially all of the porosity in the rock matrix involves interconnected pore spaces (i.e., effective porosity). Therefore all of the porosity measurements, regardless of the method, can be taken as a useful representation of the effective porosity. The effective porosity is an important parameter in contaminant transport and fate in the Chatsworth Formation because molecular diffusion causes strong contaminant mass transfer between the fractures, where active groundwater flow occurs, and the lower permeability rock matrix, where groundwater is relatively immobile, but where there is large capacity for contaminant mass storage. Laboratory measurements reported elsewhere show that the effective diffusion co-efficient for chloride and by influence for other dissolved contaminants are relatively large, as expected, given the measured values for effective porosity that are typical of the Chatsworth Formation. 4

395 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December SSFL STRATIGRAPHY AND LITHOLOGY The bedrock at the study site is part of the Upper Cretaceous Chatsworth Formation; a turbidite sequence formed approximately 100 to 65 million years ago and composed of 60-70% thick-bedded coarse to medium grained graded arkose and lithic arkose sandstone with 25-35% siltstone, 1-2% breccia, and trace amounts of limestone, (Link et al., 1984). Turbidites are formed from turbulent gravity flows down oceanic continental margins and as such, include both terrestrial- and marine-derived sediment in varying quantities that are dependent upon the frequency of turbiditic flows and the rate of deposition of both sediment types. Sage Jr., (1971) used outcrops to map the bedrock in and around the study site and reported that those found at the SSFL are from the near-shore section of a gently northwestdipping turbidite deposit. Uplift and faulting occurred during the Late Miocene, (the last 20 million years), producing the roughly 300 metres of relief that now characterizes the Simi Hills, (Sage Jr., 1971; Hanson, 1981), and causing beds at the SSFL to dip sharply at 25 to 30 degrees to the northwest. Therefore, the coreholes drilled at the SSFL are effectively angled with respect to the sedimentary bedding, despite having been drilled vertically. This increases the probability of encountering joint sets oriented perpendicular to bedding because such joints would be angled roughly 60 to 65 degrees from horizontal rather than in the same plane as the direction of drilling. A study of the geology in the northeast area of the SSFL conducted by Dr. Ross Wagner at Montgomery Watson Harza, (Montgomery Watson Harza, 2001), was based on air photo analysis, field mapping, examination of existing geologic and topographic maps, and borehole geophysical logs for holes at the SSFL, (Sage Jr., 1971; Hanson, 1981; Dibblee, 1992; Harding Lawson Associates, 1995; Montgomery Watson Harza, 2002). This study by Wagner provides the stratigraphic information used in this document. Figure 1 is a geological column showing the stratigraphic units of the Chatsworth Formation at the SSFL. This figure indicates member names for each unit. Shallow bedrock below the SSFL is composed of Upper Chatsworth Formation sandstone with relatively few finergrained, thinly interbedded units of sandstone, siltstone, and shale. The three thickest interbedded units are Shale 2 (60m), the Woolsey Member (60m), and the Happy Valley Member (20-35m). Breccia was also found to comprise roughly 1% of the formation, (Link, Squires, and Colburn, 1984), although such beds are usually less than one metre thick and discontinuous. It is important to be able to characterize rock matrix physical properties in such a way that the total statistical variance of each property is minimized because these values must be used when calculating the phase partitioning of VOC total mass concentrations. In Hurley (2003), both stratigraphic and lithologic classification schemes were systematically evaluated for suitability in grouping physical property data. Hurley showed that at the SSFL, lithological classification of rock matrix properties was more appropriate. This approach has been used previously to explain variations in permeability and sorption in unconsolidated sandy aquifers, (Allen-King et al., 1998), as well as to explain shifts in organic matter content in fluvial valley deposits collected from Switzerland and Germany (Kleineidam et al., 1999). The lithologic approach has also been successful in explaining shifts in porosity and permeability for sandstones in Columbia (Warren and Pulham, 2001). 5

396 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Figure 1. Stratigraphic Units at the SSFL Sandstone units are shown in grey; interbedded sandstone, siltstone, and shale units are shown in green. 6

397 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Currently, physical properties at the SSFL are associated with one of nine lithology types. The primary types are defined as follows: Sandstone: typical fine- to coarse-grained Chatsworth Formation sandstone; Hard sandstone: well-cemented fine-grained sandstone requiring numerous blows with a chisel to break, light grey in colour and dry; Banded sandstone: alternating bands of light and dark grey sandstone; Breccia: matrix supported, likely sedimentary in origin, where the matrix is sandstone and the clasts range in size from pebbles to cobbles, frequently composed of siltstone/shale and very occasionally quartz and igneous rock; Siltstone (Interbedded): closely spaced unweathered sandstone, siltstone, and shale beds. For simplicity, all deep fine-grained beds were considered a part of this rock type; Silty sand: unconsolidated quaternary deposits of sand and silt; Weathered Sandstone: friable, oxidized, and often highly fractured fine- to coarsegrained Chatsworth Formation sandstone; Shallow Sandstone: fine- to coarse-grained Chatsworth Formation sandstone; and, Shallow Shale/Siltstone: shallow, weathered, closely spaced sandstone, siltstone, and shale beds. For simplicity, all shallow fine-grained beds were considered a part of this rock type. Of these lithologies, the first five are associated with the Chatsworth Formation Operable Unit, while the last four are generally only associated with the Surficial Media Operable Unit. Figure 2 contains photographs of hard sandstone, banded sandstone, siltstone/shale, breccia, and sandstone as they appeared immediately after coring. Figure 3 contains enlarged photographs of banded sandstone, siltstone/shale, breccia, fine-grained sandstone, fine to medium-grained sandstone, and fine to coarse-grained sandstone. Hard sandstone could not be visually differentiated from other sandstone lithology types, but was identified in the field by the hardness of the sandstone, which could take as many as 43 blows with a hammer and chisel to break. 7

398 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Figure 2. Photographs of the lithology types identified in the field. Hard sandstone was not visually different from other lithology types, but by the number of blows required to break the rock core. Figure 3. Enlarged photographs of the lithology types identified at the SSFL Hard sandstone could not be visually differentiated from other sandstone categories, and is therefore not shown. Finegrained, fine to medium-grained, and fine to coarse-grained sandstone are differentiated by the degree of sorting of the sandstone. Each core sample is 2.5cm in diameter. 8

399 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December ROCK MATRIX POROSITY MEASUREMENTS Table 1 presents all data sources for porosity data at the SSFL. Matrix porosity values measured by Sterling (1999) at the SSFL are not presented in this document or included in the following discussions as they were shown by Hurley (2003) to have a high degree of error associated with them due to the measurement method that was used. Analyses summarized in Groundwater Resources Consultants (1992), Golder (1997), and Hurley et al. (2003) were used to determine controls on parameter variability so that the potential error associated with applying a mean property value to each volatile organic chemical (VOC) rock core sample was minimized (Hurley et al., 2003). Under the supervision of AMEC, additional samples for physical property measurements were collected from 18 shallow groundwater coreholes within the Surficial Media Operable Unit and analyzed by Keantan Laboratories in Anaheim, California. Details of the collection and analysis of these samples are provided in the Near-Surface Groundwater Characterization Report (MWH, 2003). More recently, additional new matrix porosity analyses have been performed using a variety of techniques and are summarized in Golder (2004), SWS (2009) and Loomer (2009). Table 1 lists 17 matrix porosity measurements performed by the University of Waterloo from corehole C8 have never before been reported. These results are presented in Section 3.2 and discussed in the context of the original lithology-based distribution of values. These values are used for all rock matrix contaminant phase partitioning calculations and numerical simulations at the site. The rock physical property data set is somewhat biased. This is due to the interaction between the rock types encountered at the SSFL and the methods required to measure matrix physical properties. For example, in most cases, measurement of effective matrix porosity required full wetting and subsequent drying of the sample. However, siltstone samples tended to fall apart upon wetting. This occurred with one siltstone sample from the Golder (1997) data set, two mudstone samples from Sterling (1999), and four siltstone samples from the Hurley (2003) data set. Furthermore, lithologies other than massive sandstone (such as siltstone, hard sandstone, breccia, and banded sandstone) occur much less frequently within the Chatsworth Formation, so there are fewer sections available from which to extract samples to test. These factors mean that the overall data sets for lithologies other than sandstone are smaller. 3.1 ROCK MATRIX PHYSICAL PROPERTY METHODS Table 1 summarizes the methods used to measure matrix porosity values in rock core samples from the SSFL. The methods used to determine physical property values that define the statistical properties of each lithology type define the bulk of the data in this assessment, and are thus described in more detail below. 9

400 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Reference Porosity Measurements # Samples Method Employed GWRC (1992) 6 Specific gravity (ASTM D 854) CSU (1999) 15 Sterling (1999) 35 Vacuum weight difference (Collins, 1961; Dullien, 1992) Gravity with uneven shaped core (No reference) Chatzis (1997) 11 Mercury Porosimetry Golder (1997) 11 Specific gravity (ASTM D 854) Hurley (2003) 121 Vacuum weight difference (Collins, 1961; Dullien, 1992) MWH (2003) 38 American Petroleum Institute Recommended Practice 40 Golder (2004) 22 Specific gravity (ASTM D 854) Unreported until now 17 Vacuum weight difference (Collins, 1961; Dullien, 1992) RD 45, RD 49, RD 54C, and RD 55 RD 35B and RD 46B RD 35B and RD 46B RD 45, RD 49, RD 54C, and RD 55 RD 45, RD 49, RD 54C, and RD 55 C1, C2, C3, C4, C5, C6, C7 From 18 shallow coreholes C1, C2, C3, C4, C5, C6, C7 C8 Coreholes Tested Loomer (2009) 28 Thin section examination C1, C2, C3, C4 Total porosity Effective porosity Effective porosity; data excluded; method not accurate Effective porosity Total porosity Effective porosity Effective porosity Total porosity; values always higher than rest of data set statistically valid difference of 2.8% Effective porosity Data Quality Comment Total porosity; values always lowr than rest of data set statistically valid difference of 5 7% Table 1. Summary of Matrix Porosity Measurements and Methods Summary of all references containing published matrix porosity data from the SSFL, and listing the number of porosity samples within each reference, along with the methods used in determining these values. 10

401 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December Matrix Porosity Matrix porosity measurements in the samples used to define the statistical properties of each lithology type were determined gravimetrically in most cases from initially dry samples that were re-saturated under vacuum conditions (approximately 2x10 2 mbar) as described by (Collins, 1961) and (Dullien, 1992). Collins (1961) stated that this technique is one of the best methods of porosity measurement. The technique measures only those pore spaces that are interconnected, called effective porosity, and was modified to reduce the imbibition time from a matter of days to 20 minutes by pressurizing the headspace in the imbibition chamber with nitrogen gas to 100 psi after the deionized, deoxygenated water had been introduced to the chamber under vacuum conditions. Samples were sub-cored from the field samples using a diamond core barrel and tap water for cooling to produce cylindrical samples 3.5 cm in diameter and an average of 5.4 cm high, (heights ranged from 2.75 to 7.21 cm). Some of the interbedded samples broke due to their fissile nature while being sub-cored, resulting in measurement of only 16 interbedded sandstone, siltstone and shale samples. Five percent of the samples were measured twice (run in duplicate), to assess the precision of the method. Wet and dry bulk density values were then calculated based on the mass and volume information obtained from the porosity measurements, as described by Dullien (1992) and presented below. where ρ b-wet is the wet bulk density in g/cm 3, m wet is the wet weight of the rock sample in grams, determined after re-saturating under vacuum conditions, and v is the volume of the rock sample in cm 3 calculated from the measured height and diameter of each cylindrical rock core sample. where ρ b-dry is the dry bulk density in g/cm 3, and m dry is the dry weight of the rock sample in grams, determined after drying in an oven for 24 hours at 100 ο C and storage in a desiccator. Effective matrix porosity is then determined according to the equation below. where ρ water is the density of water at 25 ο C. Dullien (1992) does not specify the type of water that should be used in this analysis. Often, a salt (NaCl) brine solution of known density is used instead of water to prevent swelling of clays in the samples, which would result in lower porosity and bulk density values. However, because the samples from this research were from shallow depths (<200m bgs) and groundwater at the site is not a brine, deionized water was used instead and is considered to be more representative of field conditions. Table 2 summarizes duplicate sample values for matrix porosity. Duplicate measurements differed by values of 1 to 2 on a percent basis in all but one sample. The anomalous sample was from corehole C4 at m bgs and the poor reproducibility of this measurement is 11

402 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 likely caused by measurement error so that for this one sample, porosity varied by 6.07%. Thus, the overall average precision for matrix porosity was 1.37%, or 0.56% when the anomalous sample from C4 is omitted. Corehole Depth Rock Type n Sample Description (m) (% v) C sandstone fine sandstone C sandstone fine sandstone dup C sandstone fine to coarse sandstone, light olive brown C sandstone fine to coarse sandstone, light olive brown Dup C sandstone fine to medium sandstone with some gravel Dup C sandstone fine to medium sandstone with some gravel C sandstone fine to coarse sandstone with some curved banding Dup C sandstone fine to coarse sandstone with some curved banding C interbedded 6.85 strongly banded Dup C interbedded 5.57 strongly banded C banded sandstone 4.83 porosity/permeability C banded sandstone 11.9 porosity/permeability Dup C sandstone porosity/permeability: grey sandstone C sandstone porosity/permeability Dup: grey sandstone C sandstone grey ss with fine clasts C sandstone grey ss with fine clasts Dup Table 2. Duplicate Analyses for Matrix Porosity Data. Eight samples were run in duplicate for rock matrix porosity with precision ranging from one to two percent in all but the sample taken from C4 at m bgs. The average precision was 1.37% when this sample is omitted. The most common alternate method used to calculate matrix porosity values in the samples used to define the statistical properties of each lithology type was the ASTM D-854 method. This method involves calculation of total porosity values from measurements of specific gravity using the following equation: 1 where G s is the measured specific gravity of the rock matrix and is measured using ASTM Method D (1992). For this method, ρ water was taken as the density of water at 23 ο C. 3.2 MATRIX POROSITY VALUES Overall, the matrix porosity ranges across one and a half orders of magnitude from 0.7% to 19.3% by volume. Figure 4 is a frequency distribution histogram of porosity results for values used in each of the lithology categories. The sandstone lithology category has been subdivided to also show variation in each grain size range. Indicated in the upper left corner of each histogram is the number of samples in each analysis, and the mean property value for each category. This figure is complemented by Table 3, which is a statistical summary of the rock matrix porosity results by lithology category. Table 4 contains published literature values of rock matrix porosity for different lithology types. Wherever possible, ranges of measured values are listed, followed by the mean value and the number of samples analyzed in each study. The values determined in this study for 12

403 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 sandstone, siltstone, and breccia are also listed and compare favourably with literaturereported values. Porosity, organic carbon content, and hydraulic conductivity were found to be the three properties with the most variation both within and between sites. Published literature is included for only two tubiditic deposits, (Kleineidam et al., 1999 and Link et al., 1984), one of which (Link et al., 1984), is based upon analysis of samples within the Chatsworth Formation, the same formation that underlies the SSFL. Hard sandstone had the lowest mean value for matrix porosity, (2.9% by volume). Zones of hard sandstone were noted in C1, C3, C4, C5, C6, C7, C10, C11, RD-31, RD-35C, and RD- 39C (both the initial and step-out holes), as well as the non-uw supervised coreholes RD-75 and RD-85. No hard sandstone zones were noted in C2; however, this was the first corehole that was drilled in Hurley (2003) and it is likely that such zones were overlooked at that time because the sampling and logging procedure had not yet become well established. The lateral continuity of these zones is uncertain. If the hard sandstone zones are continuous over the scale of hundreds of metres, it is possible that they could be used as distinctive marker beds to trace stratigraphic units across faults. The mean hard sandstone matrix porosity value was three times smaller than mean values for other sandstone lithology types. Siltstone/shale had the second lowest mean value for matrix porosity, (6.7% by volume), while all other lithology types had mean values from 11 to 15 percent. Shallow siltstone has a very high standard deviation of 7.78%; however, only two samples from this lithology type have matrix porosity measurements, which is not enough to be statistically valid. Banded sandstone and silty sand also had high standard deviations, (4.0 and 4.20 respectively, on a percent basis) which equalled the variability measured for the data set as a whole. Siltstone/shale and breccia were similarly high, (3.8 and 3.7, respectively). However, these high standard deviations are expected given the nature of these lithology types (see descriptions in the previous section). Similarly, the purity of siltstone within the Chatsworth Formation is highly variable because of its dependence on the energy of the turbidity current at the time of deposition. The percent standard deviation for matrix porosity for sandstone is small (2.7). Soil Medium Minimum Mean Maximum Standard Deviation # of Samples (% volume) (% volume) (% volume) (% volume) Banded sandstone Breccia Hard sandstone Sandstone Shallow sandstone Shallow siltstone Siltstone Silty Sand Weathered sandstone Table 3. Statistical Summary for Porosity Values for Lithology Types at the SSFL Data from Golder (1999), Hurley (2003), and MWH (2003). 13

404 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Comparison of Literature and SSFL Physical Property Values for Lithology Types Lithology n Reference (% v) Sandstone 0.1 Pankow and Cherry, 1996 Sandstone, Germany (9) Boving and Grathwohl, 2001 Triassic Sherwood Sandstone, UK (ns) Thorton, 2000 Fine Sandstone , 0.33 (55) Morris, 1967 Sandstone , 0.37 (10) Morris, 1967 Cretaceous Newcastle Sandstone, USA (12) Manger, 1963 Sandstone, UK (ns) Lawrence et al., 2006 Sandstone, UK , (2) Bourg et al., 1993 Sandstone, USA 0.15 (ns) Schaefer et al., 2009 Birmingham Aquifer Sandstone, UK 0.28 (ns) Rivett et al., 2005 Permo Triassic sandstones, UK , (30) Gooddy and Bloomfield, 2006 Chatsworth Formation Sandstone Link et al., 1984 SSFL Sandstone , (134) Hurley, 2003 Interbedded, Yavne, Israel Mercado SSFL Siltstone , (13) Hurley, 2003 Siltstone , 0.35 (7) Morris, 1967 Bear River Formation Siltstone (1) Manger, 1963 Shale , 0.06 (20) Morris, 1967 SSFL Breccia , (5) Hurley, 2003 Bear River Formation Conglomerate (1) Manger, 1963 Table 4. Summary of Selected SSFL and Literature Values for Rock Matrix Properties Values from the SSFL are consistent with literature values. The range of values is indicated, followed by the arithmetic mean value, and the number of samples for the analysis in parentheses. ns indicates that the number of samples was not stated in the literature. After the initial statistical analysis of matrix porosity distributions across defined lithology types, 17 additional samples were analyzed by the University of Waterloo from corehole C8. This corehole is located within the Former Sodium Disposal Facility (FSDF) area; the easternmost area of the SSFL and an area thought to be composed of less fractured sandstone. All 17 samples are classed in the sandstone lithology type. Table 5 presents a summary of the matrix porosity values for these samples, which range from 10.87% to 17.34%, with a mean of 14.24%. This is well within the range observed for the other 105 samples sandstone samples, which ranged from 3.70% to 19.30%, with a mean of 13.60%. 14

405 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Average Depth (m) Sample Description n (% v) 9.97 grey coarse sandstone light brown fine to very coarse sandstone grey sandstone grey sandstone grey fine to very coarse sandstone fine grey sandstone grey sandstone grey sandstone finer sandstone very coarse sandstone grey sandstone grey fine to very coarse sandstone grey fine to very coarse sandstone grey fine to very coarse sandstone coarse sandstone coarse sandstone coarse sandstone Table 5. Matrix Porosity Analytical Results for 17 Samples from Corehole C Thin Section Porosity Values Sixty-eight thin sections were prepared from rock core samples selected to represent the lithologies of the Chatsworth Formation encountered in core holes C1 through C4. The thin sections were produced at Petrographic International, of Choiceland, Saskatchewan, where thin slices of rock, cut by a diamond blade saw, are mounted onto glass slides then polished down to a thickness of 30 microns (μm) and finally immersed in epoxy. Half of the samples were prepared using a blue epoxy so that the void space, e.g. matrix porosity, can be observed visually when the thin sections are viewed using a polarizing light microscope under plane-polarized light. Figure 5 and Figure 6 are representative thin section photographs of the sandstone of the Chatsworth Formation, where the matrix porosity is shown by the blue epoxy. Table 6 presents a summary of matrix porosity values as determined in 28 samples by the University of New Brunswick (Loomer, 2009) from thin section analysis. Measured porosities range from 1.1% to 11.4% and are therefore consistent with the values of 0 to 14% reported by Link et al. (1984) and those presented by Hurley (2003). 15

406 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Figure 4. Frequency Distribution Histograms of Rock Matrix Porosity Data for each Lithology Type Values not shown for the following lithology types: shallow sandstone, shallow siltstone, silty sand, and weathered sandstone. Data from Hurley (2003). Relative to each other, the magnitude of the values reflect a similar relationship as with other methods in that sandstone has the largest average porosity, while hard sandstone and banded sandstone have the lowest porosity. However, all matrix porosity values determined from thin section analysis are 5 to 7 percent lower than the averages determined from other methods for every lithology type except hard sandstone. There are three possible reasons for this, the combined effects of which may account for this difference. First, scale is an issue when using the thin section SEM method. Up to five areas per thin section were measured, but a thin section is a two-dimensional representation of approximately 3 x 5 cm of sample, and the area actually measured in the SEM is much smaller than that. As a result, SEM- 16

407 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 porosity values cannot be easily directly compared to bulk effective porosity measurements. Second, the sample size for the SEM analysis was quite small (n=28) compared to the rest of the analyses (n=202). It is possible that the SEM subset is not large enough to be statistically representative of the entire data population. Third, the glass cover slips neededd to be ground off the thin sections in order to be able to use them for SEM analysis. This is not a trivial thing to do because it is virtually impossible to ensure a sample surface of uniform thickness after the grindingg and repolishing. In some cases, part of the sample was wedge shaped: thicker at one end and thinning to the point of virtually having no sample thickness at the other end. The effect of this when looking at them in SEM is that what's left of the grains become smaller and farther apart until there is nothing but glass slide left. Analyzing these areas would produce artificially high porosity values due to the way porosity is determined in the SEM. To account for this, the SEM technician stayed well back from thin edges. This may have introduced a slight bias towards lower porosity areas of the sample, although not enough to account for the entire discrepancy between the SEM data and the rest of the analyses. PS Bt Qtz PS Qtz Figure 5. Mineral Composition of the Chatsworth Formation Photograph of representative Chatsworth Formation. Identifiable minerals include biotite and quartz. Matrix porosity is represented by blue epoxy. Photo was taken using an objective lens of 10X, where the field of view is 2 mm. (Qtz Quartz, Bt Biotite, and PS Pore Space) 17

408 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 PS Bt Pl PS Qtz Qtz Figure 6. Mineral Composition of Chatsworthh Formation Sandstone Photograph of representative sandstone sample from the Chatsworthh Formation. The photo was taken under plane- include biotite, plagioclase feldspar and quartz. The blue in the photograph represents the matrix porosity. (Qtz polarized light using an objective lens at a power of 10X, where the field of view is 2 mm. Identifiable minerals Quartz, Pl Plagioclase Feldspar, Bt Biotite, and PS Pore Space) 18

409 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Corehole Avg Depth (m) Rock Type UNB 2009 SEM Porosity C banded sandstone 10.5 C banded sandstone 1.6 C hard sandstone 2.3 C siltstone 1.2 C siltstone 1 C siltstone 1 C sandstone 11.3 C sandstone 9.9 C sandstone 9.3 C sandstone 8.7 C sandstone 8.3 C sandstone 8.3 C sandstone 7.8 C sandstone 7.5 C sandstone 7.2 C sandstone 7.2 C sandstone 7 C sandstone 6.1 C sandstone 5.8 C sandstone 5.7 C sandstone 5.1 C sandstone 5 C sandstone 4.9 C sandstone 4.9 C sandstone 4.5 C sandstone 4.4 C sandstone 4.1 C sandstone 3.1 Table 6. University of New Brunswick Thin Section Porosity Values Average sandstone matrix porosity from these results is 6.64%. Average banded sandstone matrix porosity from these results is 6.05%. Average siltstone matrix porosity from these results is 1.07%. Only one value reported for hard sandstone: 2.3% BSEM Porosity Values The microstructure characterization of rock core samples of the Chatsworth formation at SSFL (core holes C3, C4, C6) was carried out using BSEM (Backscatter Electron Microscope) images from thin sections of the rock samples. Initially the original core samples were cut into thin sections measuring 3.85cm diameter and 1cm -2cm in length. The thin section samples were dried and impregnated with epoxy, for effective impregnation this process is carried out under vacuum. The impregnated samples were then polished with a ¼ micron diamond abrasive. The polished thin sections were then coated with 10 nanometer (nm) layer of carbon. Finally, the thin section rock samples were observed under a backscatter electron microscope (Leo 1530 FESEM) to obtain high-resolution BSEM images. An example of such an image is shown in Figure 7a, where the pore space is represented by the black color and the rock grains in gray color. Each sample (thin section) provided at least such images. 19

410 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 The BSEM images were then subjected to a process called segmentation wherein the gray scale BSEM images are converted to binary (B/W) images (Figure 7b). Figure 7b shows the pore space in black color and the rock grains and cement in white color. This process was carried out using image analysis software called SCION Image. Figure 8 to Figure 13 present more BSEM and binary image pairs that further illustrate the matrix porosity distribution within the Chatsworth Formation bedrock at the SSFL. These BSEM images form the basis of the BSEM porosity calculation technique. Figure 14 and Figure 15 present complimentary SEM and binary images from thin sections assessed by the University of New Brunswick in These images highlight the scalar quality of porosity assessments made on the basis of thin section or BSEM calculations, and highlight why values determined using such techniques may not correlate well with values from other techniques, such as gravimetric methods. The microstructure characterization involves measuring the porosity and autocorrelation function associated with each rock sample. The autocorrelation function provides an estimate of the interconnectedness and spatial organization of the pores in a sample (Ioannidis et al., 1996). A specific program was written to calculate the porosity and ACF (autocorrelation function). Porosity is calculated by taking into account the percentage of black dots against the white dots. The program calculates the ACF by prodding the binary image with yardsticks of increasing lengths. A value of one is accounted for every black spot (pore space) encountered on the image and a value of zero otherwise (empty space). This is carried out on all the images obtained from each sample and the average binary values (0 and 1) are summed up in a mathematical equation. The outcome of which reveals the average pore size present in the rock sample and the extent of interconnectedness among the pores and further to estimate the permeability of the sample (Ioannidis et al., 1999). 20

411 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 A B Pore space Rock grain Figure 7. University of Waterlooo Images of Sandstone Collected from Corehole C4 a) Backscatter electron microscopy (BSEM) image of sandstone from 137 ft bgs in corehole C4 b) Binary image of sandstone from 137 ft bgs in corehole C4 21

412 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Figure 8. UW BSEM Image (left) and Binary Image (right) from 408'4 409 ft bgs in Corehole C3 Figure 9. UW BSEM Image (left) and Binary Image (right) from '11 ft bgs in Corehole C4 Figure 10. BSEM Image (left) and Binary Image (right) from 207'9-208' '1ft bgs in Corehole C4 22

413 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Figure 11. UW BSEM Image (left) and Binary Image (right) from 174'2 175 ft bgs in Corehole C6 Figure 12. BSEM Image (left) and Binary Image (right) from '8 ft bgs in Corehole C6 Figure 13. UW BSEM Image (left) and Binary Image (right) from 495' ft bgs in Corehole C6 23

414 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Figure 14. UNB SEM Images showing Contrast between High and Low Porosity Samples Examples of SEM BSE images (left) with the corresponding images (right) isolating the porosity (black). Labels provide the sample ID and the measured porosity in percent area for each image where (a) shows one of the highest porosities measured and (b) and (c) show lower porosity. 24

415 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Figure 15. UNB SEM Images showing Scalar Quality of Porosity Measurements. SEM BSE images (left) with the corresponding images (right) isolating the porosity (black). The bottom images are higher magnification images of the area outlined by the box in the top images. The images illustrate how changes in scale change the apparent porosity for an area Specific Gravity Results Table 7 presents a summary of all measurements of specific gravity made on rock core collected at the SSFL, and related parameters. Values range from 2.61 to 2.73, with a mean value of The majority of the measurements were made on samples corrected during the shallow bedrock investigation (MWH, 2003). Hard sandstone and breccia lithologies are not represented. 25

416 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Corehole Average Depth effective n effective n pb wet pb dry Matrix Tortuosity Specific Gravity Median Pore Throat Radius Lithology (m) (%v) (%v brine) (g/cc) (g/cc) Factor (/) (um) Tests Conducted By ILBS01S silty sand MWH (2003) ILBS05S silty sand MWH (2003) ILBS31S silty sand MWH (2003) ILBS12S silty sand MWH (2003) ILBS30S silty sand MWH (2003) PZ008GT shallow sandstone MWH (2003) CLBS39S silty sand MWH (2003) ILBS01S silty sand MWH (2003) ILBS35S silty sand MWH (2003) SB (Historical) MWH (2003) RD sandstone Golder, 1992, called CH 3 PZ004GT shallow sandstone MWH (2003) BVBS02S silty sand MWH (2003) ILBS01S silty sand MWH (2003) ILBS09S silty sand MWH (2003) ILBS32S silty sand MWH (2003) ILBS35S silty sand MWH (2003) ILBS37S silty sand MWH (2003) PZ003GT silty sand MWH (2003) PZ012GT02S weathered sandstone MWH (2003) PZ003GT shallow sandstone MWH (2003) PZ006GT shallow sandstone MWH (2003) CLBS38S silty sand MWH (2003) ILBS01S silty sand MWH (2003) ILBS08S silty sand MWH (2003) PZ007GT weathered sandstone MWH (2003) PZ011GT02S weathered sandstone MWH (2003) PZ014GT02S weathered sandstone MWH (2003) SBHV 3 15 (Historical) MWH (2003) PZ002GT sand MWH (2003) PZ008GT shallow sandstone MWH (2003) AABS02S silty sand MWH (2003) AABS06S silty sand MWH (2003) AFBS09S silty sand MWH (2003) CLBS40S silty sand MWH (2003) HVBS37S silty sand MWH (2003) ILBS15S silty sand MWH (2003) ILBS36S silty sand MWH (2003) PZ002GT silty sand MWH (2003) PZ009GT weathered sandstone MWH (2003) PZ012GT01S weathered sandstone MWH (2003) PZ013GT03S weathered sandstone MWH (2003) PZ014GT01S weathered sandstone MWH (2003) PZ016GT02S weathered sandstone MWH (2003) RD sandstone Golder, 1992, called CH 2 RD sandstone Golder, 1992, called CH 1 PZ005GT shallow sandstone MWH (2003) PZ007GT shallow sandstone MWH (2003) PZ011GT01S shallow sandstone MWH (2003) ILBS02S silty sand MWH (2003) PZ015GT01S weathered sandstone MWH (2003) PZ015GT02S weathered sandstone MWH (2003) PZ017GT01S weathered sandstone MWH (2003) RD sandstone Golder, 1992, called CH 1 PZ018GT01S shallow siltstone MWH (2003) PZ016GT01S weathered sandstone MWH (2003) PZ016GT03S weathered sandstone MWH (2003) PZ017GT02S weathered sandstone MWH (2003) SB (Historical) MWH (2003) RD 54C 8.69 sandstone Golder, 1992, called CH 3 RD sandstone Golder, 1992, called CH 1 AABS06S silty sand MWH (2003) CLBS06S silty sand MWH (2003) CLBS39S silty sand MWH (2003) PZ001GT weathered sandstone MWH (2003) PZ004GT weathered sandstone MWH (2003) PZ013GT01S weathered sandstone MWH (2003) SB (Historical) MWH (2003) RD interbedded Golder, 1992, called CH 3 RD sandstone Golder, 1992, called CH 3 RD sandstone Golder, 1992, called CH 2 PZ001GT shallow sandstone MWH (2003) PZ010GT shallow sandstone MWH (2003) CLBS31S silty sand MWH (2003) PZ009GT weathered sandstone MWH (2003) PZ004GT shallow sandstone MWH (2003) PZ006GT shallow sandstone MWH (2003) PZ005GT silty sand MWH (2003) SB (Historical) MWH (2003) RD interbedded Golder, 1992, called CH 2 PZ013GT02S weathered sandstone MWH (2003) PZ009GT shallow siltstone MWH (2003) AABS03S silty sand MWH (2003) Table 7. Summary of Specific Gravity Data A summary of all measurements of specific gravity made on rock core collected at the SSFL, and related parameters. 26

417 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December MEASUREMENT METHOD COMPARISON This section presents data comparisons for some of the various techniques that were used to measure total or effective matrix porosity in rock core samples collected from the SSFL. Such comparisons are useful because they provide insight in the quality of the measurements. 3.4 GRAVIMETRIC VERSUS SPECIFIC GRAVITY CALCULATIONS Table 8 presents a comparison of matrix porosity values from samples that were analyzed by both the gravimetric (vacuum imbibition) and the Specific Gravity methods. The gravimetric data is from the Hurley (2003) data set. Specific gravity measurements were reported as part of the Golder 2004 report. The Golder 2004 values are always less than those reported by Hurley 2003 for the same samples. The overall difference is 3%. The difference for sandstone is 2.8%. This is more than the standard deviation of 2.7% noted for samples of this lithology type (see Table 3) and is therefore considered to be statistically significant. The bias is particularly apparent in Figure 16, which is an XY scatter plot of paired values and where the red line on the graph represents an ideal 1:1 correlation. By definition, total matrix porosity values should be greater than effective matrix porosity values; however, this is not what is shown in this data set. Corehole Average Depth (m) Rock Type Vacuum Imbibition (% volume) Specific Gravity (% volume) Difference (% volume) C banded sandstone C breccia C hard sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone C sandstone Table 8. Comparison of Vacuum Imbibition and Golder 2004 Specific Gravity Matrix Porosity Data The Golder (2004) values are always greater than those reported by Hurley 2003 for the same samples. The overall difference is 3%. The difference for sandstone is 2.8%. This is more than the standard deviation of 2.7% noted for samples of this lithology type (see Table 3). 27

418 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Specific Gravity Calculated Matrix Porosity (% v) Vacuum Imbibition Method Matrix Porosity (% v) Figure 16. Hurley (2003) Vacuum Imbibition Effective Matrix Porosity versus Golder (2004) Specific Gravity Total Matrix Porosity The red line indicates the ideal 1:1 correlation. By definition, total n values should be greater than effective n values. 3.5 GRAVIMETRIC VERSUS MERCURY INTRUSION To assess the accuracy of the gravimetric method used to measure matrix porosity, a subset of the physical property samples was evaluated by both the gravimetric method using samples of known geometry, as well as by mercury intrusion. This comparison was based on two different data sets; six samples were analyzed using both techniques by Amirtharaj (2003), and 11 samples were analyzed gravimetrically by Golder (1992) and subsequently by Chatzis (1997) using mercury intrusion. Table 9 shows that only two of the 17 samples have percent differences greater than 30%, both of which are from the Amirtharaj data set, and nine of the samples have percent differences that are less than 15%. With only two exceptions, the gravimetric method produced higher porosity values than the mercury intrusion method. The two exceptions had percent differences of only eight and five percent. These data are also presented graphically in Figure 17, which is a plot of the paired gravimetric and mercury intrusion measurements, where the red line shows an ideal 1:1 correlation. The large porosity values provided by the gravimetric method were expected because mercury is non-wetting with respect to the mineral solids, while water is wetting. This allows water to coat the mineral grains and penetrate smaller pores than the mercury and is more consistent with what would be encountered by groundwater at the site. Thus, gravimetric values for porosity and wet and dry bulk densities form the basis for all necessary 28

419 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 calculations with regard to corehole rock and matrix porewater contaminant concentrations because they are more consistent with the pore spaces available to groundwater at the SSFL. Corehole Depth Sample Description n n Diff % Diff (m) grav mercury C3b grey fine to coarse sandstone % C4b grey fine to medium sandstone % C4b fine to coarse sandstone, grey % C6b grey sandstone % C6b sandstone % C6b fine to coarse sandstone % RD 45a light grey coarse sandstone % RD 45a light grey fine to coarse sandstone % RD 45a light brown very fine to fine sandstone % RD 49a light brown fine to coarse sandstone % RD 49a grey coarse sandstone % RD 49a grey very fine siltstone % RD 54Ca 8.69 light grey fine to coarse sandstone % RD 55a light brown fine to coarse sandstone % RD 55a grey siltstone % RD 55a 8.31 siltstone/shale dark grey % RD 55a light grey fine to coarse sandstone, some gravel % Table 9. Comparison of Mercury Intrusion and Gravimetric Methods for Rock Matrix Porosity a indicates samples that were analyzed by Golder (1997) and Chatzis (1997). b indicates samples that were analyzed by Amirtharaj (2003). The gravimetric and mercury intrusion methods compare favourably. All Chatzis/Golder samples have percent differences less than those found for Amirtharaj (2003). 18 Mercury Intrusion Porosity Values (n %v) Gravimetric Fresh Water Porosity Values (n %v) Figure 17. Gravimetric Fresh Water versus Mercury Intrusion Matrix Porosity Values 29

420 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 The red line indicates the ideal 1:1 correlation. By definition, gravimetric n values should be greater than mercury intrusion n values as mercury is a non-wetting fluid with respect to rock, while water is a wetting fluid. The data above are consistent with this. 3.6 INFLUENCE OF DEPTH ON ROCK MATRIX PROPERTIES It is conceivable that depth may influence rock matrix properties because of increased confining pressures which promote compaction and lithification of the rock. Increased confining pressures are encountered at depth in the coreholes and therefore are addressed in this section. Literature on the Ventura Field, an oil-bearing sandstone reservoir located just west of the Chatsworth Formation in Ventura County, shows that depth can influence rock matrix porosity values, although these effects were observed over depths of kilometres, (Hsu, 1977). Franklin and Dusseault (1989, page 34) state that porosity decreases systematically with depth, at a rate of about 1.3% for every 300-m depth. The coreholes at the SSFL penetrated to a maximum depth of roughly 426m bgs and thus, using the relation from Franklin and Dusseault, the influence of depth on properties would be negligible. However, Figure 18 shows that apart from the weathered upper bedrock within the surficial operable unit, the Chatsworth Formation Operable Unit (CFOU) shows no apparent correlation between the physical property data and depth within the environmental depth range relevant to the SSFL contaminant work Depth (m) Porosity (% v) Figure 18. Matrix Porosity Correlation with Depth for Samples from the SSFL Data from Hurley (2003). 30

421 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December CONCLUSIONS Matrix porosity values were measured on 259 rock core samples from 13 coreholes drilled at the SSFL from within both the Chatsworth Formation, and the Surficial Operable Unit. These samples are distributed across a wide range of lithology types so that a strong representation of matrix porosity variations was obtained. The values from 202 of these samples were used to conduct an assessment based on nine different lithological groupings. Overall, the porosity varies from a minimum of 0.73% by volume to a maximum of 23% by volume, with finer-grained rock and well-cemented rock exhibiting the lowest values, and poorly cemented clean sandstone generally showing the highest values. Comparison between matrix porosity values measured by methods that provide total porosity and those from methods providing the porosity of the interconnected pore space indicate that the total pore space is well connected, and that total porosity measurements are the same as effective porosity values. The well-connected nature of the rock matrix pore space supports other lines of evidence indicating that contaminants under chemical gradients can readily diffuse in the Chatsworth Formation. 31

422 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December REFERENCES Amirtharaj, E., Statistical synthesis of image analysis and mercury porosimetry for multiscale pore structure characterization, M.Sc. thesis, Dept. of Chemical Engineering, University of Waterloo. Chatzis, J., Mercury Porosimetry Test Results, Rock Core Samples, Rocketdyne Santa Susana Field Laboratory, Chatsworth, California, California Department of Chemical Engineering Porous Media Laboratory Report, Technical report. Colburn, I., L. Saul, and A. Almgren, The Chatsworth Formation: A New Formation Name for the Upper Cretaceous Strata of the Simi hills, California, in Simi Hills Cretaceous Turbidites, Southern California, The Pacific Section of the Society of Economic Paleontologists and Mineralogists, Los Angeles, USA. Collins, R., Flow of fuilds through porous materials, Reinhold Publishing Corporations, New York, USA. Dibblee, T., Geologic Map of the Calabassas Quadrangle, Los Angeles and Ventura Counties, California, Dibblee Geologic Map Foundation Map DF-37. Dullien, F., Porous media: fluid transport and pore structure, second edition, Academic Press, Inc, San Diego, USA. Franklin, J., and M. Dusseault, Rock engineering, McGraw-Hill Publishing Company, New York, USA. Golder Associates, Ltd., Matrix Diffusion Testing on Rock Core Samples: Rocketdyne Santa Susana Field Laboratory, Chatsworth, California, Technical Report, Golder Associates Ltd. Groundwater Resources Consultants,1992. Results of Collection and Analysis of Rock Cores, Technical report, GWRC. Hanson, D., Surface and Subsurface Geology of the Simi Valley Area, Ventura county, California, M.Sc. thesis, Oregon State University. Harding Lawson Associates, Geophysical and Hydrologic Testing Surveys: Rocketdyne/Santa Susana field laboratory, Ventura county, California, Technical Report Hurley, J., Rock Core Investigation of DNAPL Penetration and Persistence in Fractured Sandstone, M. Sc. thesis, University of Waterloo. Hurley, J., B. Parker, and J. Cherry, Source Zone Characterization at the Santa Susana Field Laboratory: Rock Core VOC Results for Core Holes C1 through C7. Department of Earth Sciences, University of Waterloo. 32

423 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Hurley, J., B. Parker, and S. Chapman, 2005a. Rock Core VOC Results for Corehole C8 Source Zone Characterization at the Santa Susana Field Laboratory Addendum Report No. 1. Department of Earth Sciences, University of Waterloo. Hurley, J., B. Parker, and S. Chapman, 2005b. Rock Core VOC Results for Corehole C6 Deepening Source Zone Characterization at the Santa Susana Field Laboratory Addendum Report No. 2. Department of Earth Sciences, University of Waterloo. Hurley, J., B. Parker, and S. Chapman, 2005c. Rock Core VOC Results for Corehole C9 Source Zone Characterization at the Santa Susana Field Laboratory Addendum Report No. 3. Department of Earth Sciences, University of Waterloo. Hurley, J., B. Parker, and J. Cherry, 2005d. Source Zone Characterization: Rock Core VOC Results for Shallow Groundwater Coreholes Santa Susana Field Laboratory. Department of Earth Sciences, University of Waterloo. Hsu, K., Studies of Ventura field, California; ii: Lithology, Compactions, and Permeability of Sands, American Association of Petroleum Geologists Bulletin, 61(2), Ioannidis, M.A., M.J. Kwicien and I. Chatzis, Computer Enhanced Core Analysis for Petrophysical Properties. Journal of Canadian Petroleum Technology, 38(18). Kleineidam, S., H. Rugner, and P. Grathwohl, Influence of Petrographic Composition/Organic Matter Distribution of Fluvial Aquifer Sediments on the Sorption of Hydrophobic Contaminants. Sedimentary Geology, 129, Link, M., R. Squires, and I. Colburn, Simi Hills Cretaceous Turbidites, Southern California, The Pacific Section of the Society of Economic Paleontologists and Mineralogists, New York, USA. Link, M., R. Squires, and I. Colburn, Slope and Deep-Sea Fan Facies and Paleogeography of Upper Cretaceous Formation, Simi Hills, California, The American Association of Petroleum Geologists Bulletin, 68(7), Loomer, D., Mineralogical characterization of drill core samples from the Santa Susana Field Laboratory, Ventura County, California, University of New Brunswick. Manger, G., Porosity and Bulk Density of Sedimentary Rocks, Geological Survey Bulletin 1144-E, United States Government Printing Office, Washington, USA. Montgomery Watson Harza, Technical Memorandum: Geologic Characterization of the Eastern Portion of the Santa Susana Field Laboratory, Ventura County, California. 33

424 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-2 DRAFT December 2009 Montgomery Watson Harza, Technical Memorandum: Collection of Various Bedrock Properties using Borehole Geophysical Logging Methods in the Eastern Portion of the Santa Susana Field Laboratory, Ventura county, California. Morris, D., and A. Johnson, Summary of Hydrologic and Physical Properties of Rock and Soil Materials, as Analysed by the Hydrologic Laboratories of the U.S. Geological Survey, Technical Report 1839-D, US Geological Survey Water Supply Paper. Sage Jr., O., Geology of the Eastern Portion of the "Chico" Formation, Simi Hills, California, M.Sc. thesis, University of California. Sterling, S, Comparison of Discrete Depth Sampling Using Rock Core and a Removable Multilevel System in a TCE Contaminated Fractured Sandstone. Master of Science Thesis. Department of Earth Science, University of Waterloo. Sterling, S., and B. Parker, Rock Core Sampling and Analysis for Volatile Organic Concentrations and Hydraulic Parameters in Boreholes RD-35B and RD-46B at the Santa Susana Field Laboratory, California. Warren, E., and A. Pulham, Anomalous Porosity and Permeability Preservation in Deeply Buried Tertiary and Mesozoic Sandstones in the Cusiana Field, llanos foothills, Columbia, Journal of Sedimentary Research, 71,

425 20 ELEMENTS OF THE SANTA SUSANA FIELD LABORATORY SITE CONCEPTUAL MODEL OF CONTAMINANT TRANSPORT SITE CONCEPTUAL MODEL ELEMENT 2-3 DRAFT An Analysis of the Distribution of Matrix Vertical Hydraulic Conductivity within the Chatsworth Formation. Prepared for: THE BOEING COMPANY THE NATIONAL AERONAUTICS AND SPACE ADMINISTRATION UNITED STATES DEPARTMENT OF ENERGY Prepared by: Jennifer C. Hurley 1, John A. Cherry 2, Beth L. Parker 2, 1 Schlumberger Water Services, Waterloo, Ontario, Canada 2 School of Engineering, University of Guelph, Guelph, Ontario, Canada Date: December 11, 2009

426 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-3 DRAFT December 2009 Table of Contents 1 Abstract SSFL Stratigraphy and Lithology Rock Matrix Permeability, and Hydraulic Conductivity Measurements Vertical Hydraulic Conductivity Methods Vertical Hydraulic Conductivity Values Relationship between Porosity and Permeability Matrix Permeability and Hydraulic Conductivity Methods Comparison Influence of Depth on Rock Matrix Properties Occurrence of Soft and Loose Zones Conclusions References List of Tables Table 1. Summary of Permeability Measurements and Methods... 9 Table 2. Duplicate Analyses for Matrix Hydraulic Conductivity Data Table 3. Viscosity and bulk density of water with temperature Table 4. Summary of Selected SSFL and Literature Values for Rock Matrix Properties Table 5. Statistical Summary for Hydraulic Conductivity Values for Lithology Types at the SSFL Table 6. Comparison between Vacuum Imbibition and Specific Gravity Matrix Porosity Data Table 7. Summary of Occurrence of Soft and Loose Zones within the Coreholes List of Figures Figure 1. Stratigraphic Units at the SSFL... 5 Figure 2. Photographs of the lithology types identified in the field Figure 3. Enlarged photographs of the lithology types identified at the SSFL... 7 Figure 4. Frequency Histograms of Matrix Vertical Hydraulic Conductivity Data Figure 5. Box Plots of the Hydraulic Conductivity Statistics for Each Lithology Figure 6. Comparison of Matrix and Bulk Hydraulic Conductivity Values for the Chatsworth Formation 16 Figure 7. Calculated versus Measured Matrix Hydraulic Conductivity Values Figure 8. Matrix Vertical Hydraulic Conductivity Correlation with Depth for Samples from the SSFL Figure 9. Graphical Distribution of Vertical Hydraulic Conductivity Values by Lithology and Study

427 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-3 DRAFT December ABSTRACT The rock matrix vertical hydraulic conductivity has been measured in a total of 194 samples from 51 coreholes collected during various drilling episodes beginning in All of the tested core samples were from vertically drill holes. The measurements were made on cylindrical samples cut from the rock core with the long axis of the samples parallel to the core. Therefore, all results represent the matrix hydraulic conductivity (K mv ) in the vertical direction. Depending on the drilling episode providing the cores, the measurements were made in five different laboratories, using two primary methods: water flow in a triaxial cell using EPA method 9100, and air permeability testing using ASTM method (2001). The values from 194 of these samples were used to conduct an assessment based on nine different lithological groupings. Almost all of the tests were conducted on sandstone samples, which provide a geometric mean value of 4.1x10-7 cm/s for conditions at 20 ο C. Relatively few shale/mudstone samples were tested because of three primary reasons: 1) this lithology occurs with less frequency within the Chatsworth Formation so there are fewer lengths of core from which to collect such samples, 2) the coring commonly provided less intact samples of this lithology because of an increase in the number of bedding plane fractures and a tendency to split along bedding in these units, and 3) the wetting process used in the triaxial tests typically caused sample deterioration for this lithology type. However, seven siltstone samples were analyzed and provided arithmetic and geometric mean values of 2.77x10-8 and 1.37x10-10 cm/s respectively, as expected, much lower than the sandstone values. The highest values within the Chatsworth Formation Operable Unit were obtained from samples indentified as breccia (with a geometric mean of 2.19x10-6 cm/s); the lowest values were obtained from samples identified as very well cemented (hard) sandstone, with a geometric mean of 4.73x10-9 cm/s. Occasionally in the core drilling, intervals were encountered where the sandstone was poorly lithified such that no core was obtained. Therefore, no lab K mv values were obtained for such intervals. It is likely that the K mv of the least lithified and therefore least cemented sandstone units is somewhat higher than the K mv for the sandstone units for which test results are available. 3

428 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-3 DRAFT December SSFL STRATIGRAPHY AND LITHOLOGY The bedrock at the study site is part of the Upper Cretaceous Chatsworth Formation; a turbidite sequence formed approximately 100 to 65 million years ago and composed of 60-70% thick-bedded coarse to medium grained graded arkose and lithic arkose sandstone with 25-35% siltstone, 1-2% breccia, and trace amounts of limestone, (Link et al., 1984). Turbidites are formed from turbulent gravity flows down oceanic continental margins and as such, include both terrestrial- and marine-derived sediment in varying quantities that are dependent upon the frequency of turbiditic flows and the rate of deposition of both sediment types. Sage Jr., (1971) used outcrops to map the bedrock in and around the study site and reported that those found at the Santa Susana Field Laboratory (SSFL) are from the nearshore section of a gently northwest-dipping turbidite deposit. Uplift and faulting occurred during the Late Miocene, (the last 20 million years), producing the roughly 300 metres of relief that now characterizes the Simi Hills, (Sage Jr., 1971; Hanson, 1981), and causing beds at the SSFL to dip sharply at 25 to 30 degrees to the northwest. Therefore, the coreholes drilled at the SSFL are effectively angled with respect to the sedimentary bedding, despite having been drilled vertically. This increases the probability of encountering joint sets oriented perpendicular to bedding because such joints would be angled roughly 60 to 65 degrees from horizontal rather than in the same plane as the direction of drilling. A study of the geology in the northeast area of the SSFL conducted by Dr. Ross Wagner at Montgomery Watson Harza, (Montgomery Watson Harza, 2001), was based on air photo analysis, field mapping, examination of existing geologic and topographic maps, and borehole geophysical logs for holes at the SSFL, (Sage Jr., 1971; Hanson, 1981; Dibblee, 1992; Harding Lawson Associates, 1995; Montgomery Watson Harza, 2002). This study by Wagner provides the stratigraphic information used in this document. Figure 1 is a geological column showing the stratigraphic units of the Chatsworth Formation at the SSFL. This figure indicates member names for each unit. Shallow bedrock below the SSFL is composed of Upper Chatsworth Formation sandstone with relatively few finergrained, thinly interbedded units of sandstone, siltstone, and shale. The three thickest interbedded units are Shale 2 (60m), the Woolsey Member (60m), and the Happy Valley Member (20-35m). Breccia was also found to comprise roughly 1% of the formation, (Link, Squires, and Colburn, 1984), although such beds are usually less than one metre thick and discontinuous. It is important to be able to characterize rock matrix physical properties in such a way that the total statistical variance of each property is minimized because these values must be used when calculating the phase partitioning of VOC total mass concentrations. In Hurley (2003), both stratigraphic and lithologic classification schemes were systematically evaluated for suitability in grouping physical property data. Hurley showed that at the SSFL, lithological classification of rock matrix properties was more appropriate. This approach has been used previously to explain variations in permeability and sorption in unconsolidated sandy aquifers, (Allen-King et al., 1998), as well as to explain shifts in organic matter content in fluvial valley deposits collected from Switzerland and Germany (Kleineidam et al., 1999). The lithologic approach has also been successful in explaining shifts in porosity and permeability for sandstones in Columbia (Warren and Pulham, 2001). 4

429 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-3 DRAFT December 2009 Figure 1. Stratigraphic Units at the SSFL Sandstone units are shown in grey; interbedded sandstone, siltstone, and shale units are shown in green. 5

430 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-3 DRAFT December 2009 Currently, physical properties at the SSFL are associated with one of nine lithology types. The primary types are defined as follows: Sandstone: typical fine- to coarse-grained Chatsworth Formation sandstone; Hard sandstone: well-cemented fine-grained sandstone requiring numerous blows with a chisel to break, light grey in colour and dry; Banded sandstone: alternating bands of light and dark grey sandstone; Breccia: matrix supported, likely sedimentary in origin, where the matrix is sandstone and the clasts range in size from pebbles to cobbles, frequently composed of siltstone/shale and very occasionally quartz and igneous rock; Siltstone (Interbedded): closely spaced unweathered sandstone, siltstone, and shale beds. For simplicity, all deep fine-grained beds were considered a part of this rock type; Silty sand: unconsolidated quaternary deposits of sand and silt; Weathered Sandstone: friable, oxidized, and often highly fractured fine- to coarsegrained Chatsworth Formation sandstone; Shallow Sandstone: fine- to coarse-grained Chatsworth Formation sandstone; and, Shallow Shale/Siltstone: shallow, weathered, closely spaced sandstone, siltstone, and shale beds. For simplicity, all shallow fine-grained beds were considered a part of this rock type. Of these lithologies, the first five are associated with the Chatsworth Formation Operable Unit, while the last four are generally only associated with the Surficial Media Operable Unit. Figure 2 contains photographs of hard sandstone, banded sandstone, siltstone/shale, breccia, and sandstone as they appeared immediately after coring. Figure 3 contains enlarged photographs of banded sandstone, siltstone/shale, breccia, fine-grained sandstone, fine to medium-grained sandstone, and fine to coarse-grained sandstone. Hard sandstone could not be visually differentiated from other sandstone lithology types, but was identified in the field by the hardness of the sandstone, which could take as many as 43 blows with a hammer and chisel to break. 6

431 20 Elements of the Santa Susana Field Laboratory Site Conceptual Model of Contaminant Transport SCM Element 2-3 DRAFT December 2009 Figure 2. Photographs of the lithology types identified in the field. Hard sandstone was not visually different from other lithology types, but by the number of blows required to break the rock core. Figure 3. Enlarged photographs of the lithology types identified at the SSFL Hard sandstone could not be visually differentiated from other sandstone categories, and is therefore not shown. Finegrained, fine to medium-grained, and fine to coarse-grained sandstone are differentiated by the degree of sorting of the sandstone. Each core sample is 2.5cm in diameter. 7