MODELING APPROACH MEMORANDUM (1) REMEDIAL INVESTIGATION/FEASIBILITY STUDY NEWTOWN CREEK

Transcription

1 Photo by Bill Rhodes MODELING APPROACH MEMORANDUM (1) REMEDIAL INVESTIGATION/FEASIBILITY STUDY NEWTOWN CREEK Prepared by Anchor QEA, LLC 305 West Grand Avenue, Suite 300 Montvale, New Jersey March 2012

2 MODELING APPROACH MEMORANDUM (1) REMEDIAL INVESTIGATION/FEASIBILITY STUDY, NEWTOWN CREEK Prepared by Anchor QEA, LLC 305 West Grand Avenue, Suite 300 Montvale, New Jersey March 2012

3 TABLE OF CONTENTS 1 INTRODUCTION Overview of Newtown Creek Physical Setting General Description of Modeling Framework Modeling Study Goals and Objectives DESCRIPTION OF MODELING FRAMEWORK Combined Sewer Overflow/Stormwater Model Hydrodynamic Model Sediment Transport Model MODEL DEVELOPMENT AND CALIBRATION Development and Calibration of CSO/SW Model Data Requirements Model Development Model Calibration and Validation Sensitivity Analysis Development and Calibration of Hydrodynamic Model Data Requirements Model Development Model Calibration and Validation Sensitivity Analysis Development and Calibration of Sediment Transport Model Data Requirements Model Development Model Calibration and Validation Sensitivity Analysis REFERENCES Newtown Creek RI/FS i

4 List of Figures Figure 1-1 Schematic of Newtown Creek Modeling Framework Figure 2-1 Schematic of Newtown Creek Hydrodynamic and Sediment Transport Models Figure 2-2 Primary Processes and External Forcing Incorporated into 3-D Hydrodynamic Model Figure 2-3 Schematic of Approximation Method for Estimating Feedback Between Hydrodynamic and Sediment Transport Models Figure 3-1 Spatial Extent of Model Domain List of Appendices Appendix A Details of Sediment Transport Model Theory and formulation Appendix A Figures Figure A-1 Probability of Deposition for Cohesive Sediment Using the Krone Formulation Figure A-2 Probability of Deposition for Non-cohesive Sediment as a Function of Bed Shear Stress and Particle Diameter Figure A-3 Settling Speed of Discrete Sediment Particles as a Function of Particle Diameter Figure A-4 Probability of Suspension as a Function of Bed Shear Stress for Particle Diameters of 130 and 540 µm Figure A-5 Particle Shielding Factor as a Function of Particle Size Figure A-6 Schematic of Interactions Between the Water Column, Active Layer, and Parent-Bed Layer When the Active-buffer Layer is Present Figure A-7 Schematic of Interactions Between the Water Column, Active Layer, and Parent-Bed Layer When the Active-buffer Layer is Not Present Figure A-8 Initial Structure of Bed with No Active-buffer Layer at Time = t1 Figure A-9 Active-surface Layer Thickness Increases as Shear Stress Increases (τ2 > τ1) at Time = t2 Figure A-10 Active-surface Layer Thickness Decreases and Active-buffer Layer is Created as Shear Stress Decreases (τ3 < τ2) at Time = t3 Figure A-11 Active-surface Layer Thickness Decreases and Active-buffer Layer Thickness Increases as Shear Stress Continues to Decrease (τ4 < τ2) at Time = t4 Newtown Creek RI/FS ii

5 Figure A-12 Active-surface Layer Thickness Increases and Active-buffer Layer Thickness Decreases as Shear Stress Increases (τ5 > τ4) at Time = t5 Figure A-13 Active-surface Layer Thickness Increases and Active-buffer Layer is Destroyed as Shear Stress Increases (τ6 > τ5, τ6 > τ2) at Time = t6 Figure A-14 Active-surface Layer Thickness Decreases and New Active-buffer Layer is Created as Shear Stress Decreases (τ7 < τ6) at Time = t7 Figure A-15 Initiation of Motion and Suspension for a Current Over a Plane Bed, θ = f(d*), from van Rijn (1989) Figure A-16 Critical Shear Stress for Initiation of Suspension and Bed Load Transport as a Function of Particle Diameter Figure A-17 Schematic of Bed Model Structure for Bed Load Transport Newtown Creek RI/FS iii

6 LIST OF ACRONYMS AND ABBREVIATIONS Abbreviation Definition ADCP Acoustic Doppler Current Profiler AOC Administrative Order on Consent AQ Anchor QEA CERCLA Comprehensive Environmental Response, Compensation, and Liability Act cfs cubic feet per second cm centimeter CSM Conceptual Site Model CSO Combined Sewer Overflow EFDC Environmental Fluid Dynamics Code mm millimeter NYCDEP New York City Department of Environmental Protection RI/FS Remedial Investigation/Feasibility Study SNL Sandia National Laboratories SW Stormwater SWMM Stormwater Management Model TSS Total Suspended Sediment USEPA United States Environmental Protection Agency USGS United States Geological Survey µm micrometer Newtown Creek RI/FS iv

7 1 INTRODUCTION This memorandum presents the approach for developing and calibrating hydrodynamic and sediment transport models for the Newtown Creek Remedial Investigation/Feasibility Study (RI/FS) as described in the RI/FS Work Plan (AECOM 2011). This work is being performed under an Administrative Order on Consent (AOC) between the Respondents to the AOC and the United States Environmental Protection Agency (USEPA) in the USEPA Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) program. The AOC was entered into voluntarily by the USEPA and Phelps Dodge Refining Corporation, Texaco, Inc., BP Products North America Inc., The Brooklyn Union Gas Company d/b/a National Grid New York, ExxonMobil Oil Corporation, and The City of New York. The RI/FS Study Area is defined in the AOC as Newtown Creek and its tributaries (Dutch Kills, Maspeth Creek, Whale Creek, East Branch, and English Kills) having an approximate 3.8-mile reach to the high water mark. 1 The purpose of this memorandum is to provide a description of the hydrodynamic and sediment transport models that will be applied to Newtown Creek, including approaches for developing and calibrating these models. The memorandum is organized into three sections. Section 1 provides a general description of the overall modeling framework, along with the goals and objectives of the modeling study. Section 2 presents a description of the hydrodynamic and sediment transport models proposed for use in this study. The methods used to develop and calibrate the hydrodynamic and sediment transport models are discussed in Section 3. Due to characteristics of the urban watershed that surrounds Newtown Creek, 1 The Newtown Creek Superfund Site Study Area is described in the AOC as encompassing the body of water known as Newtown Creek, situated at the border of the boroughs of Brooklyn (Kings County) and Queens (Queens County) in the City of New York and the State of New York, roughly centered at the geographic coordinates of 40 42' north latitude ( ) and 73 55' west longitude ( ), having an approximate 3.8-mile reach, including Newtown Creek proper and its five branches (or tributaries) known respectively as Dutch Kills, Maspeth Creek, Whale Creek, East Branch, and English Kills, as well as the sediments below the water and the water column above the sediments, up to and including the landward edge of the shoreline, and also including any bulkheads or riprap containing the waterbody, except where no bulkhead or riprap exists, then the Study Area shall extend to the ordinary high water mark, as defined in 33 CFR 328(e) of Newtown Creek and the areal extent of the contamination from such area, but not including upland areas beyond the landward edge of the shoreline (notwithstanding that such upland areas may subsequently be identified as sources of contamination to the waterbody and its sediments or that such upland areas may be included within the scope of the Newtown Creek Superfund Site as listed pursuant to Section 105(a)(8) of the CERCLA). Newtown Creek RI/FS

8 Introduction a combined sewer overflow (CSO) and stormwater (SW) model will be used to simulate freshwater inflow to Newtown Creek. The CSO/SW model is also described and discussed in Sections 2 and Overview of Newtown Creek Physical Setting Newtown Creek and its five tributaries (Dutch Kills, Whale Creek, Maspeth Creek, East Branch, English Kills) comprise an estuarine waterbody with an overall length of about 3.8 miles. A substantial portion of the shoreline is contained by bulkheads or protected by riprap. The typical width of Newtown Creek is about 200 to 300 feet. Newtown Creek is a federally designated navigation channel with authorized project water depths of 20 feet or greater. Circulation within Newtown Creek is primarily affected by two mechanisms: tidal motion in the East River and freshwater inflows. The semi-diurnal tidal cycle has a vertical range of approximately 5 to 7 feet. Freshwater sources to Newtown Creek include CSOs, stormwater outfalls, municipal/industrial outfalls, and groundwater. Generally, fresher (less dense) water flows in a surface layer toward the East River, with saltier (more dense) water flowing upstream (away from the East River) in a bottom layer of the water column. 1.2 General Description of Modeling Framework The Newtown Creek modeling framework will consist of four sub-models that are linked together: 1) hydrodynamic; 2) sediment transport; 3) chemical fate and transport; and 4) bioaccumulation. The hydrodynamic model simulates the movement of water (e.g., current velocity, tidal elevation). The sediment transport model predicts the erosion, deposition and transport of solids (e.g., clay, silt, sand) in the system. The chemical fate and transport model will be used to predict temporal and spatial changes in: 1) particulate and dissolved chemical concentrations in the water column; and 2) particulate and porewater concentrations in the sediment bed. The bioaccumulation model will simulate temporal and spatial changes in chemical concentrations in benthic and water column biota. The information generated by each of the sub-models, along with the linkage between the models, is shown in Figure 1-1. The hydrodynamic and sediment transport models will produce the following information that will be used in the chemical fate and transport model: 1) tidal elevation (water depth); 2) current velocity; 3) vertical turbulent mixing Newtown Creek RI/FS

9 coefficients; 4) suspended sediment concentrations; 5) deposition fluxes of sediment; 6) erosion fluxes of sediment; and 7) net deposition rates of sediment. Introduction The first phase of developing the Newtown Creek modeling framework is focused on the hydrodynamic and sediment transport models, which are described in this memorandum. The second phase of the modeling study will entail the development, calibration, and application of the chemical fate and transport and bioaccumulation models. A memorandum describing the approach for developing those models is currently scheduled to be submitted to USEPA in summer Modeling Study Goals and Objectives The primary objective of the modeling study is to develop a reliable management tool that can be used to inform the conceptual site model (CSM) for the study area, support risk assessment, and evaluate the efficacy of remedial alternatives. The initial questions that the complete modeling framework (i.e., linked hydrodynamic, sediment transport, chemical fate and transport, and bioaccumulation models) will ultimately help answer include: What effects do chemical concentrations in the sediment bed have on total chemical concentrations in the water column and in biota? What is the annual load of chemicals from Newtown Creek to the East River under current conditions? What is the effect of chemical loads from the East River on chemical concentrations in the sediment bed and in biota? What are the effects of ongoing sources on chemical concentrations in the sediment bed and in biota? What are the effects of high-flow events or storms on chemical concentrations in the sediment bed, in the water column, and in biota? What is the potential for erosion, transport, and re-deposition of particle-associated chemicals in the sediment bed during high-flow events or storms at different locations within Newtown Creek? What is the rate of natural attenuation of chemical concentrations in the sediment bed and in biota under future conditions? Newtown Creek RI/FS

10 Introduction How would various remedial alternatives affect the rate of attenuation of chemical concentrations in the sediment bed and in biota? How would various remedial alternatives affect the annual load of chemicals from Newtown Creek to the East River? What is the potential for recontamination of remediated areas due to inputs from the East River and/or ongoing sources? The questions listed above will be answered after the chemical fate and transport and bioaccumulation models are developed and calibrated. Modeling Approach Memorandum (2), which will describe the development and calibration of the chemical fate and transport and bioaccumulation models, will be submitted to USEPA during The hydrodynamic and sediment transport models, which are the focus of this memorandum, will be used to answer questions related to sediment transport processes in Newtown Creek during multiyear periods and episodic high-flow events or storms, including: What areas are net depositional, net erosional, or in dynamic equilibrium? What is the net sedimentation rate in areas that are net depositional? What is the potential depth of scour during high-flow events or storms in areas that are net depositional, net erosional, or in dynamic equilibrium? Newtown Creek RI/FS

11 2 DESCRIPTION OF MODELING FRAMEWORK Hydrodynamics and sediment transport within Newtown Creek are influenced by: tidal exchange and mixing between the East River and Newtown Creek waters; freshwater inflows from CSOs and SW outfalls; and groundwater. Tidal exchange between the East River and Newtown Creek is characterized as semi-diurnal with a tidal range of approximately 5 to 7 feet (AECOM 2011), which, under typical tidal conditions, equates to an exchange of water of about 1,000 cubic feet per second (cfs; estimated using a simplified tidal box model of Newtown Creek). The main source of freshwater to Newtown Creek from the surrounding urban watershed is from CSOs and SW outfalls. Due to the complex nature of runoff from an urban watershed, a reliable method for predicting time-variable freshwater inflow from the CSO/SW outfalls is through use of a computer model. The InfoWorks CS software package, which was developed by Wallingford Software, has been used by the New York City Department of Environmental Protection (NYCDEP) to estimate freshwater inflow from CSO/SW outfalls and direct runoff from the Newtown Creek watershed. The total annual average discharge from CSO/SW outfalls to Newtown Creek is estimated to be about 9 cfs (NYCDEP 2007); freshwater inflow rates from CSO/SW outfalls are significantly higher than 9 cfs during precipitation events (i.e., wet weather conditions). Groundwater is a secondary source of freshwater to Newtown Creek. Based on the results of a regional groundwater model developed by the United States Geological Survey (USGS), the total annual average groundwater flow to Newtown Creek is about 0.3 cfs (Misut and Monti 1999), which is negligible compared to CSO/SW inflows and tidal exchange between Newtown Creek and the East River. Given the potential importance of freshwater inflows from the watershed, especially during precipitation events, a CSO/SW model (e.g., InfoWorks or similar) will be integrated into the hydrodynamic and sediment transport modeling framework. Groundwater inflow will not be included in the hydrodynamic modeling due to its negligible contribution to overall freshwater flow to the creek. This approximation is typically used in an estuarine hydrodynamic model. Even though groundwater inflow will not be incorporated into the hydrodynamic model, chemical loads from groundwater sources will be incorporated into the chemical fate and transport model as appropriate. The hydrodynamic and sediment Newtown Creek RI/FS

12 Description of Modeling Framework transport models of Newtown Creek will be linked together as shown in Figure 2-1. A summary of each model is provided below. 2.1 Combined Sewer Overflow/Stormwater Model Freshwater inputs from the watershed will be described using a CSO/SW model such as InfoWorks (or similar). InfoWorks is composed of hydrologic and hydraulic models that are applicable to urban watersheds. The Stormwater Management Model (SWMM) is incorporated into InfoWorks; SWMM is a runoff routing model that is capable of simulating flow through complex hydraulic structures (Rossman 2007). Output from the CSO/SW model will be used as input to the hydrodynamic model. The effects on freshwater inputs to Newtown Creek due to potential future changes in the operational characteristics of the Newtown Creek and Bowery Bay Water Pollution Control Plants will be incorporated into the CSO/SW model as needed. 2.2 Hydrodynamic Model The hydrodynamic model simulates the movement of water in Newtown Creek, and it accounts for the effects of the following factors on water movement: freshwater inflow from CSO/SW outfalls and direct runoff; tides; spatially variable bathymetry and geometry; and estuarine circulation resulting from density differences between seawater and freshwater, as well as temperature gradients in the water column. The hydrodynamic model is used to simulate temporal and spatial changes in water depth, current velocity, bed shear stress, salinity, and temperature. Newtown Creek is an estuarine system with freshwater inflows from CSO/SW outfalls and direct runoff, as well as other sources (e.g., municipal and industrial outfalls, groundwater). During high-flow events, large volumes of freshwater flow into Newtown Creek and density-driven circulation will occur. Thus, use of a three-dimensional hydrodynamic model is warranted, as is typically the case when modeling an estuarine system. However, it is unclear at this time whether or not density-driven circulation is a primary process in Newtown Creek. Current velocity and salinity data will be collected at five locations in Newtown Creek for a minimum of three months during The data from that field study Newtown Creek RI/FS

13 Description of Modeling Framework will be used to develop an improved understanding of density-driven circulation in the system. The hydrodynamic model that will be applied in this study is the Environmental Fluid Dynamics Code (EFDC), which is supported by USEPA. EFDC is a three-dimensional hydrodynamic model capable of simulating time-variable flow in rivers, lakes, reservoirs, estuaries, and coastal areas. The model solves the conservation of mass, momentum, energy (temperature), and salt equations, which are the fundamental equations governing the movement of water in an estuary. The effects of density-driven processes on circulation in an estuary are incorporated into EFDC. In addition, the model includes a sophisticated turbulence closure algorithm that simulates the effects of vertical turbulence on estuarine circulation. A characteristic of EFDC that is important for this study is the flooding-drying feature, which makes it possible to realistically simulate the flooding and drying of intertidal areas caused by tidal action in Newtown Creek. The model has been applied to a wide range of environmental studies in a large number of rivers, estuaries, and coastal ocean areas. A complete description of the model is given in Hamrick (1992). The primary hydrodynamic processes and external forcing functions incorporated into the three-dimensional hydrodynamic model are shown in Figure 2-2. Other three-dimensional hydrodynamic models that are widely used for simulating estuarine hydrodynamics include: 1) RMA-10 (Resource Management Associates); 3) MIKE 3 (Danish Hydraulic Institute); and 5) DELFT-3D (Deltares, The Netherlands). While these three models have various strengths and weaknesses, all of them have similar capabilities to EFDC and any of them might be adequate for simulating hydrodynamic circulation within Newtown Creek. The RMA-10 and MIKE 3 models are proprietary, so neither of these models could be used for this study, even though each model provides adequate technical capabilities. The DELFT-3D model recently became an open source code (i.e., January 2012). While DELFT-3D is a sophisticated hydrodynamic model, it is not linked to a sediment transport model that has the capability to utilize erosion rate data obtained from Sedflume testing (i.e., SEDZLJ capabilities). Water-column transport information (e.g., water depths, current velocities, turbulent diffusivity) from the hydrodynamic model will be used as input to the sediment transport Newtown Creek RI/FS

14 Description of Modeling Framework model. Current velocity information is used in the sediment transport model to calculate bed shear stress, which affects erosion and deposition processes. Episodic rare events will be simulated for two situations: 1) storm surge; or 2) high-flow event. Various tidal, atmospheric and flow conditions in the East River and other regions of New York Harbor and Long Island Sound can create a storm surge that enters at the mouth of Newtown Creek. An analysis of historical tide gauge data collected at the Battery will be conducted to determine the time history of water surface elevation to specify at the mouth of Newtown Creek for a storm surge simulation. High-flow events in Newtown Creek are primarily driven by CSO/SW inflows during storms with high rainfall. Thus, the CSO/SW model will be used to generate freshwater inflows for rare events (e.g., 100-year storm). The inflow hydrograph for a rare storm predicted by the CSO/SW model will be used as input to the hydrodynamic model. 2.3 Sediment Transport Model The sediment transport model is used to simulate temporal and spatial changes in: suspended sediment concentrations in the water column; bed elevation (i.e., bed scour depth, net sedimentation rate); sediment bed composition (i.e., relative amounts of clay, silt, and sand from different sources); and deposition and erosion fluxes across the sediment-water interface. The sediment transport model is capable of simulating the movement of sediment by suspended load (i.e., primarily clay, silt, fine sand) and bed load transport (i.e., near-bed movement of coarse sand and gravel). Bed load transport is the movement of sand and gravel in a thin layer (i.e., about 1 millimeter [mm] to 1 centimeter [cm] thick) just above the sediment surface. Based on experience at other sites and the characteristics of Newtown Creek, it is unlikely that bed load transport is a significant process in the Study Area. However, the importance of bed load transport will be evaluated during the development of the sediment transport model and a decision will be made as to whether or not to include bed load transport in the model. Note that many sediment transport modeling studies at contaminated sediment sites have been successfully completed without including bed load transport in the model (e.g., Upper Hudson River [New York], Grasse River [New York], Lower Duwamish Waterway [Washington], Lower Willamette River [Oregon]). Newtown Creek RI/FS

15 Description of Modeling Framework Mechanistic formulations and algorithms are used in the sediment transport model to simulate deposition and erosion of cohesive (muddy) and non-cohesive (sandy) sediment. The formulations and algorithms used to simulate deposition and erosion are based on empirical information and data from a wide range of laboratory and field studies. In addition, site-specific data will be used to determine various parameters for the sediment transport model, which provides additional constraints on the model. The sediment transport model that will be applied in this study is referred to as SEDZLJ, and is capable of simulating erosion and deposition of sediment within cohesive and noncohesive bed areas (Ziegler et al. 2000; Jones and Lick 2001; QEA 2008). A detailed description of the formulations used in and the structure of the sediment transport model are provided in Appendix A. The sediment transport model has the following characteristics and capabilities: 1) three-dimensional transport of suspended sediment in the water column; 2) use of Sedflume core data to specify erosion rate parameters; 3) specification of spatially variable bed properties, including vertical variations in erosion properties caused by consolidation effects; and 4) inclusion of a sediment bed model that tracks temporal changes in bed composition (i.e., sediment particle size, sediment source). Explicit simulation of morphodynamic updating of the sediment bed requires direct feedback between the hydrodynamic and sediment transport models (i.e., the two models are run in parallel). This approach will not be used because it would result in simulation times that are unacceptably long and make it difficult to meet the objectives of this study. The sediment transport model has the capability to use an algorithm that approximates the effects of bed elevation changes due to erosion and deposition on the hydrodynamics in the Study Area (i.e., current velocity and bed shear stress). The approximation approach, which does not affect simulation time, will be used in this study to estimate the effects of bed elevation change on water depth, current velocity, and bed shear stress. The approximation method for estimating feedback between the hydrodynamic and sediment models uses adjusted values of water depth (ha) and current velocity (Ua) as follows (Figure 2-3). First, use local conservation of mass to develop a relationship between the water depth and current velocity predicted by the hydrodynamic model (i.e., h and U) and the adjusted values: Newtown Creek RI/FS

16 Description of Modeling Framework Uh = U a h a (2-1) The adjusted water depth is calculated using: h a = h z (2-2) where z is bed elevation change calculated by the sediment transport model. Rearranging Equation 2-1 and solving for the adjusted current velocity: U a = (h/h a )U (2-3) Finally, the adjusted current velocity is used to calculate the adjusted bed shear stress: τ sf,a = ρ w C f U a 2 (2-4) where ρw is the density of water and Cf is the bottom friction coefficient (see Equation A-2 in Appendix A). If net deposition (bed aggradation) occurs (i.e., z greater than zero), then the bed shear stress increases. If net erosion (bed degradation) occurs (i.e., z less than zero), then the bed shear stress decreases. Thus, the approximation feedback method produces results that are qualitatively consistent with the effects of direct feedback between hydrodynamic and sediment transport processes in the river. Application of the approximation feedback method does not have any numerical limitations (e.g., maximum allowable bed elevation change per timestep) because timesteps in the sediment transport model are typically less than 1 minute and bed elevation changes less than 1 mm during a single timestep. Typically, a sediment transport model is used to predict the transport and fate of inorganic sediment; the transport and fate of organic solids are generally not simulated during a contaminated sediment modeling study. Simulating the transport of organic solids is problematic for several reasons: 1) specifying the organic solids content of external solids loads; 2) determining the settling characteristics of organic particles; 3) determining the fraction (content) of organic particles in the sediment bed, which is needed for calculating the erosion flux of organic solids; and 4) estimating water-column production of organic Newtown Creek RI/FS

17 Description of Modeling Framework solids. However, due to the potential importance of organic solids in this study, a review and analysis of existing data, used in conjunction with data collected during Phase 1 field studies, will be conducted to determine if organic solids transport needs to be incorporated into the sediment transport model, and if it does, how that can be accomplished. If organic solids are included in the modeling framework as a distinct class of sediment/solids, then the transport and fate of organic solids will be simulated by the sediment transport model, and not within the chemical fate and transport model. The transport of organic solids is implicitly simulated by SEDZLJ because organic solids in the sediment bed are lumped together with the clay/silt sediment class in the model. The transport of organic solids has not been explicitly simulated using SEDZLJ in previous modeling studies. If organic solids need to be incorporated into the model as a separate sediment/solids class, then a methodology will need to be developed to specify or determine: 1) amount of organic solids in external sediment loads; 2) content of organic solids in the sediment bed; and 3) settling characteristics of organic solids. It would also need to be determined whether or not water column production of organic solids needs to be incorporated into the model. If organic solids are included in the model, the data collected during Phase 2 field studies will be used to develop boundary condition inputs for the organics solids class. The version of EFDC used in this study has been modified by Anchor QEA to include the SEDZLJ sediment transport model and the AQ-FATE chemical fate and transport model. This modeling framework has been used at other contaminated sediment sites, including: 1) Upper Hudson River; 2) Lower Duwamish Waterway (Washington); 3) Lower Willamette River (Oregon); and 4) San Jacinto River estuary (Texas). The model is non-proprietary and has been made available to USEPA for review (i.e., source code and input/output files) during modeling studies at other sites. The SEDZLJ model used in this study was developed at Anchor QEA (SEDZLJ-AQ) and is based on the original version of SEDZLJ developed by Jones and Lick (2001). The Jones-Lick version of SEDZLJ was incorporated into EFDC by researchers at Sandia National Laboratories (SNL; Thanh et. al. 2008). The primary differences between the SEDZLJ models incorporated into the Anchor QEA and SNL versions of EFDC are: Newtown Creek RI/FS

18 Description of Modeling Framework SEDZLJ-AQ separates the active layer of the bed model into two sub-layers (i.e., active-surface and active-buffer layers), see Appendix A. Erosion rate parameters are vertically constant within each layer of the bed model in SEDZLJ-AQ, with discontinuities in parameter values occurring at the interface between two adjacent layers. In SEDZLJ-SNL, erosion rate parameters are assumed to decrease at a logarithmic rate with increasing depth within each layer of the bed model. The calculation of erosion flux of class k sediment in SEDZLJ-SNL (Jones and Lick 2001) has been modified to include a particle shielding factor in SEDZLJ-AQ, see Equations A-15 and A-20 in Appendix A. SEDZLJ-AQ has the capability to track sediment originating from different locations in the study area or from different external sources. Newtown Creek RI/FS

19 3 MODEL DEVELOPMENT AND CALIBRATION Development and calibration of the hydrodynamic and sediment transport models will be accomplished using a phased approach. The first phase will include development and calibration of the CSO/SW and hydrodynamic models. These two models will be developed in parallel, with calibration of the CSO/SW model being completed prior to starting the hydrodynamic model calibration process. The second phase of this study will focus on development and calibration of the sediment transport model. As discussed in Section 1.1, a chemical fate and transport model will be linked to the hydrodynamic and sediment transport models. The possibility exists during the development and calibration of the chemical fate and transport that adjustments will need to be made to the hydrodynamic and sediment transport models in order to improve the performance of the chemical fate and transport model. This iterative process implies that the calibration of the CSO/SW, hydrodynamic and sediment transport models discussed below may not necessarily be finalized during this phase of the modeling study. The calibration of those three models will be finalized either before or during the final phase of the modeling study (i.e., chemical fate and transport and bioaccumulation modeling). Prior to starting the model development and calibration process, a data gaps analysis will be conducted to determine if additional hydrodynamic and sediment transport data, beyond what is being collected during the Phase 1 field studies, needs to be obtained. If additional data needs are identified, then field studies will be proposed to be conducted during the Phase 2 data collection efforts. 3.1 Development and Calibration of CSO/SW Model Data Requirements Developing a CSO/SW model of Newtown Creek requires the following data and information for creating inputs to the model: Precipitation and meteorological data Delineation of sub-drainage area Physical characteristics of sub-drainage areas Newtown Creek RI/FS

20 Model Development and Calibration Infrastructure as-built information for separate and combined piping systems Pumping station locations and related information Design/operation information for CSO regulators Calibration and validation of the CSO/SW model will use these types of data: Flow rates within piping system and at discharge locations Pump station flow rates Water level data in manholes or other infrastructure Model Development The CSO/SW model will be developed using information and data from previous NYCDEP CSO/SW modeling studies conducted for the Newtown Creek watershed (NYCDEP 2007) where applicable. Inputs to the existing NYCDEP models will be revised and expanded as needed to simulate freshwater inflow to Newtown Creek from the surrounding urban watershed during dry and wet weather conditions. Data and information collected on CSO/SW outfalls during 2012, as well as other discharges to the creek, will be incorporated in the CSO/SW model Model Calibration and Validation The strategy for calibrating and validating the CSO/SW model will be similar to the methodology used for the NYCDEP model (NYCDEP 2007). The primary model parameters that may be adjusted during calibration are: percentage of impervious and pervious areas in the watershed; catchment slope; overland catchment width; infiltration rate; depression storage; and pipe roughness. Calibration and validation of the model will involve comparisons of predicted and measured values of: 1) total runoff volume; 2) pipe water depth; and 3) velocity. Model performance will be evaluated during dry and wet weather conditions. Calibration and validation data are available for March 2005 through January 2006, with a number of precipitation events of varying duration and intensity occurring during this 11-month period. The data collected during the 11-month period will be separated into two sub-periods, with one data set being used for model calibration (e.g., 6- month period from March 2005 through August 2005) and the other data set used for validation (e.g., 5-month period from September 2005 through January 2006). Newtown Creek RI/FS

21 Model Development and Calibration Sensitivity Analysis An evaluation of the sensitivity of CSO/SW model predictions to parameters adjusted during the calibration process will be conducted. The first step in the sensitivity analysis will be to determine lower- and upper-bound values of each calibration parameter. The second step will be to repeat the calibration and validation simulations using the lower- and upperbound parameter values. For example, if three parameters (i.e., parameter A, B, C) were adjusted during calibration, then a total of six sensitivity simulations would be conducted (i.e., A-low, A-high, B-low, B-high, C-low, C-high). The final step will be to quantitatively and qualitatively analyze the results of the sensitivity simulations. 3.2 Development and Calibration of Hydrodynamic Model Data Requirements Development of inputs for the hydrodynamic model will need these types of data and information: Bathymetry and geometry Water surface elevation, salinity, and temperature at the East River boundary Meteorological data for specifying heat flux at the water surface Time-variable wind speed and direction Freshwater inflows (predicted by CSO/SW model) Data sources for model inputs are discussed in Section Calibration and validation of the hydrodynamic model will require the following data: Current velocities (magnitude and direction) Water surface elevation Salinity Temperature These data will be obtained during a current meter study that will be conducted as part of the RI/FS (Anchor QEA 2011). Acoustic Doppler current profiles (ADCPs) will be deployed Newtown Creek RI/FS

22 Model Development and Calibration at five locations for a period of at least 3 months: 1) mouth of Newtown Creek; 2) immediately downstream of junction with Dutch Kills; 3) mouth of Maspeth Creek; 4) mouth of East Branch; and 5) mouth of English Kills. In addition to current velocity measurements, water surface elevation, temperature, and salinity data will be collected at the five locations. A complete description of the current meter study is provided in Anchor QEA (2011) Model Development The hydrodynamic model domain will include all of Newtown Creek, with the boundary of the model located at the confluence of the creek with the East River (Figure 3-1). It is envisioned that a curvilinear, boundary-fitted numerical grid will be generated to represent the geometry and bathymetry of Newtown Creek in the model. A preliminary numerical grid will be generated using about 1,000 to 1,500 grid cells in the horizontal plane and ten layers in the vertical direction. Initial testing of the model will be conducted using this numerical grid to determine simulation times and qualitatively evaluate model performance (i.e., ensure that the hydrodynamic model is producing realistic results). The numerical grid will be refined and modified as needed during this initial testing period. The sediment transport and chemical fate and transport models will use the same numerical grid as the hydrodynamic model. The effects of the in-channel aeration system located in English Kills on water column mixing processes (i.e., turbulent mixing) will be incorporated into the hydrodynamic model. The aeration system may cause the water column to be vertically well mixed (i.e., vertical profiles of temperature and salinity are approximately uniform). Thus, vertical turbulent diffusion coefficients in the model will be increased to achieve a well-mixed water column (e.g., vertically uniform salinity and temperature predicted by the model) in the portions of English Kills where the aeration system is operating. Temporal variations in the operation of the in-channel aeration system will be included in the model as needed. If aeration systems are installed in other portions of Newtown Creek in the future, then the effects of those systems will be incorporated in the model. Newtown Creek RI/FS

23 Model Development and Calibration Bathymetry and geometry information are available from two sources: 1) existing data; and 2) planned surveys. Existing data includes the following hydrographic surveys: Laurel Hill OU6 bathymetric survey (2003); NYCDEP bathymetric survey for select portions of Newtown Creek (2005/2006); NYCDEP maintenance dredging surveys (2008); and USACE controlling depth reports and surveys (2009). A single-beam bathymetry survey of Newtown Creek was conducted during A complete description of this bathymetry survey is provided in Anchor QEA (2011). Data from this survey will be used to define the bathymetry and geometry of the creek in the model. The time-variable water surface elevation at the model boundary at the mouth of Newtown Creek (i.e., confluence with the East River) can be specified using two different methods: 1) it can be estimated using data collected at tidal gauging stations located in New York Harbor; or 2) it can be predicted by a far-field hydrodynamic model. The empirical approach is preferred for this study because it provides more flexibility for conducting multiyear simulations due to the long historical record of tide data in the New York Harbor area. Meteorological and wind speed/direction data will be obtained from weather stations located in the vicinity of the Study Area. Output from the CSO/SW model will be used as input to the hydrodynamic model. Time variable freshwater inflows to Newtown Creek from CSO and SW outfalls will be predicted during dry and wet weather conditions on either an hourly or daily basis (e.g., hourly output during wet weather conditions). The time series of freshwater inflows that is output from the CSO/SW (InfoWorks) model will be used as boundary condition input to the hydrodynamic model Model Calibration and Validation Calibration and validation of the hydrodynamic model will involve comparisons of predicted and measured values of: 1) current velocity; 2) water surface elevation; 3) temperature; and 4) salinity. The capability of the model to simulate temporal and spatial variations in current velocity, water surface elevation, temperature, and salinity will be evaluated using data collected at the five current meter locations discussed in Section Model performance will be evaluated over a range of freshwater inflow and tidal conditions. The calibration and Newtown Creek RI/FS

24 Model Development and Calibration validation periods will each be at least 1 month in length. A review of the data collected during the 3-month current meter deployment period (minimum length) will be conducted to determine the calibration and validation periods. During the calibration process, model input parameters (e.g., effective bed roughness) will be adjusted so that the agreement between predicted and observed quantities is optimized. After the model is calibrated, the validation period will be simulated with no adjustment of input parameters Sensitivity Analysis The sensitivity of hydrodynamic model predictions to calibration parameters will be evaluated. First, lower- and upper-bound values of each calibration parameter will be determined. The calibration and validation simulations will be repeated using the lowerand upper-bound parameter values. For example, if two parameters (i.e., parameter A, B) were adjusted during calibration, then a total of four sensitivity simulations would be conducted (i.e., A-low, A-high, B-low, B-high). The final step will be to quantitatively and qualitatively analyze the results of the sensitivity simulations. 3.3 Development and Calibration of Sediment Transport Model Data Requirements Developing a sediment transport model of Newtown Creek requires the following data and information for creating inputs to the model: Magnitude and composition of sediment loads from the CSO/SW outfalls Magnitude and composition of sediment loads from the East River Bulk bed properties, including grain size distribution and dry density Delineation of cohesive (muddy) and non-cohesive (sandy) bed areas Erosion properties of cohesive bed sediment (Sedflume core testing data) These inputs will be developed using existing data and information in conjunction with data collected during Phase 1 field studies (Anchor QEA 2011). Calibration and validation of the sediment transport model will use these types of data: Net sedimentation rates Newtown Creek RI/FS

25 Model Development and Calibration TSS concentrations Bed elevation change (bathymetry data) Net sedimentation rates will be determined from an analysis of radioisotope data obtained from sediment cores collected during the geochronology study (Anchor QEA 2011). TSS concentration data will be collected at 15 sampling locations within Newtown Creek over a wide range of flow and tidal conditions (Anchor QEA 2011). Bed elevation change data may be available from bathymetry surveys conducted within Newtown Creek at different times in the past. This type of data will only be used for model calibration or validation if reliable estimates of bed elevation change can be derived from the bathymetry data sets Model Development The two primary challenges for developing and applying a sediment transport model of Newtown Creek are: Specifying the spatial distribution of bed properties Estimating the magnitude and composition of external sediment loads Specific inputs that need to be determined for the sediment transport model are: Bed map, which delineates areas of cohesive and non-cohesive sediment Effective particle diameters of sediment size classes (i.e., clay/silt, sand, gravel) Spatial distribution of median particle diameter (D50) and initial composition of bed sediment Spatial distribution of erosion rate parameters for cohesive sediment Boundary condition: magnitude and composition of sediment loads from freshwater inflows Boundary condition: magnitude and composition of sediment load from the East River Delineating areas of cohesive and non-cohesive sediment throughout Newtown Creek is the first step in model development. This delineation is necessary because the erosion properties of these two types of sediment are significantly different. Results of the side-scan sonar survey of the creek (Anchor QEA 2011) will be used to delineate cohesive and non-cohesive Newtown Creek RI/FS

26 Model Development and Calibration bed areas. The erosion properties of cohesive sediments will be determined from an analysis of erosion rate data collected from five sediment cores during the Sedflume study (Anchor QEA 2011). The erosion properties of non-cohesive sediment depend on: 1) median particle diameter (D50); 2) bed composition (i.e., relative amounts of clay/silt, sand and gravel); and 3) dry density. Within non-cohesive bed areas, specification of the spatial distributions of D50 and bed composition (i.e., clay/silt/sand/gravel content) is necessary for model simulations. Grain size distribution and dry density data obtained from surficial sediment samples collected within Newtown Creek (Anchor QEA 2011) will be used to determine the spatial distribution of bed properties. Within the cohesive bed area, spatial variations in erosion properties (vertical and horizontal) and bed composition need to be incorporated into the model. External sediment loads have a primary controlling effect on net sedimentation rates within the system. The composition of the incoming loads (i.e., relative amounts of clay/silt/sand) is as important as the load magnitude. External sediment loads originate from two sources: 1) discharge from CSO/SW outfalls; and 2) tidal transport from the East River at the downstream boundary of the model (Figure 3-1). High-flow events are the focus of a sediment load study because, typically, a majority of the annual load occurs during a small number of high-flow events. A review of existing sediment load data and information will be conducted. The existing data will be combined with total suspended sediment (TSS) concentration data collected during various surface water sampling studies to be conducted within Newtown Creek; see Section 10 of Anchor QEA (2011). The sediment bed in Newtown Creek is composed of sediment particles that range from clay to gravel. Simulation of the entire particle size spectrum is impractical for several reasons: simulation times and array-storage requirements increase with each particle-size class that is added; limitations in grain size distribution data for the sediment bed make it difficult to specify initial conditions for the entire spectrum; and sparse data for the composition of the external sediment load make it problematic to specify this boundary condition for the entire spectrum. Therefore, particles will be separated into four size classes: 1) clay and silt with particle diameters less than 62 micrometers (µm); 2) fine sand (62 to 250 µm); 3) medium and coarse sand (250 to 2,000 µm); and 4) gravel (greater than 2,000 µm). Grain size distribution data collected from the Study Area will be used to estimate the effective particle diameters of Newtown Creek RI/FS

27 Model Development and Calibration the four sediment size classes. The selection of these four sediment size classes was based on experience gained from previous modeling studies at contaminated sediment sites (e.g., QEA 2008), where use of these four size classes has produced reliable models that met the study objectives. The clay/silt component is represented as a single size class because: 1) clay/silt particles suspended in the water column flocculate and are not transported as discrete particles; 2) the erosion rates of different particle sizes in the clay/silt size range cannot be measured; and 3) the particle size distribution of clay/silt sediment in the incoming sediment load cannot be estimated Model Calibration and Validation The primary calibration target for the sediment transport model will be net sedimentation rate within the creek. A multi-year simulation (e.g., 10 to 30 years in length) will be conducted, and predicted net sedimentation rates will be compared to observed values. Comparisons of predicted and measured TSS concentrations at multiple locations within Newtown Creek will also be used to evaluate model performance. An iterative process will be used during model calibration to optimize agreement between predicted and observed values while adjusting a minimum number (i.e., three or less) of model parameters and inputs. Additional validation of the predictive capabilities of the model will be conducted if reliable bed elevation change data are available Sensitivity Analysis Sensitivity and diagnostic analyses will be conducted to evaluate model reliability, both quantitatively and qualitatively. The sensitivity of model predictions to uncertainty in key model inputs will be evaluated. Diagnostic analyses will be used to ensure that the model is performing in a realistic manner over a wide range of flow and tidal conditions. In addition to evaluating sediment transport processes in Newtown Creek over a multi-year period, the model will be used to evaluate the effects of rare storms in the Study Area. As discussed in Section 2.3, direct feedback between the hydrodynamic and sediment transport models is not a viable option for this study and an approximation method will be used to estimate feedback between the two models. A sensitivity simulation will be conducted to evaluate potential differences on model predictions of using the approximation Newtown Creek RI/FS

28 Model Development and Calibration feedback method. A 3-month period will be selected from the multi-year calibration simulation, during which the approximation feedback method was used. That 3-month period will be simulated using the direct feedback method (i.e., hydrodynamic and sediment transport models will be run in parallel). Quantitative comparisons of the direct and approximation feedback methods will be made using results from the two simulations of the 3-month period. The approximation feedback method will be used to conduct the multi-year simulation and bed elevations, and water depths, during that multi-year period will be predicted to change. Over the course of a multi-year period, changes in bed elevation and water depth may become large enough that the approximation feedback method begins to become inaccurate. A sensitivity analysis will be conducted to determine under what conditions the approximation feedback method does not produce accurate results. Once those conditions are established, the results of the multi-year simulation will be examined to determine if the approximation feedback method has produced inaccurate results. If the approximation feedback method does produce inaccurate results at some point during the multi-year simulation, that method will be modified such that the model predictions have acceptable accuracy. As discussed in Section 3.3.2, four size classes will be used in the sediment transport model: 1) clay and silt; 2) fine sand; 3) medium and coarse sand; and 4) gravel. The sensitivity of model predictions to the number of sediment size classes will be investigated by conducting a simulation with five size classes: 1) clay and silt (less than 62 m); 2) fine sand (62 to 250 m); 3) medium sand (250 to 500 m); 4) coarse sand (500 to 2,000 m); and 5) gravel (greater than 2,000 m). A 1-year period will be selected from the multi-year calibration simulation and that 1-year period will be repeated using the five size classes discussed above. Comparisons of the 4-class and 5-class simulations of the 1-year period will be made so as to quantitatively evaluate the potential effects of number of size classes on model predictions. Newtown Creek RI/FS

29 4 REFERENCES AECOM, Remedial Investigation/Feasibility Study Work Plan. June. Anchor QEA, Field Sampling and Analysis Plan, Remedial Investigation/Feasibility Study, Newtown Creek. July Cheng, N.S., Simplified settling velocity formula for sediment particle. ASCE J Hydr Engr. 123(2): Garcia, M. and G. Parker, Entrainment of bed sediment into suspension. ASCE J Hydr Engr, 117(4): Gessler, J., The beginning of bedload movement of mixtures investigated as natural armoring in channels. W.M. Keck Laboratory of Hydraulics and Water Resources, California Institute of Technology, Translation T-5. Guy, H.P., D.B. Simons, and E.V. Richardson, Summary of alluvial channel data from flume experiments, U.S. Geological Survey Professional Paper 462-I. Washington, D.C. Hamrick, J.M., A Three-Dimensional Environmental Fluid Dynamics Computer Code: Theoretical and Computational Aspects. College of William and Mary, Virginia Institute of Marine Sciences. Special Report pp. Jepsen R, J. Roberts, and W. Lick, Effects of bulk density on sediment erosion rates. Water Air Soil Pollution, 99: Jones, C.A. and W. Lick, SEDZLJ: A Sediment Transport Model. Final report. University of California, Santa Barbara, California. May 29, Karim, M.F. and J.F. Kennedy, Computer-based predictors for sediment discharge and friction factor of alluvial streams, IIHR report no Univ. of Iowa, Iowa City, Iowa. Misut, P.E. and J. Monti, Simulation of Ground-water Flow and Pumpage in Kings and Queens Counties, Long Island, New York. U.S. Geological Survey Water-Resources Investigation Report p. New York City Department of Environmental Protection (NYCDEP), Landside Modeling Report, Volume 6, Newtown Creek WPCP, Final. City of New York Newtown Creek RI/FS

30 References Department of Environmental Protection Bureau of Engineering Design and Construction, Draft. October Parker, G., D Sediment Transport Morphodynamics, with Applications to Rivers and Turbidity Currents. E-book located at: Quantitative Environmental Analysis, LLC (QEA), Lower Duwamish Waterway Sediment Transport Modeling Report, Draft Final. Prepared for USEPA, Region 10 and Washington State Dept. of Ecology. June Rahuel, J.L., F.M. Holly, J.P. Chollet, P.J. Belleudy, and G. Yang, Modeling of riverbed evolution for bedload sediment mixtures. ASCE J Hydr Engr, 115(11): Roberts J., R. Jepsen, and W. Lick, Effects of particle size and bulk density on the erosion of quartz particles. ASCE J Hydr Engr 124(12): Rossman, L.A., Stormwater Management Model User s Manual, Version 5.0. U.S. Environmental Protection Agency. EPA/600/R-05/040. June Thanh, P.H.X., M.D. Grace, and S.C. James, Sandia National Laboratories Environmental Fluid Dynamics Code: Sediment Transport User Manual. Sandia Report SAND , 47 pages, September Van den Berg, J.H. and A. van Gelder, Prediction of suspended bed material transport in flows over silt and very fine sand. Water Resour. Res., 29(5): Van Rijn, L.C., 1984a. Sediment transport, part I: bed load transport. ASCE J Hydr Engr. 110(10): Van Rijn, L.C., 1984b. Sediment transport, part II: suspended load transport. ASCE J Hydr Engr. 110(11): Van Rijn, L.C., 1984c. Sediment transport, part III: bed forms and alluvial roughness. ASCE J Hydr Engr. 110(12): Van Rijn, L.C., Principles of Sediment Transport in Rivers, Estuaries and Coastal Seas. Aqua Publications. Delft, The Netherlands. Van Rijn, L.C., H. van Rossum, and P. Termes, Field verification of 2-D and 3-D suspended-sediment models. ASCE J Hydr Engr, 116(10): Newtown Creek RI/FS

31 References Voogt, L., L.C. van Rijn, and J.H. van den Berg, Sediment transport of fine sands at high velocities. ASCE J Hydr Engr, 117(7): Wright, S., and G. Parker, Flow resistance and suspended load in sand-bed rivers: simplified stratification model. ASCE J. Hydr. Engr. 130(8): Ziegler, C.K. and W. Lick, The transport of fine-grained sediments in shallow waters. Environmental Geology and Water Sciences 11: Ziegler, C.K, P.H. Israelsson, and J.P. Connolly, Modeling sediment transport dynamics in Thompson Island Pool, Upper Hudson River. Water Quality and Ecosystem Modeling 1:19 Newtown Creek RI/FS

32 FIGURES

33 Hydrodynamic Model Information Transfer Sediment Transport Model Chemical Fate & Transport Model Bioaccumulation Model Tidal Elevation Current Velocity TSS Concentration Particulate & Dissolved Concentrations Biota Concentrations vkl - \\montd-vlist\d_drive\newtown_creek\documents\newtown_creek_figure_1-1.pptx Salinity Temperature Deposition Erosion Net Sedimentation Rate Bed & Porewater Concentrations Biota Concentrations Figure 1-1 Schematic of Newtown Creek Modeling Framework Modeling Approach Memorandum (1) Newtown Creek RI/FS

34 CSO/Stormwater Model Freshwater Inflow Hydrodynamic Model Information Transfer Sediment Transport Model Tidal Elevation Current Velocity TSS Concentration vkl - \\montd-vlist\d_drive\newtown_creek\documents\newtown_creek_figure_2-1.pptx Salinity Temperature Deposition Erosion Net Sedimentation Rate Figure 2-1 Schematic of Newtown Creek Hydrodynamic and Sediment Transport Models Modeling Approach Memorandum (1) Newtown Creek RI/FS

35 Tidal Elevation Wind surface stress Surface heat flux Freshwater Inflows Fresher, less dense water ckz/btr- \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_2-2.pptx Exchange with East River Vertical profiles of: Current velocity Salinity Temperature Saltier, more dense water Vertical turbulent mixing Bed shear stress Figure 2-2 Primary Processes and External Forcing Incorporated into 3-D Hydrodynamic Model Modeling Approach Memorandum (1) Newtown Creek RI/FS

36 Hydrodynamic Model Sediment Transport Model Water velocity: u Adjusted water velocity: u a eg/btr- \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_2-3.pptx Water depth: h Bed shear stress: τ Adjusted water depth: h a Adjusted bed shear stress: τ a Bed elevation change due to erosion/deposition: z Figure 2-3 Schematic of Approximation Method for Estimating Feedback Between Hydrodynamic and Sediment Transport Models Modeling Approach Memorandum (1) Newtown Creek RI/FS

37 Vernon Blvd 48th Ave Jackson Ave 21st St 11th St 49th Ave Skillman Ave DUTCH KILLS Hunters Point Ave Van Dam St 495 Greenpoint Ave 39th St 51st Ave Calvary Cemetery Queens Midtown Expy 58th St Maurice Ave Maspeth 55th Dr Ave LEGEND Newtown Creek Study Area Waterbody Open Space 278 NEWTOWN CREEK MASPETH CREEK WHALE CREEK Kingsland Ave Bridgewater St Newtown Creek Study Area Grand Ave \\Montd-vlist\D_Drive\Newtown_Creek\GIS\Newtown_Figure_3-1_Model_Extent.mxd qea_user 11/11/2011 2:32:28 PM EAST EAST RIVER RIVER Mcguinness Blvd Tidal Boundary at Mouth of Newtown Creek DRAFT Berry St Nassau Ave Bedford Ave Norman Ave Manhattan Driggs Ave Ave Union Ave Humboldt St Mcguinness Blvd S Meeker Ave Apollo St Lombardy St Grandparents Ave Maspeth Ave Vandervoort Ave Bushwick Ave Metropolitan Ave ENGLISH KILLS Grand St EAST BRANCH ,000 1,500 Feet Queens Manhattan Brooklyn Figure 3-1 Spatial Extent of Model Domain Modeling Approach Memorandum (1) Newtown Creek RI/FS

38 APPENDIX A DETAILS OF SEDIMENT TRANSPORT MODEL THEORY AND FORMULATION

39 APPENDIX A DETAILS OF SEDIMENT TRANSPORT MODEL THEORY AND FORMULATION This appendix presents discussion of the theory and formulations used in the sediment transport model to calculate erosion and deposition fluxes at the sediment-water interface. For bed scour, erosion fluxes in cohesive and non-cohesive bed areas are treated differently. In a grid cell specified as hard bottom, the erosion and deposition fluxes are set to zero, so no change in bed elevation is calculated during a simulation. A.1 CALCULATION OF BED SHEAR STRESS Erosion rate is dependent on bed shear stress, which is calculated using near-bed current velocity predicted by the hydrodynamic model. The bed shear stress calculated within the hydrodynamic model is the total bed shear stress, which represents the total drag on the water column by the sediment bed. The total bed shear stress (τtot) is the sum of shear stresses associated with skin friction (τsf) and form drag (τfd): τtot = τsf + τfd (A-1) Skin friction represents the shear stress generated by sediment particles (i.e., small-scale physical features), whereas form drag corresponds to the drag generated by bedforms (e.g., ripples, dunes) and other large-scale physical features. For a cohesive bed, skin friction is considered the dominant component of the bed shear stress for most applications. When simulating the erosion of a non-cohesive bed, it is a reasonable approximation, and a standard approach, to use the skin friction component and neglect form drag for calculating bed shear stress for a non-cohesive bed. This approach is consistent with accepted sediment transport theory (Parker 2004). Skin friction shear stress is calculated using the quadratic stress law: τsf = ρw Cf u 2 (A-2) Newtown Creek RI/FS A

40 Appendix A where ρw is the density of water, Cf is the bottom friction coefficient, and u is the near-bed current velocity (i.e., predicted velocity in the bottom layer of the numerical grid). Use of the near-bed current velocity is standard practice for calculating bed shear stress in a threedimensional model. The bottom friction coefficient is determined using (Parker 2004): Cf = κ 2 ln -2 (11 zref /ks) (A-3) where zref is a reference height above the sediment bed, ks is the effective bed roughness, and κ is von Karman s constant (0.4). The reference height (zref) is spatially and temporally variable because it is equal to half of the thickness of the bottom layer of the numerical grid. Because a stretched (sigma-layer) grid is used in the vertical direction, the thickness of the bottom layer of the vertical grid varies due to changes in tidal elevation and river flow rate. Thus, the reference height properly incorporates temporal and spatial variations in water depth into the calculation of the bottom friction coefficient. The effective bed roughness is assumed to be proportional to the D90 of the surface sediment layer (Parker 2004; Wright and Parker 2004): ks = 2D90 (A-4) Grain size distribution data will be used to specify D90 values for the surface layer of Newtown Creek sediments. The validity of the above approach for calculating the bottom friction coefficient is evaluated as follows. Bottom friction coefficients were calculated over a range of water depths using an average D90 value of 3,000 microns ([µm]; see Table A-1). The range of bottom friction coefficient values in Table A-4 is consistent with expected values (van Rijn 1993). This approach provides an objective method for estimating the effective bed roughness, which will decrease the uncertainty associated with subjective estimates of roughness. Newtown Creek RI/FS A

41 Appendix A Table A-1 Bottom Friction Coefficient Values for a Range of Water Depths Water Depth Bottom Coefficient: (feet) D 90 =3,000 µm For use in formulations presented below, a demonstrated accurate equation for bed-shear velocity (u*) is defined as (van Rijn 1993): u* = (τsf /ρw) 0.5 (A-5) Current velocity in turbulent flow, which exists in Newtown Creek for all flow and tidal conditions, is the sum of two components: time-averaged mean velocity and turbulent fluctuations about the mean value. The bed-shear velocity (u*) corresponds to the turbulentfluctuation component of the current velocity. Thus, the skin friction shear stress is driven by the turbulent fluctuations in the flow, which are randomly variable with time. Random variation in turbulence along the sediment bed is the primary reason that a probabilistic approach to calculating deposition and erosion fluxes is necessary; use of probability of deposition (see Equation A-6) and suspension (see Equation A-15) formulations have been incorporated into the model to account for these turbulence effects. A.2 DEPOSITION PROCESSES The deposition flux for size class k sediment (Dk) is expressed as (Ziegler et al. 2000): Dk = Pdep,k Ws,k Ck (A-6) Newtown Creek RI/FS A

42 Appendix A where Pdep,k is probability of deposition of class k, Ws,k is settling speed of class k, and Ck is near-bed suspended sediment concentration of class k. Deposition flux has units of mass per unit area per time (e.g., g/cm 2 s). The near-bed concentration (Ck) is calculated using the sediment transport model and is represented by the value in the vertical grid cell immediately above the bed. Probability of deposition of cohesive sediment (i.e., clay/silt) is determined using the Krone formulation (van Rijn 1993): Pdep,k = 1 (τsf/τcr,dep) for τsf < τcr,dep (A-7) = 0 for τsf > τcr,dep (A-8) where τsf is bed shear stress (skin friction) and τcr,dep is the critical bed shear stress for deposition. The relationship between probability of deposition and bed shear stress for cohesive sediment is shown in Figure A-1. For non-cohesive sediment (i.e., sand and gravel), the probability of deposition depends on bed shear stress and particle diameter, and is described by a Gaussian distribution (Gessler 1967; Ziegler et al. 2000): Pdep,k = (2π) -0.5 EXP(-0.5x 2 ) dx (A-9) where the lower and upper limits of the integral are negative infinity and Y, respectively, and EXP corresponds to the exponential function with base e. The parameter Y is given by: Newtown Creek RI/FS A

43 Appendix A Y = 1.75 (τc,k /τsf - 1) (A-10) where τc,k is critical shear stress for suspension of class k sediment, which is: τc,k = ρw u*,crs,k 2 (A-11) where u*,crs,k is critical bed-shear velocity for initiation of suspension for class k: u*,crs,k = 4 Ws,k /d*,k for 1 < d*,k < 10 (A-12) = 0.4 Ws,k for d*,k > 10 and: d*,k = dk [(s-1)g/ν 2 ] 1/3 (A-13) where dk is particle diameter for class k, s is specific density of particle (i.e., 2.65), g is acceleration caused by gravity, and ν is kinematic viscosity of water. The non-dimensional particle parameter (d*,k) is commonly used in a wide range of sediment transport formulations (van Rijn 1993). The probability of deposition for 130 and 540 µm particles (i.e., fine and medium/coarse sand) as a function of bed shear stress and particle diameter is presented in Figure A-2. Numerous field and laboratory experiments have demonstrated that a physically realistic representation of the settling speed of a discrete particle is related to the particle diameter, representing size class k, as follows (Cheng 1997): Ws,k = (ν/dk) [( d*,k 2 ) 0.5 5] 1.5 (A-14) Newtown Creek RI/FS A

44 Appendix A The dependence of settling speed on particle diameter is shown in Figure A-3. A.3 EROSION PROCESSES: COHESIVE BED Within sediment bed areas designated as cohesive, the following numerical algorithm is used to calculate the erosion flux of sediment from the bed to the water column, where it is transported as suspended sediment. The erosion flux for size class k sediment (Ek) from a cohesive bed is given by: Ek = ρdry fas,k Sk Psus,k Egross (A-15) where Egross is the gross erosion rate, Psus,k probability of suspension for size class k, Sk is the particle-shielding factor for size class k, ρdry is dry density of bed sediment, and fas,k is the fraction of size class k sediment in the active-surface layer. Erosion flux has units of mass per unit area per time (e.g., g/cm 2 s). Erosion of a sediment bed depends on a number of factors, including, but not limited to: shear stress, grain size distribution, wet (bulk) density, TOC content, and gas content (Jepsen et al. 1997; Roberts et al. 1998). Factors such as TOC content, gas content and bioturbation are implicitly incorporated into the cohesive erosion algorithm through the use of sitespecific erosion rate data (i.e., Sedflume core data). The rate at which sediment is removed from the consolidated sediment bed and transported to a thin near-bed layer that exists between the consolidated sediment bed and the water column is termed the gross erosion rate (Egross). Some of the eroded sediment in the near-bed layer is re-deposited to the consolidated bed; the rate of re-deposition is referred to as the gross deposition rate (Dgross). The remainder of the eroded material in the near-bed layer is transported to the water column; this rate is referred to as the net erosion rate (Enet). The near-bed layer discussed above is incorporated into a model of the sediment bed, which is described below. Two parameters that affect Egross are shear stress and wet density (Jepsen et al. 1997). An evaluation of Sedflume data collected at other contaminated sediment sites indicate that Newtown Creek RI/FS A

45 Appendix A minimal correlation exists between wet density and erosion rate. Thus, it will be assumed in this study that erosion rate is dependent on skin friction shear stress (Jones and Lick 2001): Egross = A τsf n for τsf > τcr (A-16) = 0 for τsf < τcr where Egross is gross erosion rate (cm/s), τsf is skin friction shear stress (Pa), and τcr is critical shear stress (Pa), which is the shear stress at which a small, but measurable, rate of erosion occurs (generally less than 2 mm/hr). The erosion parameters, A and n, are site-specific and may be spatially variable, both horizontally and vertically. The erosion rate of each sediment size class is affected by the probability of suspension for that size class (Psus,k), which is given by (Jones and Lick 2001): Psus,k = 0 for τsf < τc,k (A-17) = [ln(β1) - ln(β2)]/[1.39 ln(β2)] for τsf > τ and β1 < 4 = 1 for β1 > 4 and the non-dimensional parameters are: β1 = u*/ws,k (A-18) β2 = u*,crs,k/ws,k (A-19) The formulation presented in Equation A-17 was developed from the results of flume measurements of suspended and bed load transport of sand conducted by Guy et al. (1966). Jones and Lick (2001) analyzed the Guy et al. (1966) data, with Equation A-17 resulting from their analysis. Probability of suspension as a function of bed shear stress is shown in Figure Newtown Creek RI/FS A

46 Appendix A A-4 for particle diameters of 130 and 540 µm. This figure shows that for a given shear stress value, the probability of suspension increases with decreasing particle size. The particle-shielding factor, which is a positive number with a maximum value of one, is used to reduce the erosion flux of smaller particles within a graded bed (i.e., bed with wide range of particle sizes) that are sheltered by larger particles. The particle-shielding factor (Sk) for size class k is formulated as follows (Karim and Kennedy 1981; Rahuel et al. 1989): Sk = (dk/dm) 0.85 for dk < dm (A-20) = 1 for dk > dm where dm is the mean particle diameter in the active layer. The relationship between the particle-shielding factor and particle diameter, for three values of mean particle diameter, is shown in Figure A-5. For a given particle diameter (dk), the particle-shielding effect increases (i.e., Sk decreases) as the mean particle diameter increases. The particle-shielding factor is consistent with erosion processes within a graded bed, where voids between larger particles provide areas where smaller particles may be shielded (i.e., hide ) from the turbulence at the sediment-water interface that induces erosion. Thus, the particle-shielding factor is a mechanistic parameter that accounts for real processes that affect scour from a graded bed. The sediment bed model used in the bed scour model is similar to the bed model described in Jones and Lick (2001). This bed model has been developed over the previous 20 years and used within the SEDZL and SEDZLJ algorithms (Ziegler and Lick 1988, Ziegler et al. 2000, Jones and Lick 2001). The SEDZL/SEDZLJ bed model has been successfully used in over 30 sediment transport modeling studies, including: Upper Hudson River, Lavaca Bay (Texas), Grasse River (New York), Upper Mississippi River (Minnesota), Watts Bar Reservoir/Tennessee River (Tennessee), and Patrick Bayou (Texas). A multi-layer bed model is used in the SEDZLJ algorithm, with each bed layer having specific erosion rate parameters (i.e., τcr, A, and n). The effects of consolidation on erosion properties of deposited sediment are not explicitly incorporated into the bed model. If the Newtown Creek RI/FS A

47 Appendix A initial layer 1 is present, then deposited sediment is added to layer 1 (i.e., surface layer) of the bed model and, thus, that sediment has the same erosion properties as the surface layer. If the initial layer 1 is not present (i.e., that layer has been eroded), then a new surface layer is created by the deposited sediment which has the same erosion properties as the initial layer 1. This approach produces conservative results during a high-flow event because the erosion properties of sediment deposited prior to the event will not have been reduced due to consolidation. Erosion from cohesive and non-cohesive beds is affected by bed armoring, which is a process that tends to limit the amount of bed scour during a high-flow event. Bed armoring occurs in a bed that contains a range of particle sizes (e.g., clay, silt, sand). During a high-flow event when erosion is occurring, finer particles (i.e., clay and silt) tend to be eroded at a faster rate than coarser particles (i.e., sand). The differences in erosion rates of various particle sizes creates a thin layer at the surface of the bed, referred to as the active layer, that is depleted of finer particles and enriched with coarser particles. This depletion-enrichment process can lead to bed armoring, where the active layer is primarily composed of coarse particles that have limited mobility. After bed armoring occurs during a high-flow event, various physical mixing processes in the surface layer of the bed (e.g., bioturbation) can affect the armor layer. The effects of physical mixing processes on bed armoring are not well understood at the present time; these effects are not explicitly incorporated into the bed model and bed armoring algorithm. However, the effects of physical mixing processes are implicitly included into the bed model through use of the Sedflume data, which incorporates these effects into the erosion rate data. Physical mixing in the surface layer is one reason why near-surface sediment is generally more erodible than deeper sediment. The bed armoring process is simulated using an active layer at the surface of the bed, with the gross erosion rate being affected by the composition of the active layer (Jones and Lick 2001). The active layer is a theoretical construct that approximates the near-bed layer mentioned during the description of gross deposition and erosion rates previously in this section. The active layer is part of a numerical algorithm and it was created as a holding area such that the bed model realistically represents the complex processes at the sediment- Newtown Creek RI/FS A

48 Appendix A water interface. Even though the active-layer approach used in the model is a simplification of various complex processes, it is conceptually realistic and has been shown to produce accurate results in previous modeling studies. The surface-layer in the bed model (e.g., top 5-cm layer) is divided into two zones: 1) active layer; and 2) parent bed. The active layer is at the top of the surface layer and the parent bed is below it. The active layer interacts with the water column; erosion and deposition across the sediment-water interface occurs in the active layer. Use of an active layer to simulate the effects of bed armoring is frequently used in sediment transport models (Rahuel et al. 1989). The bed model tracks changes in the composition of the active layer associated with erosion and deposition; temporal changes in active layer composition affect the erosion process. The active layer is composed of two sub-layers: 1) active-surface layer; and 2) active-buffer layer. The active-surface layer interacts with the water column, while the active-buffer layer controls interactions between the active-surface layer and the parent bed (Figure A-6). The objective of separating the active layer into two sub-layers was to produce a more realistic representation of the interactions between the active and parent-bed layers in a tidal environment. The thickness of the active-surface layer is assumed to depend on bed shear stress and grain size distribution. The formulation used to calculate active-surface layer thickness (TAS) is (Jones and Lick 2001): TAS = 2 dm (τsf /τcr) (A-21) where dm is the mean particle diameter in the active layer. The active-surface layer thickness is temporally and spatially variable, and it changes as the composition of the bed and bed shear stress change with time. The active-surface layer thickness is determined using Equation A-21, with the bed model tracking the mass per unit area using: MAS = ρdry TAS (A-22) Newtown Creek RI/FS A

49 Appendix A where MAS is the total sediment mass per unit area in the active-surface layer and ρdry is the dry density of bed sediment. The thickness, or mass per unit area, of the active-surface layer changes with time as TAS changes as a result of increases or decreases in mean particle diameter or bed shear stress. Let δsb represent changes in active-surface layer mass, for size class k, caused by temporal changes in MAS. Expansion and contraction of the active-surface thickness (i.e., TAS) causes interactions between the active-surface and active-buffer layers, which result in mass transfer between the two layers. For increasing MAS (i.e., MAS N+1 > MAS N, where the superscript N represent time-level N in the numerical model): δsb,k = fab,k (MAS N+1 - MAS N ) (A-23) where fab,k is the fraction of size class k sediment in the active-buffer layer. For decreasing or constant MAS (i.e., MAS N+1 < MAS N ): δsb,k = fas,k (MAS N+1 - MAS N ) (A-24) where fas,k is the fraction of size class k sediment in the active-surface layer. The change in active-surface layer mass is calculated using: MAS,k N+1 = MAS,k N + δsb,k+ t (Dk - Ek - fas,k Dtot + fab,k Etot) (A-25) where MAS,k is active-surface layer mass per unit area for size class k sediment, Ek is the erosion flux for size class k sediment, Dk is the deposition flux for size class k sediment, and t is the numerical time-step. The total deposition and erosion fluxes are given by: Newtown Creek RI/FS A

50 Appendix A Dtot = Σ Dk (A-26) Etot = Σ Ek (A-27) where the summations are over all of the sediment size classes. In Equation A-27, the values of Ek are calculated using Equation A-15 for each size class k. Thus, Etot is affected by the composition of the active-surface layer. Note that the deposition and erosion flux terms in the parentheses on the right-hand side of Equation A-25 do not sum to zero for a specific size class k. This characteristic of the algorithm generates bed armoring effects due to unequal mass transfer of different sediment size classes between the active-surface, active-buffer and parent-bed layers. However, conservation of mass is assured when Equation A-25 is summed over all sediment size classes, which results in the sum of the deposition and erosion flux terms being equal to zero. The terms on the right-hand-side of Equation A-25 correspond to the following changes in the mass of the active-surface layer: 1) δsb,k is an increase in mass of class k sediment if the total active-surface layer mass is increasing (i.e., mass added from active-buffer layer) and it is a decrease in mass of class k sediment if the total active-surface layer mass is decreasing (i.e., mass lost to active-buffer layer); 2) t Dk is an increase in mass of class k sediment due to deposition from the water column to the bed; 3) t Ek is a decrease in mass of class k sediment due to erosion from the bed to the water column; 4) t fas,k Dtot is a decrease in mass of class k sediment caused by movement of sediment from the active-surface layer to the active-buffer layer due to deposition; and 5) t fab,k Etot is an increase in mass of class k sediment caused by movement of sediment from the active-buffer layer to the active-surface layer due to erosion (see Figure A-6). The change in active-buffer layer mass for size class k (MAB,k) is calculated using: MAB,k N+1 = MAB,k N - δsb,k + t [(fas,k - fab,k)dtot - fab,k Etot] (A-28) Newtown Creek RI/FS A

51 Appendix A It is assumed that there is no mass transfer between the buffer layer and the parent bed due to erosion processes. The terms on the right-hand-side of Equation A-28 correspond to the following changes in the mass of the active-buffer layer: 1) δsb,k is a decrease in mass of class k sediment if the total active-surface layer mass is increasing (i.e., mass lost to active-surface layer) and it is an increase in mass of class k sediment if the total active-surface layer mass is decreasing (i.e., mass added from active-surface layer); 2) t fas,k Dtot is an increase in mass of class k sediment caused by movement of sediment from the active-surface layer to the activebuffer layer due to deposition; 3) t fab,k Dtot is a decrease in mass of class k sediment caused by movement of sediment from the active-buffer layer to the parent-bed layer due to deposition; and 4) t fab,k Etot is a decrease in mass of class k sediment caused by movement of sediment from the active-buffer layer to the active-surface layer due to erosion. When the buffer layer is depleted of sediment (typically during an erosion event), the activesurface layer interacts directly with the parent bed (Figure A-7). Let δsp,k represent changes in active-surface layer mass, for size class k, caused by temporal changes in MAS and expansion/contraction interactions between the active-surface and parent-bed layers. For increasing MAS: δsp,k = fp,k (MAS N+1 - MAS N ) (A-29) where fp,k is the fraction of size class k sediment in the parent-bed layer. For decreasing or constant MAS: δsp,k = fas,k (MAS N+1 - MAS N ) (A-30) The change in active-surface layer mass for size class k is calculated using: MAS,k N+1 = MAS,k N + δsp,k + t (Dk - Ek - fas,k Dtot + fp,k Etot) (A-31) Newtown Creek RI/FS A

52 Appendix A The terms on the right-hand-side of Equation A-31 correspond to the following changes in the mass of the active-surface layer: 1) δsp,k is an increase in mass of class k sediment if the total active-surface layer mass is increasing (i.e., mass added from parent-bed layer) and it is a decrease in mass of class k sediment if the total active-surface layer mass is decreasing (i.e., mass lost to parent-bed layer); 2) t Dk is an increase in mass of class k sediment due to deposition from the water column to the bed; 3) t Ek is an decrease in mass of class k sediment due to erosion from the bed to the water column; 4) t fas,k Dtot is a decrease in mass of class k sediment caused by movement of sediment from the active-surface layer to the parent-bed layer due to deposition; and 5) t fp,k Etot is an increase in mass of class k sediment caused by movement of sediment from the parent-bed layer to the active-surface layer due to erosion. The change in parent-bed layer mass for size class k (MP,k) is determined from: MP,k N+1 = MP,k N - δsp,k + t (fas,kdtot fp,k Etot) (A-32) The terms on the right-hand-side of Equation A-32 correspond to the following changes in the mass of the parent-bed layer: 1) δsp,k is a decrease in mass of class k sediment if the total active-surface layer mass is increasing (i.e., mass lost to active-surface layer) and it is an increase in mass of class k sediment if the total active-surface layer mass is decreasing (i.e., mass added from active-surface layer); 2) t fas,k Dtot is an increase in mass of class k sediment caused by movement of sediment from the active-surface layer to the parent-bed layer due to deposition; and 3) t fp,k Etot is a decrease in mass of class k sediment caused by movement of sediment from the parent-bed layer to the active-surface layer due to erosion. After the buffer layer is depleted, a new active-buffer layer is created when the activesurface layer decreases in thickness as a result of decreasing bed shear stress. For the condition when MAB,k N+1 equals zero and MAS is decreasing (i.e., MAS N+1 < MAS N ), then the initial mass of the new active-buffer layer, for size class k, is: Newtown Creek RI/FS A

53 Appendix A MAB,k N+1 = fp,k (MAS N - MAS N+1 ) (A-33) This amount of mass is removed from the parent-bed layer, so that mass is conserved. The fractions of each sediment size class are updated after the new sediment masses are calculated in each layer: fas,k = MAS,k N+1 / MAS N+1 (A-34) fab,k = MAB,k N+1 / MAB N+1 (A-35) fp,k = MP,k N+1 / MP N+1 (A-36) where MAS N+1, MAB N+1,and MP N+1 are total sediment mass per unit area in the active-surface, active-buffer, and parent-bed layers, respectively. The numerical algorithm presented above for the interactions between the active-surface, active-buffer, and parent-bed layers may be difficult to understand from a conceptual viewpoint. The following sequence of figures is intended to clarify the mechanistic interactions between the three layers due to temporal variations in bed shear stress, which result in expansion and contraction of the active layer. It is assumed that initially (i.e., time = t1) two layers exist: 1) active-surface layer (with thickness TAS,1 corresponding to a shear stress value of τsf,1); and 2) parent-bed layer (see Figure A-8). As the shear stress increases to τsf,2 (which is greater than τsf,1) at time = t2, the active-surface layer thickness increases to TAS,2 and sediment is transferred from the parent-bed layer to the active-surface layer (Figure A- 9). The shear stress reaches a maximum value at time = t2 and decreases to a value of τsf,3 at time = t3. As the shear stress decreases during this time interval (i.e., t2 to t3), an active-buffer layer is created as the active-surface layer contracts in size, which is the process that generates an active-buffer layer (Figure A-10). This new active-buffer layer was created from a portion of the active-surface layer that existed at time = t2; sediment was transferred Newtown Creek RI/FS A

54 Appendix A from the active-surface layer to the active-buffer layer. As the shear stress continues to decrease during the time interval between t3 and t4, the active-surface and active-buffer layers decrease and increase in thickness, respectively (Figure A-11). The shear increases during the time interval between t4 and t5, which causes sediment to be transferred from the active-buffer layer (which is contracting) to the active-surface layer (which is expanding) (see Figure A-12). Note that during the time interval between t2 and t5, when the shear stress is less than the maximum value of τsf,2, the sum of the thicknesses of the active-surface and active-buffer layers remains constant at a value of TAS,2 (assuming that no deposition or erosion occurs). During the time interval between t5 and t6, the active-buffer layer is destroyed, and sediment is transferred from the parent-bed layer to the active-surface layer, as the shear stress exceeds the original maximum value of τsf,2 and the active-surface layer expands to a thickness greater than TAS,2 (Figure A-13). As the shear stress decreases from the new maximum value of τsf,6, a new active-buffer layer is created from the active-surface layer as that layer contracts in size (Figure A-14). The structure of the bed model described above is based on heuristic concepts that were developed from a general understanding of cohesive bed processes. The overall concepts applied to, and general behavior of, the model are consistent with known processes. However, uncertainty exists in some details of the model structure (e.g., transfer of sediment between the active-surface, active-buffer, and parent-bed layers as the active layer expands and contracts). Due to the complexity of the model structure, a unique methodology does not exist and a wide range of alternatives can be constructed from proposed general structure. However, the approach that is described above is consistent with a general understanding of cohesive bed processes and it does produce reasonable results. A.4 EROSION PROCESSES: NON-COHESIVE BED Non-cohesive sediment bed transport is dominated by gravitational, lift, and drag forces acting on individual particles. Cohesive forces are negligible compared to these other forces and are not evident in non-cohesive bed behavior. Non-cohesive beds generally contain only a small amount of clay and silt particles. Numerous laboratory and field studies have been conducted on the erosion properties of non-cohesive sediments; see van Rijn (1993) for an overview. These investigations have led to the development of various formulations for quantification of non-cohesive suspended and bed load transport. Several investigators have Newtown Creek RI/FS A

55 Appendix A evaluated the accuracy of different quantitative approaches using laboratory and field data (Garcia and Parker 1991; Voogt et al. 1991; van den Berg and van Gelder 1993). The results of these investigations have shown that the formulations developed by van Rijn (1984a, 1984b, 1984c) provide one of the best methods for calculating suspended load transport of non-cohesive sediments. The van Rijn equation have been successfully used in sediment transport modeling studies of riverine (Ziegler et al. 2000) and estuarine (van Rijn et al. 1990) systems over a wide range of flow and sediment conditions. The numerical algorithm discussed below is used to calculate the erosion flux of sediment from a non-cohesive bed to the water column, where it is transported as suspended sediment. Following the van Rijn method, the equations presented below are used to calculate the erosion flux for sediment size class k, which is represented by an effective particle diameter (dk). The critical bed-shear velocity for initiation of bed load transport (u*,crb,k) is calculated using the Shields criteria (see Figure A-15): u*,crb,k = [(s-1) g dk θcr,k] 0.5 (A-37) where θcr is the critical mobility parameter, which is approximated by (van Rijn 1993): θcr,k = 0.24 d*,k -1 for d*,k < 4 (A-38) = 0.14 d*,k for 4 < d*,k < 10 = 0.04 d*,k for 10 < d*,k < 20 = d*,k 0.29 for 20 < d*,k < 150 = for d*,k > 160 and d*,k is calculated using Equation A-13. Equation A-38 is a piece-wise fit to the Shields curve that was developed by van Rijn (1993). Critical shear stresses for initiation of bed load (τcrb,k) and suspended load (τcrs,k) transport are calculated as follows: τcrb,k = ρw u*,crb,k 2 (A-39) Newtown Creek RI/FS A

56 Appendix A τcrs,k = ρw u*,crs,k 2 (A-40) The relationships between particle diameter and the critical bed shear stresses for bed load and suspended load transport are shown in Figure A-16. For the cohesive sediment class, which represents clay and silt, it is assumed that Equations A-12 and A-37 through A-40 can be extrapolated to particle sizes less than 62 µm (i.e., d* less than 1.47). This assumption is commonly used for simulation of non-cohesive sediment transport with a graded bed (i.e., mixture of sediment particle sizes), and it has a minimal effect on model predictions in noncohesive bed areas. If the bed shear stress exceeds the critical shear stress for suspended load transport, then the equilibrium sediment concentration (Ceq,k) at a reference height (z = a) above the bed is calculated using: Ceq,k = (dk Tk 1.5 ) / (a d*,k 0.3 ) (A-41) where Tk is the transport stage parameter, given by: Tk = (u*/u*,crs,k) 2-1 for u* > u*,crs,k (A-42) The reference height (a) is calculated using: a = MAX (0.01 h, knik) (A-43) where h is water depth and knik is the Nikuradse roughness height: knik = 33 D90 (A-44) Newtown Creek RI/FS A

57 Appendix A The erosion flux for size class k sediment for a non-armoring sediment bed is calculated using: Ena,k = - Ws,k (Ca,k Ceq,k) for Ca,k < Ceq,k (A-45) where Ca,k is the suspended sediment concentration of size class k at z = a. For a threedimensional model, Ca,k is set equal to the suspended sediment concentration, as predicted by the water-column transport model, in the first grid cell above the bed. Similar to the cohesive bed discussed in Section A.3, bed armoring processes occur in the non-cohesive bed and those processes affect the erosion flux from that bed type. An active layer is assumed to exist at the surface of the non-cohesive bed, with the thickness of that layer calculated using Equation A-21. A bed model tracks changes in the composition of the non-cohesive active layer associated with erosion and deposition, as well as interactions between the active and parent bed layers. Thus, the erosion flux for size class k sediment from an armoring bed (Enon,k) is given by: Enon,k = fnon,a,k Sk Ena,k (A-46) where fnon,a,k is the fraction of class k sediment in the active layer of the non-cohesive bed and Sk is the particle-shielding factor (see Equation A-20). The particle-shielding factor (Karim and Kennedy 1981; Rahuel et al. 1989) was included in the erosion flux for an armoring bed because this factor accounts for the effects of differential erosion rates A.5 BED LOAD TRANSPORT Bed load transport consists of the movement of sand and gravel within a thin layer (i.e., about 1 cm thick or less) along the surface of the sediment bed. Simulation of bed load transport uses a control volume approach that consists of a surface layer of constant thickness (TB) that tracks the horizontal movement of bed load sediment into and out of a grid cell. This surface layer interacts with the parent bed layer below it (Figure A-17). A mass balance is constructed for each grid cell and it consists of summing for bed load fluxes across the four Newtown Creek RI/FS A

58 Appendix A boundaries of the grid cells. If the sum of the bed load fluxes is negative, then more sediment leaves the grid cell than enters it and bed degradation (net erosion) occurs. If the sum of the bed load fluxes is positive, then more sediment enters the grid cell than leaves it and bed aggradation (net deposition) occurs. The volumetric bed load transport rate at the (i-1/2, j) grid cell boundary for size class k sediment (van Rijn 1993) is [cm 2 /s]: q b,i 1 2,j,k = 0.053[(s 1)g]0.5 d k 0.3 T B,i 1 2,j,k P sus,k for u,i 1/2,j > u,crb (A-47) where the transport stage for bed load at the (i-1/2,j) grid cell boundary for size class k is: T B,i 1 2,j,k = u,i 1/2,j/u,crb,k 2 1 (A-48) The shear hear velocity at the (i-1/2,j) grid cell boundary is [cm/s]: u,i 1/2,j = C f 0.5 u i 1/2,j (A- 49) where ui-1/2,j is the current velocity predicted by the hydrodynamic model at the (i-1/2,j) grid cell boundary. Similarly, the volumetric bed load transport rate at the (i,j-1/2) grid cell boundary for size class k [cm 2 /s] is: q 1 = 0.053[(s b,i,j 2,k 1)g]0.5 d 1.5 k d 0.3,k T B,i,j 1 P 2,k sus,k for v,i,j 1/2 > u,crb,k (A-50) where the transport stage for bed load at the (i,j-1/2) grid cell boundary for size class k is: Newtown Creek RI/FS A

59 Appendix A T B,i,j 1 2,k = v,i,j 1/2/u,crb,k 2 1 (A-51) The shear hear velocity at the (i,j-1/2) grid cell boundary is [cm/s]: v,i,j 1/2 = C f 0.5 v i,j 1/2 (A-52) where vi,j-1/2 is the current velocity predicted by the hydrodynamic model at the (i,j-1/2) grid cell boundary. The sediment mass transport flux due to bed load across the (i-1/2, j) grid cell boundary for size class k is [g per timestep]: Γ b,i 1 2,j,k = t y ρ dry f SL,i 1 2,j,kq b,i 1 2,j,kSGN u i 1/2,j (A-53) where t is the timestep and y is the width of (i-1/2, j) grid cell boundary. The content of size class k in the surface bed load layer in the upwind grid cell is given by: f SL,i 1 2,j,k = 0.5 f SL,i,j,k SGN u i 1/2,j f SL,i,j,k f SL,i 1,j,k SGN u i 1/2,j f SL,i 1,j,k (A-54) The sediment mass transport flux due to bed load across the (i,j-1/2) grid cell boundary for size class k is [g per timestep]: Γ b,i,j 1 2,k = t x ρ dry f SL,i,j 1 2,kq b,i,j 1 2,kSGN v i,j 1/2 (A-55) where x is the width of (i,j-1/2) grid cell boundary. The content of size class k in the surface bed load layer in the upwind grid cell is given by: Newtown Creek RI/FS A

60 Appendix A f SL,i,j 1 2,k = 0.5 f SL,i,j,k SGN v i,j 1/2 f SL,i,j,k f SL,i,j 1,k SGN v i,j 1/2 f SL,i,j 1,k (A-56) Let Msum correspond to the sum of mass transport fluxes due to bed load for all size classes for grid cell (i,j): M sum = Σ Γ 1 b,i+ 2,j,k Γ b,i 1 2,j,k + Γ b,i,j+ 1 2,k Γ b,i,j 1,k (A-57) 2 where the summation is from k=1 to k= Ktotal. Now, the thickness (TB) and total mass (MSL,total) of the surface bed load layer are constant. If Msum is positive, then degradation (net erosion) occurred, which causes sediment movement from the parent bed to the surface layer and results in a net decrease of sediment mass in the parent bed. If Msum is negative, then aggradation (net deposition) occurred, which causes sediment movement from the surface layer to the parent bed and results in a net increase of sediment mass in the parent bed. Thus, the change in total mass of the parent bed is: n+1 M PB,total = n M PB,total M sum (A-58) If Msum is negative (net deposition), then the changes in mass for size class k in the surface layer and parent bed layers are given by: n+1 M SL.k = n M SL.k Γ b,i+ 1 2,j,k Γ b,i 1 2,j,k + Γ b,i,j+ 1 2,k Γ b,i,j 1 2,k f SLkδM total (A-59) n+1 M PB,k = n M PB,k + f SLk δm total (A-60) If Msum is positive (net erosion), then the changes in mass for size class k in the surface layer and parent bed layers are given by: Newtown Creek RI/FS A

61 Appendix A n+1 M SL.k = n M SL.k Γ b,i+ 1 2,j,k Γ b,i 1 2,j,k + Γ b,i,j+ 1 2,k Γ b,i,j 1 2,k f PB,kδM total (A-61) n+1 M PB,k = n M PB,k + f PB,k δm total (A-62) Finally, the content of class k sediment in the surface and parent bed layers is updated using: f SLk = n+1 M SL,k /M SL,total (A-63) f PB,k = n+1 M PB,k /M PB,total (A-64) Newtown Creek RI/FS A

62 APPENDIX A FIGURES

63 btr-\\montl-braymond\d_drive\newtown_creek\documents\newtown_creek_figure_a-1.docx Figure A-1 Probability of Deposition for Cohesive Sediment Using the Krone Formulation Modeling Approach Memorandum (1) Newtown Creek RI/FS

64 btr-\\montl-braymond\d_drive\newtown_creek\documents\newtown_creek_figure_a-2.docx Figure A-2 Probability of Deposition for Non-Cohesive Sediment as a Function of Bed Shear Stress and Particle Diameter Modeling Approach Memorandum (1) Newtown Creek RI/FS

65 btr-\\montl-braymond\d_drive\newtown_creek\documents\newtown_creek_figure_a-3.docx Figure A-3 Settling Speed of Discrete Sediment Particles as a Function of Particle Diameter Modeling Approach Memorandum (1) Newtown Creek RI/FS

66 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-4.pptx Figure A-4 Probability of Suspension as a Function of Bed Shear Stress for Particle Diameters of 130 and 540 µm Modeling Approach Memorandum (1) Newtown Creek RI/FS

67 btr-\\montl-braymond\d_drive\newtown_creek\documents\newtown_creek_figure_a-5.docx Figure A-5 Particle Shielding Factor as a Function of Particle Size Modeling Approach Memorandum (1) Newtown Creek RI/FS

68 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-6.pptx Figure A-6 Schematic of Interactions Between the Water Column, Active Layer, and Parent-Bed Layer when the Active-Buffer Layer is Present Modeling Approach Memorandum (1) Newtown Creek RI/FS

69 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-7.pptx Figure A-7 Schematic of Interactions Between the Water Column, Active Layer, and Parent-Bed Layer when the Active-Buffer Layer is not Present Modeling Approach Memorandum (1) Newtown Creek RI/FS

70 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-8.pptx Figure A-8 Initial Structure of Bed with no Active-Buffer Layer at Time = t 1 Modeling Approach Memorandum (1) Newtown Creek RI/FS

71 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-9.pptx Figure A-9 Active-Surface Layer Thickness Increases as Shear Stress Increases (τ 2 > τ 1 ) at Time = t 2 Modeling Approach Memorandum (1) Newtown Creek RI/FS

72 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-10.pptx Figure A-10 Active-Surface Layer Thickness Decreases and Active-Buffer Layer is Created as Shear Stress Decreases (τ 3 < τ 2 ) at Time = t 3 Modeling Approach Memorandum (1) Newtown Creek RI/FS

73 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-11.pptx Figure A-11 Active-Surface Layer Thickness Decreases and Active-Buffer Layer Thickness Increases as Shear Stress Continues to Decrease (τ 4 < τ 3 ) at Time = t 4 Modeling Approach Memorandum (1) Newtown Creek RI/FS

74 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-12.pptx Figure A-12 Active-Surface Layer Thickness Increases and Active-Buffer Layer Thickness Decreases as Shear Stress Increases (τ 5 > τ 4 ) at Time = t 5 Modeling Approach Memorandum (1) Newtown Creek RI/FS

75 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-13.pptx Figure A-13 Active-Surface Layer Thickness Increases and Active-Buffer Layer is Destroyed as Shear Stress Increases (τ 6 > τ 5, τ 6 > τ 2 ) at Time = t 6 Modeling Approach Memorandum (1) Newtown Creek RI/FS

76 btr - \\MONTL-BRAYMOND\D_Drive\Newtown_Creek\Documents\Newtown_Creek_Figure_A-14.pptx Figure A-14 Active-Surface Layer Thickness Decreases and New Active-Buffer Layer is Created as Shear Stress Decreases (τ 7 < τ 6 ) at Time = t 7 Modeling Approach Memorandum (1) Newtown Creek RI/FS

77 btr-\\montl-braymond\d_drive\newtown_creek\documents\newtown_creek_figure_a-15.docx Figure A-15 Initiation of Motion and Suspension for a Current Over a Plane Bed, θ=f(d * ), from Van Rijn (1989) Modeling Approach Memorandum (1) Newtown Creek RI/FS