Attribution of Haze Report (Phase I) Geographic Attribution for the Implementation of the Regional Haze Rule

Transcription

1 Attribution of Haze Report (Phase I) Geographic Attribution for the Implementation of the Regional Haze Rule Prepared for the: Western Regional Air Partnership Western Governors Association 1515 Cleveland Place, Suite 200 Denver, CO Prepared by: Air Resource Specialists, Inc Sharp Point Drive, Suite E Fort Collins, Colorado (970) March 14, 2005

2 TABLE OF CONTENTS Section Page 1.0 INTRODUCTION Background Attribution of Haze Project Analysis Methods Phase I Data Summary Products ANALYTICAL APPROACH Emissions Inventories Point and Area Source Emissions Mobile Source Emissions Fire Emissions Biogenic and Windblown Dust Emissions Ammonia Emissions Boundary Conditions Emissions Summaries Monitoring Data The IMPROVE Monitoring Network Modeling Air Quality Modeling Back Trajectory Modeling Attribution Assessments Tagged Species Source Apportionment (TSSA) Method Trajectory Regression Attribution Method ATTRIBUTION OF HAZE WEB SITE DESCRIPTION REGIONAL ATTRIBUTIONS AND ASSESSMENTS Source Attribution Results Regional Attributions 4-1 i

3 TABLE OF CONTENTS (continued) Section Page Results for Selected Sites Fire Episode Impacts on IMPROVE Samples Regional Assessments Fire Assessment Carbon Assessment Dust Assessment RECOMMENDATIONS FOR PHASE II Recommendations 5-2 LIST OF FIGURES Figure Page 2-1 Interim 2002 Emissions Totals for the WRAP Region a WRAP Annual 2002 Interim Emissions Estimates from all Source Categories for NO x by 36 km Grid Cell b WRAP Annual 2002 Interim Emissions Estimates from all Source Categories for SO 2 by 36 km Grid Cell State Emissions Map for 2002 SO 2 Emissions in California Map of WRAP Region Tribal and Mandatory Federal Class I Areas and IMPROVE Monitoring Sites Map Depicting the 2002 Annual Average IMPROVE Aerosol Extinction (in inverse megameters, Mm -1 ) by species in the WRAP States Comparison of the Extinction (B ext ), Deciview, and Visual Range (V.R.) Visibility Metrics West Elk Mountains The WRAP 2002 Air Quality Modeling Domain 2-20 ii

4 LIST OF FIGURES (continued) Figure Page 2-8a The Mean Fractional Bias (MFB) for the Model Simulated Data to the Ambient Monitoring Data for Each of the PM Ambient Data Networks b The Mean Fractional Bias (MFB) for the Model Simulated Data to the Ambient Monitoring Data for Each of the PM Ambient Data Networks Back Trajectory Residence Time Map for Great Sand Dunes NP TSSA Source Area Mapping File Graphical Representation of the Edge Effect Map Indicating the Selected Source Regions for the Rocky Mountain NP Monitoring Site State Report Page for North Dakota Tribal Report Page Class I Area Report Page for Rocky Mountain National Park Initial Grouping of Class I Areas by TSSA Source Region Attribution of Sulfate and Nitrate a TSSA Sulfate Contributions to Class I Areas Attributed to Washington b TSSA Nitrate Contributions to Class I Areas Attributed to Washington TSSA Sulfate and Nitrate Contribution at Each Class I Area Attributed to the Other (Non-point, Non-mobile) Source Category a 2002 Timeline of IMPROVE Monitoring Data for Rocky Mountain NP b 2002 Timeline of CMAQ Model Results for Rocky Mountain NP a Sulfate TSSA and TRA Source Apportionment Method Comparison for Rocky Mountain NP b Sulfate and Nitrate TSSA Source Apportionment Results for Rocky Mountain NP 4-35 iii

5 LIST OF FIGURES (continued) Figure Page 4-6a Colorado Emissions Map for SO b Colorado Emissions Map for NO x Residence Time Map for Rocky Mountain NP ( ) Simulated Visibility Based on the 20% Best Days and 20% Worst Days a 2002 Timeline of IMPROVE Monitoring Data for Yellowstone NP b 2002 Timeline of Raw IMPROVE Monitoring Data for Yellowstone NP c 2002 Timeline of CMAQ Model Results for Yellowstone NP a Sulfate TSSA and TRA Source Apportionment Method Comparison for Yellowstone NP b Sulfate and Nitrate TSSA Source Apportionment Results for Yellowstone NP a Wyoming Emissions Map for SO b Wyoming Emissions Map for NO x Residence Time Map for Yellowstone NP ( ) a 2002 Timeline of IMPROVE Monitoring Data for Mount Rainier NP b 2002 Timeline of CMAQ Model Results for Mount Rainier NP a Sulfate TSSA and TRA Source Apportionment Method Comparison for Mount Rainier NP b Sulfate and Nitrate TSSA Source Apportionment Results for Mount Rainier NP a Washington Emissions Map for SO b Washington Emissions Map for NO x Residence Time Map for Mount Rainier NP ( ) 4-56 iv

6 LIST OF FIGURES (continued) Figure Page 4-17a 2002 Timeline of IMPROVE Monitoring Data for Yosemite NP b 2002 Timeline of CMAQ Model Results for Yosemite NP a Sulfate TSSA and TRA Source Apportionment Method Comparison for Yosemite NP b Sulfate and Nitrate TSSA Source Apportionment Results for Yosemite NP a California Emissions Map for SO b California Emissions Map for NO x Residence Time Map for Yosemite NP ( ) Simulated Visibility Based on the 20% Best Days and 20% Worst Days a 2002 Timeline of IMPROVE Monitoring Data for Zion NP b 2002 Timeline of CMAQ Model Results for Zion NP a Sulfate TSSA and TRA Source Apportionment Method Comparison for Zion NP b Sulfate and Nitrate TSSA Source Apportionment Results for Zion NP a Utah Emissions Map for SO b Utah Emissions Map for NO x Residence Time Map for Zion NP ( ) a 2002 Timeline of IMPROVE Monitoring Data According to the RHR Guidance for Crater Lake NP b 2002 Timeline of Raw IMPROVE Monitoring Data, Including Data from Clogged Filter Events, for Carter Lake NP a 2002 Timeline of IMPROVE Monitoring Data According to the RHR Guidance for Petrified Forest NP 4-79 v

7 LIST OF FIGURES (continued) Figure Page 4-27b 2002 Timeline of Raw IMPROVE Monitoring Data, Including Data from Clogged Filter Events, for Petrified Forest NP Modeled Annual Average Contribution to Light Extinction by all Fire Categories for Modeled Annual Average Contribution to Light Extinction by Natural Fires for Modeled Annual Average Contribution to Light Extinction by Anthropogenic Fires for IMPROVE Organic Carbon Extinction as a Fraction of Total Aerosol Extinction Based on the Average of the 20% Highest Values for IMPROVE Elemental Carbon Extinction as a Fraction of Total Aerosol Extinction Based on the Average of the 20% Highest Values for Ratio of Organic to Elemental Carbon for all IMPROVE Sties during Summer PM 10 Emissions by Source Category 4-90 LIST OF TABLES Table Page 1-1 Approaches Used to Determine and Support Source Attribution by Species Summary of AoH Major Data Products Analyses and Information Available for Attribution Assessment Site Specifications, IMPROVE Network CMAQ Model Performance Compared to IMPROVE Monitoring Data for 2002, Monthly Bias (Positive/Negative) and Error (%) Back Trajectory Model Parameters Selected for AoH Analysis 2-24 vi

8 LIST OF TABLES (continued) Table Page 2-5 Source Tags Available for Use in the TSSA Simulation Class I Area Group Characteristics, Range of TSSA Sulfate and Nitrate Contributions in Percent (Rounded to Nearest 5%) a TSSA Analysis Attribution Results for SO4 (%), Based on 20% Highest Modeled Extinction Days in b TSSA Analysis Attribution Results for NO3 (%), Based on 20% Highest Modeled Extinction Days in a TSSA Analysis SO4 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in b TSSA Analysis NO3 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in Uncertainty Analysis of TRA Results and Comparison to TSSA Results Summary of Clogged Filters at WRAP IMPROVE Sites by Module in Phase I Fire Emissions Inventory Data PM 10 Dust Emissions by Source Category for Attribution of Haze Project, Phase II, Tentative Schedule 5-2 vii

9 List of Acronyms Acronym AIRS AoH APACE AQS ARL Definition Aerometric Information Retrieval System Attribution of Haze (WRAP Workgroup) The Atmospheric Particulate Carbon Exchange Air Quality System Air Resources Laboratory BC Boundary condition BCON Boundary concentration BEIS3 Biogenic Emissions Inventory System version 3 BRAVO Big Bend Regional Aerosol & Visibility Observations study CARB CASTNet CEM CENRAP CMAQ COHA DRI DV California Air Resources Board Clean Air Status and Trends Network (monitoring network) Continuous Emissions Monitoring Central Regional Air Planning Association Community Multiscale Air Quality Causes of Haze Assessment Desert Research Institute Deciview EC Elemental carbon EDAS Eta Data Assimilation System (meteorological data fields) EDMS Emissions Data Management System EGAS Economic Growth Analysis System EGU Electric generating unit EI Emissions inventory EIA-767 Energy Information Administration Form 767 EPA Environmental Protection Agency ETS Emissions Tracking System FEJF FNL GCVTC GEAR GEOSCHEM HYSPLIT IAS IC ICON Fire Emissions Joint Forum (WRAP Forum) Global Forecast System Final analysis (meteorological data fields) Grand Canyon Visibility Transport Commission Gas phase chemistry solver. Global chemical-transport model Hybrid-Single Particle Lagrangian Integrated Trajectory (dispersion/trajectory model) Integrated Assessment System Initial condition Initial concentration viii

10 List of Acronyms (Cont.) Acronym IMPROVE km mi Mm -1 MM5 data MOBILE6 MSF NADP NCAR NEI NH 3 NOAA NONROAD NO x NP OC PM PM 2.5 PM 10 PMC PSU RH RHR RMC RPO RRF SEARCH SIP SMOKE SO 2 SO 4 STN Definition Interagency Monitoring of Protected Visual Environments (monitoring network) Kilometer Mile Inverse megameter (units of light extinction) Fifth-generation Penn State/NCAR Mesoscale Model (meteorological field) Motor vehicle emissions factor model Mobile Sources Forum (WRAP Forum) National Atmospheric Deposition Program (monitoring network) National Center for Atmospheric Research National Emissions Inventory Ammonia National Oceanic and Atmospheric Administration Model for off-road emissions Oxides of nitrogen National Park Organic carbon Particulate matter Particulate matter less than 2.5 microns in diameter Particulate matter less than 10 microns in diameter Coarse particulate matter, between 2.5 and 10 microns in diameter Pennsylvania State University Relative humidity Regional Haze Rule Regional Modeling Center Regional Planning Organization Relative Reduction Factors Southeastern Aerosol Research and Characterization (monitoring network) State implementation plan Sparse Matrix Operator Kernel Emissions Sulfur dioxide Sulfate Speciation Trends Network (monitoring network) ix

11 List of Acronyms (Cont.) Acronym TIP TRA TSSA VIEWS VMT VOC VR WA WinHaze WRAP Definition Tribal implementation plan Trajectory Regression Analysis (attribution method) Tagged Species Source Apportionment (attribution method) Visibility Information Exchange Web System (visibility web site) Vehicle miles traveled Volatile organic carbon compounds Visual range Wilderness Area Visual Air Quality Modeler Western Regional Air Partnership x

12 1.0 INTRODUCTION This report is designed as a gateway to the detailed information integrated into Phase I of the Attribution of Haze (AoH) project. Most of this information resides on the Western Regional Air Partnership (WRAP) Web site ( and with WRAP data centers. The major sections of the report include: Section 1.0, Introduction This section steps through the AoH project background and goals, and provides context for the project results. Section 2.0, Analytical Approach This section describes the data sets and methods used to determine attribution of haze. Section 3.0, Attribution of Haze Web Site Description This section introduces the data summary and analysis products available on the AoH Web site. Section 4.0, Regional Attributions and Assessments This section presents the sulfate and nitrate attribution results, and summarizes other mass and visibility impact assessments that were performed. Section 5.0, Recommendations for Phase II This section presents the recommendations for future AoH work. 1.1 BACKGROUND The Western Regional Air Partnership is comprised of representatives from cooperating western states, tribes, and federal agencies. The WRAP was primarily established to implement the recommendations of the Grand Canyon Visibility Transport Commission (GCVTC) and to develop technical and policy tools to assist western states and tribes to comply with the U.S. Environmental Protection Agency s (EPA) Regional Haze Rule (RHR). The WRAP carries out its responsibilities through a network of committees and forums, composed of members and stakeholders who represent a wide range of viewpoints. The WRAP Strategic Plan for outlines a two-phased approach for developing state and tribal implementation plans (SIPs and TIPs) in accordance with the RHR (see: Phase I is intended as a trial run for Phase II. With Phase I the WRAP hopes to gain insight into which strategies are successful in meeting its goals, where data sets and other information are incomplete and need revision, and how to refine its approach to meet the needs of states and tribes in Phase II. 1-1

13 1.2 ATTRIBUTION OF HAZE PROJECT The Attribution of Haze Workgroup was established to prepare a policy-level report describing the emission source categories and geographic source regions presently contributing to visibility impairment at each of the over 100 tribal and mandatory federal Class I areas in the WRAP region. A broad representation of technical and policy representatives were selected for the workgroup and an open meeting format was established to foster additional input. Most meetings were attended by a mix of industry, government, tribal and environmental groups. The AoH Workgroup understands the importance of the guidance contained in the WRAP Tribal Template document ( Tribes, along with states and federal agencies, are full partners in the WRAP, having equal representation on the WRAP Board as states. Whether Board members or not, all tribes are governments, as distinguished from the stakeholders (private interest) which participate on Forums and Committees but are not eligible for the Board. Despite this equality of representation on the Board, tribes are very differently situated than states. There are over four hundred federally recognized tribes in the WRAP region, including Alaska. The sheer number of tribes makes full participation impossible. Moreover, many tribes are faced with pressing environmental, economic, and social issues, and do not have the resources to participate in an effort such as the WRAP, however important its goals may be. These factors necessarily limit the level of tribal input into and endorsement of WRAP products. The tribal participants in the WRAP make their best effort to ensure that WRAP products are in the best interest of the tribes, the environment, and the public. One interest is to ensure that WRAP policies, as implemented by states and tribes, will not constrain the future options of tribes who are not involved in the WRAP. With these considerations and limitations in mind, the tribal participants have joined the state, federal, and private stakeholder interests in approving this report as a consensus document. The specific goals of Phase I of the AoH project are: To provide state and tribal air regulators with an initial, regional assessment of the attribution of haze in their Class I areas; To provide an initial assessment of how and to what extent natural and anthropogenic emissions from each state affect western Class I areas; and Ultimately, to provide air regulators with the information and tools they need to prepare state and tribal implementation plans (SIPs and TIPs) under the RHR. The attribution results from Phase I are designed neither to explicitly single out individual sources nor to identify the amount of reduction needed by a given source or group of sources in order to meet the RHR goals. Phase II of the project will build upon the experiences, results, and recommendations of Phase I. Important additional steps in Phase II will be the modeling of future pollution levels based on emissions control strategies defined by WRAP forums, an assessment of the impact from WRAP states to nearby Class I areas in the Central Regional Air Planning Association (CENRAP), and the development of Web-based analytical tools. A tentative schedule for Phase II and recommendations for Phase II are presented in Section

14 1.3 ANALYSIS METHODS The term Attribution used in this report refers to an assessment of the natural and anthropogenic emission from geographic source regions that contribute to aerosol concentration and extinction measured or estimated at a Class I area. It was not the intent or design of the AoH project to perform original research. Rather, the project was designed to pull together existing information from various analyses and use that data to determine the source types and source areas impacting each of the WRAP Class I areas. The three major sources of data for this project include: Emissions inventories (EIs) While in some cases uncertain or incomplete, the EIs defined geographic source regions and provided estimates of emissions magnitudes. Monitoring data Light extinction calculated from measured speciated fine mass and total coarse mass define the scope of visibility impacts in or near Class I areas. Modeling results Atmospheric chemistry and transport models were used to make the connection between emissions from geographic source regions and fine mass collected in or near Class I areas. A weight of evidence approach was used to evaluate and describe source attribution. The methodology consisted of reviewing emissions inventories, monitoring data, and modeling results, for the 2002 calendar year. One or two independent source apportionment methods were applied to each Class I area, results were compared, and supporting data and information were used to corroborate or question apportionment results. The WRAP Regional Modeling Center (RMC) performed modeling runs for 2002 using the EPA Models-3/Community Multiscale Air Quality (CMAQ) model. The RMC developed a new algorithm in CMAQ to tag emissions inputs from particular source categories and source regions and then track the chemical transformations and transport of those emissions. Due to the complexity of the modeling technique, this Tagged Species Source Apportionment (TSSA) method was used to track only sulfate (SO 4 ) and nitrate (NO 3 ) associated with point and mobile sources from broad source regions. The point and mobile source categories were chosen because together these represent approximately 80% of the WRAP sulfur dioxide (SO 2 ) and oxides of nitrogen (NO x ) emissions. Together, sulfate and nitrate compounds account for 25 70% of the visibility impairment at Class I areas in the WRAP region. The WRAP Causes of Haze Assessment (COHA) contractor, Desert Research Institute (DRI), performed meteorological back trajectory analyses for all Class I area monitoring locations within WRAP using the National Oceanic and Atmospheric Administration (NOAA) Hybrid-Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model. Using back trajectory results and monitoring data, DRI was able to perform a Trajectory Regression Analysis (TRA) for each monitoring location. This analysis relates the amount of time air spends over a source region (determined by a compilation of many back trajectories) to the aerosol species measured at a receptor site. This method was applied to sulfate and aerosol extinction only. DRI also generated many other back trajectory products in support of the AoH project. 1-3

15 The TSSA method fosters an understanding of source-receptor relationship, including emissions, chemistry, and transport. The TRA method does not incorporate emissions or chemistry, and thus treats all potential source regions similarly. Given these differences, comparisons between TSSA and TRA results are encouraging. The major contributing source regions identified by both methods are generally the same, though there is often some disagreement in the magnitude of each region s impact. Often the major differences between methods occur in the attribution results associated with Canada, Mexico, and the Pacific Ocean. The TSSA method shows only minor contributions from Canada and Mexico in border states where international impacts would be expected to be important. Since the emissions inventories used for Canada and Mexico were older than U.S. EIs (from 1995 and 1999, respectively) and are believed to be incomplete, the modeled impacts from these regions, based on the TSSA method, are more uncertain than modeled impacts from WRAP states. The TSSA method includes a wide band of the Pacific Ocean in its designation of coastal states as source regions, so impacts from coastal shipping is not reported as separate from coastal states. Emissions inventories for the Pacific region are far from complete. The TRA method identifies a Pacific Coast region close to shore independent from the rest of the Pacific Ocean. The TRA method can indicate a significant (>10%) contribution from the Pacific that is not explicitly indicated by the TSSA method. While these differences are important, the attribution from WRAP states was considered the highest priority during Phase I. States and tribes will look to control the contributions from within WRAP in their efforts to reduce regional haze. States and tribes must cooperatively determine how best to use the attribution results. AoH has not defined a minimum threshold below which attribution results should be ignored. It may be more beneficial for states and tribes to review and understand source categories than to stipulate these minimum values. State and regional emissions inventories, and other analyses of back trajectories were compared to source attribution results, and often added confidence to the attribution findings. Not all emissions types were tracked by TSSA and not all species were analyzed by TRA, so attribution results are not available for all species. Other methods were used to determine the regional affects of fire, carbon, and wind blown dust within the WRAP. Class I areas were assigned to clusters based on the similarity of their attribution results. A total of twenty distinct clusters were identified. The motivation for grouping sites by species attribution was to understand if specific source regions might impact a group of Class I areas in a similar way, thus allowing those Class I areas to be treated together in Phase II. A closer review of these clusters is also anticipated in Phase II. Table 1-1 presents an overview of the approaches used to determine and support the attribution of each aerosol species. 1-4

16 Table 1-1 Approaches Used to Determine and Support Source Attribution by Species Attribution/Assessment Methods Supporting Analyses and Data Species CIA Regional TSSA TRA Modeling Back Trajectories Emissions Monitoring Data Sulfate x x x x 1 x Nitrate x x x 1 x Organic Carbon x x x 1 x Elem. Carbon x x x x Soil x x x x Coarse Mass x x x x Extinction x x x 1 Species defined in Table 1 are not necessarily emitted as primary particles, but may be result of chemical transformation of direct emissions. 1.4 PHASE I DATA SUMMARY PRODUCTS Analysis results for the AoH project cover a broad region of the country, with some results focusing on individual Class I areas and others on the entire modeling domain. This report outlines some of the major data summary products and results generated as part of Phase I of the AoH project. The regional, state, tribal, and individual Class I area products themselves are too numerous to include here, but are available on the project Web site at: Section 3.0 describes what products are available on this Web site and how to find them. A summary of the major products with sample Web links is provided in Table

17 Table 1-2 Summary of AoH Major Data Products State and Regional Reports Sample Link: State maps Regional maps showing the showing 2002 locations of visibility in WRAP, visibility monitors as measured in and in and near Class I near the Class I areas areas Regional and state maps showing gridded emissions data used for the 2002 AoH modeling Summaries of individual state impacts on all Class I areas Individual Federal and Tribal Class I areas Reports Sample link: monitored 2002 attribution and modeled results by TSSA visibility and TRA methods summaries Links to more detailed supporting data (RMC and DRI raw results, back trajectory maps) Links to VIEWS Web site and WRAP EDMS Other Products (Section 4.0 of this report) Identification of clusters of Class I areas which have similar attribution results Regional assessments of fire, carbon, and wind blown dust 1-6

18 2.0 ANALYTICAL APPROACH The WRAP Attribution of Haze project synthesized emissions inventories, monitoring data, and modeling results to attribute haze in tribal and mandatory federal Class I areas in the western U.S. AoH conclusions are based on a weight of evidence approach rather than any single data set, analysis result, or line of reasoning. Table 2-1 presents the analyses and information available for consideration. There is more information supporting results for some species, and the credibility of results depends on how well all categories of information agree. For example, the AoH Phase I work resulted in attribution results for sulfate, nitrate, and extinction only. Attribution results for each of these parameters can be supported by one or two independent attribution methods, meteorological back trajectory summaries, and emissions and monitoring data. Each data set or subsequent analysis of the data plays a well-defined role in an integrated analysis: Emissions inventories (EIs) are an important part of understanding the anthropogenic and natural sources of regional haze. EIs characterize the location and type of emission sources, as well as the mass emission rates of particulate matter and the precursors of particulate matter that affect visibility. Also, EIs can provide information on existing control levels so that engineering and policy judgments can be made as to the degree of additional control that may be achievable. By their very nature, EIs have limitations, including known or unknown errors in source location, temporal variation, and emission rates. Monitoring data (particulate, or aerosol measurements) are the most accurate indicators of the spatial and temporal variation of particulate matter and its chemical constituents throughout the country. Monitoring data is collected on a specific schedule which often integrates samples over a long period of time (24 hours) and does not occur on all days (1 in 3 days is common). The time resolution and spatial and temporal coverage of aerosol monitoring is often insufficient to characterize some short term or regional episodes. Modeling predicts the transformation, movement, and fate of emissions in the atmosphere. Modeling performance can be evaluated by comparing model predictions with monitoring data. Modeling involves the integration of meteorological data with complex dispersion and chemical algorithms and emissions data to produce predicted aerosol concentrations anywhere in the modeling domain. As with the emissions inventory and monitoring data, this approach cannot by itself adequately describe the relationship between sources and impacts, but it is an essential link between emissions inventories and monitoring data. Each of these data inputs is described briefly in the following three subsections. This chapter finishes with a description of the specific attribution assessment tools used for this project. Web links to more detailed information are offered throughout the discussion. 2-1

19 Table 2-1 Analyses and Information Available for Attribution Assessment Attribution Results Attribution Support Other Supporting Data Species Trajectory Back Trajectory TSSA Regression Summaries Emissions Monitoring Sulfate x x x x x Nitrate x x x x Org. Carbon x x x Elem. Carbon x x x Soil x x x Coarse Mass x x x Extinction x x x 2.1 EMISSIONS INVENTORIES To facilitate air quality modeling for 2002 prior to completion of the EPA National Emissions Inventory (NEI), WRAP developed an interim 2002 emissions inventory. The six pollutants included in the inventory are: oxides of nitrogen (NO x ); sulfur dioxide (SO 2 ); volatile organic carbon compounds (VOCs); ammonia (NH 3 ); fine particulate matter, less than 2.5 microns in diameter (PM 2.5 ); and coarse particulate matter, between 2.5 and 10 microns in diameter (PMC). This interim EI is most complete for the WRAP states and tribes. Tribal data was not obtained directly, but EPA inventories used were prepared with input from state, local, and tribal organizations. Emissions data generated by the Regional Modeling Center (RMC) (e.g. biogenic emissions) were produced using a variety of databases (e.g., meteorology and land cover) which provided inventories irrespective of tribal or state boundaries. For modeling purposes, several inventories were developed by the RMC, with the simulation labeled pre02c being the final inventory used for Phase I modeling. This section describes the development and coverage of each major piece of the inventory that was used in simulation pre02c. (A detailed Excel spreadsheet containing the AoH interim 2002 EI is available on the RMC Web site: The RMC continues to apply source refinements to the emissions inventories for source attribution and sensitivity tests. In the pre02d emissions scenario, which is the scenario used for graphical emissions representations in Phase I, refinements include separation of prescribed fires into anthropogenic and natural categories, and dividing area emissions into area, anthropogenic dust, and ammonia emissions. The most recent EI information, including individual point source emissions, can be obtained from the WRAP Emissions Data Management System (EDMS) Web site: Point and Area Source Emissions On behalf of the Emissions Forum, EH Pechan and Associates developed 2002 interim" EIs to represent 2002 emissions from point and area sources in the WRAP and the Central Regional Air Planning Association (CENRAP) regions. Point and area source emissions in the eastern U.S. were prepared by the Regional Modeling Center (RMC) emissions modeling team. 2-2

20 Mexican and Canadian inventories were available from the 1999 Big Bend Regional Aerosol & Visibility Observations (BRAVO) Study. Also, the 1995 Canadian national inventory used for the U.S. EPA Clear Skies modeling project, was available and was used to compliment the BRAVO inventory. Point sources are those sources from which emissions are released at a fixed location (e.g., a factory exhause stack). Point sources are often subdivided into electric generating units (EGUs) and non-egu sources, particularly in criteria inventories in which EGUs are a primary source of NO x and SO 2. Examples of non-egu major point sources include chemical manufacturers, refineries, smelters, and pulp and paper mills. (This is not the method used to estimate 2003 emissions for the SO 2 milestone report under development by the WRAP for the five states which submitted SIPs under Section 309 of the RHR. That report uses actual 2003 emissions for all SO 2 sources greater than 100 tons per year, slightly adjusted in some cases for purposes of comparison to the regional milestone.) For point sources in the WRAP region, the 2002 estimates began with the 1996 WRAP point source database, version 3 (generated from actual emissions). For EGUs, the emission estimates were based on 2002 Emissions Tracking System/Continuous Emissions Monitoring (ETS/CEM) data, and 2002 EIA-767 based estimates for the larger plants/units that submit emission estimates to this system. For copper smelters, 2002 sulfur dioxide (SO 2 ) emission estimates were updated with emissions provided by smelter companies. For non-smelter, nonutility point sources, the WRAP 1996 point source emissions database was used as the starting point for a six-year projection to 2002 using growth factors and Integrated Assessment System (IAS) model algorithms. For the top 100 emitting SO 2 facilities, the IAS model-estimated 2002 values for SO 2 were replaced with actual 2000 SO 2 emission estimates provided by the State air pollution control agencies. Also, all pollutants for facilities that were indicated as shutdown/closed were set to zero. In the CENRAP states, EGU estimates were retained at 1996 levels and non-utility, nonsmelter emissions were projected to 2002 from the 1996 WRAP point source emissions database using growth factors and IAS model algorithms. For the eastern U.S., the 1996 EPA NEI was projected to 2002 using EGAS 4.0 growth factors. For the point source 2002 emission estimates, the most certain values are those for the largest EGUs in the WRAP states, and copper smelter SO 2 emission estimates. Because emissions updates were limited to the WRAP region, changes outside of the WRAP are not captured in the current analysis and emissions levels are more uncertain. For non-utility, nonsmelter point sources, the approach used to estimate 2002 emissions does not account for all of the plant closures, new unit start-ups, and new installations of air pollution control equipment that have occurred since The approach for developing the point source EIs is described in detail on the WRAP Web site ( Area source emissions, by definition, are not emissions released at a well-defined fixed location, but emissions from sources that are too small, numberous, or difficult to be inventoried individually. Area source emissions are treated as a spatial distribution, usually spread over a county or air district. For this analysis, area sources include all source categories traditionlly inventoried as area sources by the states, with the following exceptions, which were accounted 2-3

21 for in separate categories: highway vehicles, nonroad engines/vehicles, dust from eigher paved or unpaved roads, or emissions from wildfire, prescribed burning, and agricultural burning. For the WRAP and CENRAP regions, the 1999 EPA NEI Version 3 was used as a starting point for a three-year projection to 2002 using growth factors from the Economic Growth Analysis System (EGAS) 4.0 for the 1999 to 2002 period. For the eastern U.S. the 1996 EPA NEI was projected to 2002 using EGAS 4.0 growth factors. The approach for developing the area source EIs is described in detail on the WRAP Web site ( Mobile Source Emissions At the direction of the WRAP Mobile Sources Forum (MSF), as part of the work to fulfill requirements of Section 309 of the Regional Haze Rule, mobile emissions inventories for the WRAP region were prepared by ENVIRON. These inventories were used as the basis for the mobile source portion of the 2002 interim inventory. Complete details of the methods and assumptions utilized to develop these inventories are available on the WRAP Web site ( The annual inventories for On-road and Offroad mobile sources were estimated for the 1996 base year and for four future years 2003, 2008, 2013 and The projected 2003 emissions were used as the interim 2002 emission estimates. Emissions for interim 2002 Road Dust was interpolated between 1996 and On-road mobile sources include vehicles certified for highway use. Sources can be computed either as being spread over a spatial extent or as being assigned to a line location, and on-road inventories can be reflected as either emissions or activity data. On-road mobile source emissions are estimated as the products of emissions factors and activity estimates. Samples of on-road mobile sources include light-duty gasoline vehicles and heavy-duty diesel vehicles. With the exception of California, the on-road mobile source emission factors for WRAP states were derived from EPA on-road emissions factor models. EPA s MOBILE6 model was used to estimate vehicle VOC, NO x and CO emissions and a modified version of EPA s PART5 model was the basis for estimating PM 10, PM 2.5, sulfate, ammonia, elemental carbon, and organic carbon emissions. Activity for on-road mobile sources is vehicle miles traveled (VMT) and vehicle speeds by roadway types. States were provided default modeling inputs and VMT for base and future years to review and update; several states provided revised data. The California Air Resources Board (CARB) has developed its own model for estimating California on-road emissions, and California provided on-road emissions estimates by county. For states east of the WRAP, the MOBILE6 model was used with VMT from the 1999 EPA NEI. Mexican and Canadian inventories were available from the 1999 BRAVO study and the 1995 Canadian national inventory used for the U.S. EPA Clear Skies modeling project. Non-road mobile sources include vehicular and otherwise movable sources not traveling on roadways. These sources are also computed as being spread over a spatial extent (a county or air district). Non-road mobile source emissions are estimated as the products of emissions factors and activity estimates. Examples of nonroad mobile sources include locomotives, lawn and garden equipment, construction vehicles, and boat emissions. 2-4

22 With the exception of California, EPA s June 2000 version of the NONROAD model was used to estimate off-road sources in the WRAP states. The NONROAD model includes estimates of emission factors, activity levels and growth factors for all traditional off-road sources. The default activity levels were provided to state agencies for input and update; however, no state provided updated off-road activity data. California provided off-road emission estimates on a county level. For this project, CARB provided county level emissions estimates for all off-road sources from an internal working version of its own off-road emissions model. Aircraft, commercial marine vessel and locomotive emissions are not addressed in EPA s NONROAD model. Emissions from these sources were estimated using EPA guidance and/or EPA methodology documented in the ENVIRON report. Due to identified model errors, NONROAD2000 overestimated emissions of some species. EPA has updated the NONROAD model twice in the last three years and the NONROAD2000 emission estimates are noticeably higher than the newer NONROAD2004 emissions estimates. Non-road mobile emissions for states east of the WRAP were projected from the 1996 EPA NEI to 2002 using EGAS 4.0 growth factors. Mexican and Canadian inventories were available from the 1999 BRAVO study and the 1995 Canadian national inventory used for the U.S. EPA Clear Skies modeling project, respectively. Because the eastern inventories relied on older versions of the EPA NONROAD model, it is expected that the interim inventory overestimates mobile sources in the east to a different extent than in the WRAP region. In 2005 these inventories will updated to produce the final 2002 working inventory and remodeling will be completed to refine the final results Fire Emissions On behalf of the Fire Emissions Joint Forum, Air Sciences, Inc., developed 2002 interim" EIs for the entire WRAP region, based on actual 2002 wildland and prescribed fires, and agricultural fire emissions estimates from work performed in the development of the Section 309 SIPs. Specific locations, dates, sizes and fuel loadings for each fire event were collected and included for fire within the WRAP states. Emissions outside of the WRAP region were collected from other Regional Planning Organizations (RPOs) and other sources, but were generally incomplete. The approach for developing the WRAP fire EIs is described in detail on the WRAP Web site ( Biogenic And Windblown Dust Emissions The WRAP Regional Modeling Center (RMC) developed biogenic and windblown dust EIs to support air quality modeling efforts. These emissions models required data inputs from a variety of sources (Land use/land cover; Windblown dust soil characteristics, Soil Landscape of Canada, International Soil Reference and Information Centre; Meteorological data including km gridded MM5 runs; and Agricultural data). Biogenic emissions were modeled using the Biogenic Emissions Inventory System version 3 (BEIS3). Windblown dust emissions were 2-5

23 modeled using the WRAP Windblown Dust Emissions Model. Emissions for the entire modeling domain were generated, although outside of the WRAP states the emissions modeling of windblown dust was problematic. No information regarding when agricultural lands were covered or barren was available. Without this information modeling of agricultural lands will contain inaccuracies. If the model assumes agricultural lands are barren when in fact they are not, emissions of windblown dust will be overestimated. The approach for developing the WRAP biogenic and windblown dust EIs is described on the RMC Web site ( Ammonia Emissions The WRAP RMC developed interim 2002 ammonia emissions for the WRAP states and tribes by improving the 1996 NH 3 emissions inventory that was used to meet requirements of Section 309 of the Regional Haze Rule. Significant ammonia emissions include livestock operations, fertilizer use, waste management, mobile sources, industrial point sources, wild and prescribed fires and various biological sources including human respiration, wild animals and soil microbial processes. In the current ammonia emissions inventory, only livestock operation, fertilizer use, soil microbial processes and domestic sources are considered. Ammonia emissions from mobile sources, industrial point sources and wild and prescribed fires were included in other inventories. The ammonia emission modeling system was populated with data that included data sources, activity data, emissions factors and environmental variables. Details of the approach for developing the WRAP ammonia inventory were described in a paper presented by the RMC at the EPA 13th International Emission Inventory Conference "Working for Clean Air in Clearwater" ( Boundary Conditions Establishing conditions for the boundary of the WRAP modeling domain is necessary to account for emissions from outside the domain and their transformation and transport to the WRAP region. The RMC facilitated the development of modeling domain boundary conditions EI to support air quality modeling efforts. The boundary conditions were derived from an annual simulation of the GEOSCHEM global chemical-transport model, developed at Harvard University. Early simulations used boundary conditions based on seasonal averages. This was later modified for the interim 2002 EI to include 3-hour resolution of boundary conditions. The initial conditions used for the 2002 modeling run were generated by running the model for the period December 17 31, A description of the boundary conditions and initial conditions used for the model run are described on the RMC Web site ( Emissions Summaries Figure 2-1 presents a summary of the WRAP annual 2002 interim emissions. The attribution methods used in the AoH Phase I analysis focused on particulate nitrate and sulfate, which are formed from chemical transformations of NO x and SO 2 emissions. Due to computational and time limitations, only point and mobile (on-road and non-road) sources were 2-6

24 tagged for the attribution analysis. As can be seen in the figure, these sources account for approximately 80% of the total NO x and SO 2 emissions in the WRAP region. Together, the sulfate and nitrate particulate formed from these emissions account for approximately 25 70% of the visibility impairment in the WRAP region. Figures 2-2a and 2-2b presents the total interim NO x and SO 2 emissions from all sources by model grid cell (36 x 36 km). These and additional regional maps are available on the project Web site ( Figure 2-3 presents a sample SO 2 emissions map for a state. This map displays the same information as the regional map in Figure 2-2b, but focuses on the state of interest and its surroundings. Additional state emissions maps (all states, all emission types) are available on the project Web site under individual State Reports ( 2-7

25 Interim 2002 Emissions Totals for the WRAP Region Oxides of Nitrogen (NO X ) 4,479 thousand tons/yr Sulfer Dioxide (SO 2 ) 1,326 thousand tons/yr Volatile Organic Carbon (VOC) 17,288 thousand tons/yr Fine Particulates (PM 2.5 ) 2,345 thousand tons/yr Point Area On-Road Mobile Off-Road Mobile Rx Fires Ag Fires Off-Shore Windblown Dust Road Dust Wildland Fire Biogenics Animals/Soils NH3 Ammonia (NH 3 ) 1,938 thousand tons/yr Coarse Particulates (PMC) 3,524 thousand tons/yr Figure 2-1. interim 2002 Emissions Totals for the WRAP Region. 2-8

26 2-9 Figure 2-2a. WRAP Annual 2002 Interim Emissions Estimates from all Source Categories for NO x by 36 km Grid Cell.

27 Figure 2-2b. WRAP Annual 2002 Interim Emissions Estimates from all Source Categories for SO 2 by 36 km Grid Cell

28 California SO 2 Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Sulfur oxide gases (SOx) are formed when sulfur containing fuels, such as oil or coal, are burned, when gasoline is extracted from oil or when metals are extracted from ore. In California, 2002 emissions of SO 2 were dominated by point and off-road sources. SO 2 dissolves in water vapor to form acid, and contributes to the formation of sulfate compounds (e.g. (NH 4 ) 2 SO 4 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. SO 2 Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% California 2002 SO 2 Emissions 123 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Offshore Natural (N) Rx Fires Wildland Fire 0% A N Figure 2-3. State emissions map for 2002 SO 2 emissions in California.

29 2.2 MONITORING DATA Ambient monitoring data covering aerosol and meteorological parameters played an important role in the AoH project. The WRAP did not conduct this monitoring or contract it out; rather, data were retrieved from national programs designed for routine monitoring The IMPROVE Monitoring Network The Interagency Monitoring of Protected Visual Environments (IMPROVE) monitoring program collects speciated PM 2.5, and PM 2.5 and PM 10 mass. IMPROVE is a nation-wide network which began in 1988 and expanded significantly in 2000 in response to the EPA s Regional Haze Rule. It is, in fact, data from this program that states and tribes must use to track progress under the RHR. The primary purposes of the IMPROVE network are to characterize spatial patterns of regional haze, identify long-term and seasonal trends, and support source apportionment. A listing of site affiliations, names, abbreviations, locations, and operational start dates is presented in Table 2-2. Some Class I areas do not operate aerosol samplers but are represented by samplers located at other, nearby Class I areas. The representative monitoring site for each Class I area is indicated in the Site Name and Site Code fields in Table 2-2. A map of the IMPROVE sites within the WRAP region and their proximity to Class I areas is presented in Figure 2-4. The design of the IMPROVE network and sampling procedures is dictated by the purposes of the network identified above, the need to control costs, maintain consistency, and the often remote locations of the monitoring sites. The IMPROVE network collects 24-hour integrated filter samples every three days (Wednesday and Saturday prior to 2000). Each monitoring location operates 4 samplers, designated Module A through D: Module A utilizes a Teflon filter for PM 2.5 gravimetric and elemental analysis. The visibility-related species derived from this module are sulfate, soil, and coarse mass (in conjunction with module D). Module B utilizes a nylon filter preceded by a carbonate denuder for PM 2.5 ion analysis. The visibility-related species derived from this module are nitrate and sulfate (backup measurement). Module C utilizes a quartz filter for PM 2.5 carbon analysis. The important visibilityrelated species derived from this module are organic carbon and elemental carbon. Module D utilizes a Teflon filter for PM 10 gravimetric analysis. The difference between module D PM10 and module A PM 2.5 yields an estimate of coarse mass, the final visibility-related species. Detailed information regarding IMPROVE sampling protocols is available on the IMPROVE Web site ( and the Visibility Information Exchange Web System (VIEWS) Web site ( The IMPROVE program has developed a method for estimating light extinction from speciated aerosol and relative humidity data. Each major species is assigned a constant dry mass extinction efficiency. This accounts for the fact that an elemental carbon particle is ten times more efficient at absorbing light than a particle of sand is at scattering light. The sum of species

30 mass for a given sample will not necessarily agree with the gravimetric mass (determined by weighing the filter) due to assumptions based on average values, which may be inaccurate on a given day or under certain circumstances. IMPROVE makes the assumption that all sulfur present and all sulfate ions present exists as ammonium sulfate. In reality, there are other forms of particulate sulfate, and the mix of sulfate types affects both the total sulfate mass and its contribution to extinction. IMPROVE makes the assumption that all nitrate ions exist as ammonium nitrate. Some nitrate may be in the form of sodium nitrate, though the percentage on a given sample or on average at individual sites is not currently known. Sulfate and nitrate species are known to absorb water and thus their contribution to extinction is enhanced above certain values of relative humidity (RH) as the particles increase in size. As the RH increases, IMPROVE assumes an increase in scattering by these species. EPA Regional Haze guidance and current IMPROVE protocol call for the use of a climatologically representative monthly average RH enhancement factor. This approach removes much of the short-term variability of RH effects and allows calculation of extinction at sites which do not routinely monitor RH. However, extinction calculated using a long-term average of RH will likely not represent the actual visibility conditions on a given day. Once extinction has been calculated for each species at a monitoring site, there are complicated rules for data completeness and handling missing data under the Regional Haze Rule (see Guidance for Tracking Progress Under the Region Haze Rule: RHR guidance allows for data substitution of missing species mass under certain conditions. It is possible, for example, to lose the nitrate or coarse mass data for a sampling day (due to a variety of errors) yet not lose that day in the annual data set. These rules define what data can be used by states and tribes for the purposes of tracking regional haze progress. Figure 2-5 presents an overview of 2002 annual average extinction in the WRAP region by species, modified based on the allowable data substitution rules. While this report presents visibility measurements as extinction, there are other useful metrics. The three most common metrics used to describe visibility impairment are: Extinction Extinction is a measure of the fraction of light lost per unit length along a sight path due to scattering and absorption by gases and particles, expressed in inverse Megameters (Mm -1 ). This metric is useful for representing the contribution of each aerosol species to visibility impairment and can be practically thought of as the units of light lost in a million meter distance. Visual Range Visual range is the greatest distance a large black object can be seen on the horizon, expressed in kilometers (km) or miles (mi). Deciview The deciview index was designed to be linear with respect to human perception of visibility. A one deciview change is approximately equivalent to a 10% change in extinction, whether visibility is good or poor. A one deciview change in visibility is generally considered to be the minimum change the average person can detect. This is the metric used for tracking regional haze in the RHR. Figure 2-6 presents a comparison of these metrics and a set of images with simulated visibility impairment. As can be seen in the figure, an increase in extinction or deciview is equivalent to a decrease in visual range.

31 Table 2-2 Site Specifications IMPROVE Network Class I Area Site Name Site Code Agency State Site Lat. Site Long. Site Elev. Start Date Denali NP and Preserve Denali DENA1 NPS AK /2/1988 Simeonof W Simeonof SIME1 FWS AK /10/2001 Tuxedni W Tuxedni TUXE1 FWS AK /18/2001 Bering Sea W N/A N/A FWS AK N/A Mount Baldy W Mount Baldy BALD1 FS AZ /29/2000 Chiricahua NM Chiricahua CHIR1 NPS AZ /2/1988 Chiricahua W Chiricahua CHIR1 FS AZ 3/2/1988 Galiuro W Chiricahua CHIR1 FS AZ 3/2/1988 Grand Canyon NP Hance Camp GRCA2 NPS AZ /24/1997 Hualapai Tribe Hance Camp GRCA2 Tribal AZ 9/24/1997 Mazatzal W Ike's Backbone IKBA1 FS AZ /2/2000 Pine Mountain W Ike's Backbone IKBA1 FS AZ 4/2/2000 Grand Canyon NP - In Canyon Indian Gardens INGA1 NPS AZ /4/1989 Petrified Forest NP Petrified Forest PEFO1 NPS AZ /2/1988 Saguaro NP - East Saguaro SAGU1 NPS AZ /4/1988 Saguaro NP - West Saguaro West SAWE1 NPS AZ /19/2001 Sierra Ancha W Sierra Ancha SIAN1 FS AZ /10/2000 Sycamore Canyon W Sycamore Canyon SYCA1 FS AZ /26/2000 Yavapai-Apache Nation Sycamore Canyon SYCA1 Tribal AZ 4/26/2000 Superstition W Tonto TONT1 FS AZ /23/1988 Agua Tibia W Agua Tibia AGTI1 FS CA /15/2000 Desolation W Bliss State Park BLIS1 FS CA /17/1990 Mokelumne W Bliss State Park BLIS1 FS CA 11/17/1990 Dome Lands W Dome Lands DOME1 FS CA /1/2000 Hoover W Hoover HOOV1 FS CA /1/2001 Joshua Tree NP Joshua Tree JOSH1 NPS CA /22/2000 Kaiser W Kaiser KAIS1 FS CA /26/2000 Ansel Adams W Kaiser KAIS1 FS CA 1/26/2000 John Muir W Kaiser KAIS1 FS CA 1/26/2000 Lava Beds NM Lava Beds LABE1 NPS CA 3/25/2000 South Warner W Lava Beds LABE1 FS CA /25/2000 Lassen Volcanic NP Lassen Volcanic LAVO1 NPS CA /2/1988 Caribou W Lassen Volcanic LAVO1 FS CA 3/2/1988 Thousand Lakes W Lassen Volcanic LAVO1 FS CA 3/2/1988 Pinnacles NM Pinnacles PINN1 NPS CA /2/1988 Ventana W Pinnacles PINN1 FS CA 3/2/1988 Point Reyes NS Point Reyes PORE1 NPS CA /2/1988 San Rafael W San Rafael RAFA1 FS CA /2/2000 Redwood NP Redwood REDW1 NPS CA /2/1988 San Gabriel W San Gabriel SAGA1 FS CA /15/2000 Cucamonga W San Gabriel SAGA1 FS CA 12/15/2000 San Gorgonio W San Gorgonio SAGO1 FS CA /2/1988 San Jacinto W San Gorgonio SAGO1 FS CA 3/2/1988 Sequoia NP Sequoia SEQU1 NPS CA /4/1992 Kings Canyon NP Sequoia SEQU1 NPS CA 3/4/1992 Marble Mountain W Trinity TRIN1 FS CA /19/2000 Yolla Bolly-Middle Eel W Trinity TRIN1 FS CA 7/19/2000 Yosemite NP Yosemite YOSE1 NPS CA /9/1988 Emigrant W Yosemite YOSE1 FS CA 3/9/1988 Great Sand Dunes NP Great Sand Dunes GRSA1 NPS CO /4/1988 Mesa Verde NP Mesa Verde MEVE1 NPS CO /5/1988 Mount Zirkel W Mount Zirkel MOZI1 FS CO /30/1994 RaWh W Mount Zirkel MOZI1 FS CO 7/30/1994 Rocky Mountain NP Rocky Mountain ROMO1 NPS CO /19/1990 Weminuche W Weminuche WEMI1 FS CO /2/1988 Black Canyon of Gunnison NP Weminuche WEMI1 NPS CO 3/2/1988 La Garita W Weminuche WEMI1 FS CO 3/2/1988 Eagles Nest W White River WHRI1 FS CO /17/1993 Flat Tops W White River WHRI1 FS CO 7/17/1993 Maroon Bells-Snowmass W White River WHRI1 FS CO 7/17/1993 West Elk W White River WHRI1 FS CO 7/17/1993 Haleakala NP Haleakala HALE1 NPS HI /16/1991 Hawaii Volcanoes NP Hawaii Volcanoes HAVO1 NPS HI /23/1988 Craters of the Moon NM Craters of the Moon CRMO1 NPS ID /13/1992 Sawtooth W Sawtooth SAWT1 FS ID /26/1994 (cont.)

32 Table 2-2 (cont.) Site Specifications IMPROVE Network Class I Area Site Name Site Code Agency State Site Lat. Site Long. Site Elev. Start Date Cabinet Mountains W Cabinet Mountains CABI1 FS MT /24/2000 Confederated Salish and Flathead FLAT1 Tribal MT /19/2002 Kootenai Tribes Fort Peck Tribes Fort Peck FOPE1 Tribal MT /25/2002 Gates of the Mountains W Gates of the Mountains GAMO1 FS MT /25/2000 Glacier NP Glacier GLAC1 NPS MT /2/1988 Medicine Lake W Medicine Lake MELA1 FWS MT /15/1999 Bob Marshall W Monture MONT1 FS MT /28/2000 Mission Mountains W Monture MONT1 FS MT 3/28/2000 Scapegoat W Monture MONT1 FS MT 3/28/2000 Northern Cheyenne Tribe Northern Cheyenne NOCH1 Tribal MT /22/2002 Selway-Bitterroot W Sula Peak SULA1 FS MT /10/1994 Anaconda-Pintler W Sula Peak SULA1 FS MT 8/10/1994 U.L. Bend W UL Bend ULBE1 FWS MT /25/2000 Red Rocks Lakes W Yellowstone 2 YELL2 FWS MT 7/1/1996 Lostwood Wilderness Lostwood LOST1 FWS ND /15/1999 Theodore Roosevelt NP Theodore Roosevelt THRO1 NPS ND /15/1999 Bandelier NM Bandelier BAND1 NPS NM /2/1988 Bosque del Apache W Bosque del Apache BOAP1 FWS NM /5/2000 Gila W Gila GICL1 FS NM /6/1994 Carlsbad Caverns NP Guadalupe Mountains GUMO1 NPS NM /2/1988 Salt Creek W Salt Creek SACR1 FWS NM /6/2000 San Pedro Parks W San Pedro Parks SAPE1 FS NM /15/2000 White Mountain W White Mountain WHIT1 FS NM /15/2002 Wheeler Peak W Wheeler Peak WHPE1 FS NM /15/2000 Pecos W Wheeler Peak WHPE1 FS NM /15/2000 Jarbidge W Jarbidge JARB1 FS NV /2/1988 Crater Lake NP Crater Lake CRLA1 NPS OR /2/1988 Diamond Peak W Crater Lake CRLA1 FS OR 3/2/1988 Gearheart Mountain W Crater Lake CRLA1 FS OR 3/2/1988 Mountain Lakes W Crater Lake CRLA1 FS OR 3/2/1988 Hells Canyon W Hells Canyon HECA1 FS OR /1/2000 Kalmiopsis W Kalmiopsis KALM1 FS OR /7/2000 Mount Hood W Mount Hood MOHO1 FS OR /7/2000 Eagle Cap W Starkey STAR1 FS OR /7/2000 Strawberry Mountain W Starkey STAR1 FS OR 3/7/2000 Three Sisters W Three Sisters THSI1 FS OR 7/24/1993 Mount Jefferson W Three Sisters THSI1 FS OR /24/1993 Mount Washington W Three Sisters THSI1 FS OR 7/24/1993 Badlands NP Badlands BADL1 NPS SD /2/1988 Wind Cave NP Wind Cave WICA1 NPS SD /15/1999 Bryce Canyon NP Bryce Canyon BRCA1 NPS UT /2/1988 Canyonlands NP Canyonlands CANY1 NPS UT /2/1988 Arches NP Arches CANY1 NPS UT 3/2/1988 Capital Reef NP Capitol Reef CAPI1 NPS UT /28/2000 Zion NP Zion ZION1 NPS UT /21/2000 Mount Rainier NP Mount Rainier MORA1 NPS WA /2/1988 North Cascades NP North Cascades NOCA1 NPS WA 3/1/2000 Glacier Peak W North Cascades NOCA1 FS WA /1/2000 Olympic NP Olympic OLYM1 NPS WA /11/2001 Pasayten W Pasayten PASA1 FS WA /15/2000 Alpine Lakes W Snoqualmie Pass SNPA1 FS WA /3/1993 Spokane Tribe of Indians Spokane Res. SPOK1 Tribal WA /11/2001 Goat Rocks W White Pass WHPA1 FS WA /15/2000 Mount Adams W White Pass WHPA1 FS WA 2/15/2000 Bridger W Bridger BRID1 FS WY /2/1988 Fitzpatrick W Bridger BRID1 FS WY 3/2/1988 North Absaroka W North Absaroka NOAB1 FS WY /25/2000 Washakie W North Absaroka NOAB1 FS WY 1/25/2000 Yellowstone NP Yellowstone 2 YELL2 NPS WY /1/1996 Grand Teton NP Yellowstone 2 YELL2 NPS WY 7/1/1996 Teton W Yellowstone 2 YELL2 FS WY 7/1/1996

33 Map of WRAP Region Class I Areas and IMPROVE Monitoring Sites WRAP Class I Areas Federal Tribal IMPROVE Monitoring Sites CIA Other > 3 years (through 2002) 1-3 years < 1 year Discontinued site* *Not operational in AK HI Figure 2-4. Map of WRAP region tribal and mandatory federal Class I areas and IMPROVE monitoring sites. Monitoring site type and length of data record through 2002 are indicated by color and shape.

34 2002 Annual Average Aerosol Extinction by Species at WRAP IMPROVE Sites 2002 Annual Average IMPROVE Aerosol Extinction (Mm-1) Ammonium Sulfate Ammonium Nitrate Organic Material Elemental Carbon Soil Coarse Material AK HI Figure 2-5. Map depicting the 2002 annual average IMPROVE aerosol extinction (in inverse megameters, Mm -1 ) by species in the WRAP states.

35 Figure 2-6. Comparison of the Extinction (B ext ), Deciview, and Visual Range (V.R.) visibility metrics. The six images of the West Elk Mountains were simulated to represent varying degrees of visibility impairment. 2-18

36 2.3 MODELING The Regional Monitoring Center (RMC) ran air quality and meteorological models to track emissions and ultimately to attribute pollutants to source categories and geographic regions. Desert Research Institute (DRI) ran a trajectory model in support of an analysis to attribute measured species mass concentrations to geographic regions. This section describes the major model components used by these two organizations Air Quality Modeling The RMC modeling effort focused on seasonal and annual regional visibility using a deterministic photochemical grid model capable of simulating the formation, transport, and fate of atmospheric pollutants in one hour increments. The air quality model selected for this work is the EPA Models-3/Community Multiscale Air Quality (CMAQ) model. The RMC developed a 2002 annual air quality modeling database to use with CMAQ, comprised of the EPA Sparse Matrix Operator Kernel Emissions (SMOKE) emissions model, and the Mesoscale Model version 5 (MM5) meteorological model (developed and maintained by the Pennsylvania State University (PSU) and National Center for Atmospheric Research (NCAR)). SMOKE was used to process the input data for anthropogenic and biogenic gas and PM emission rate estimates for CMAQ. The emissions input data were initially based on the NEI data for 1999 with many updates and improvements from a variety of data sets and emissions modeling activities supported by WRAP. MM5 was used to develop the required hourly meteorological fields. (Indepth descriptions of CMAQ and SMOKE can be found at: The CMAQ model requires inputs of three-dimensional gridded wind, temperature, humidity, cloud/precipitation, and boundary layer parameters. The current version of CMAQ can only utilize output fields from the MM5 meteorological model. MM5 is a state-of-the science model proven useful for air quality applications and has been used extensively in past local, state, regional, and national modeling efforts. MM5 has undergone extensive peer-review, with all of its components continually undergoing development and scrutiny by the modeling community. The MM5 modeling system software is freely provided and supported by the Mesoscale Prediction Group in the Mesoscale and Microscale Meteorology Division of NCAR. For these reasons, MM5 is the most widely used public-domain prognostic model. (In-depth descriptions of MM5 can be found at Air quality models are used in the SIP/TIP development process in two stages: first, the models are validated by performing model simulations of visibility for an historical episode; second, the models are operated for future projections, assuming that the meteorology is unchanged in the future, to assess the adequacy of various emissions reduction strategies for attaining visibility goals. In this work the MM5/SMOKE/CMAQ modeling system was applied for an annual simulation period during calendar year 2002 over a spatial domain that includes the continental US and large areas of Canada and Mexico (see Figure 2-7). The master domain for this model effort was determined necessary by the Regional Planning Organization (RPO) Modeling Discussion Group due to the regional nature of haze and the need to create a continental-scale meteorological modeling database that can be used seamlessly with other RPO modeling efforts around the country. The model spatial domain was represented as a 3- dimensional matrix of grid cells with a horizontal grid size of 36 by 36 km and with 19 layers 2-19

37 extending from the surface to the top of the troposphere, where the model surface layer had a nominal thickness of 36 meters and the layer thickness progressively increased toward the model top. Figure 2-7. The WRAP 2002 air quality modeling domain. A key step in the application of the modeling system is the model performance evaluation for the historical episode in which model performance for 2002 was evaluated by comparing simulated aerosol and gas species mass with ambient monitoring data at receptor sites. Ambient monitoring data have been collected by multiple national or regional monitoring networks including the IMPROVE, Clean Air Status and Trends Network (CASTNet), National Acid Deposition Program (NADP) and Speciation Trends Network (STN) for PM species, and the EPA s Air Quality System (AQS) network for gas phase species. Table 2-3 presents a summary of the mean fractional bias and error results for comparison to IMPROVE sites in the WRAP states, and Figures 2-8a and 2-8b show the mean fractional bias (MFB) for the model simulated data to the ambient monitoring data for each of the PM ambient data networks. (Note that not all networks monitor all species). In Table 2-3, the bias is noted as positive (+) or negative (-), indicating the monthly average of the predicted results were higher or lower than monthly measured values. For example, the modeled nitrate was generally higher than measured nitrate in cooler months (Oct Apr), and lower than measured nitrate in warmer months (May Sep). The error presented in the table is approximate (rounded to the nearest 10%), and gives an indication of the magnitude of the uncertainty in the predicted results for each month. Error in modeled sulfate was lowest in the spring and fall months (Apr, May, Sep, Oct) and highest in the winter months (Nov Feb). Complete evaluation results can be found in the 2004 Interim Report for the Western Regional Air Partnership (WRAP) Regional Modeling Center (RMC) ( 2-20

38 Table 2-3 CMAQ Model Performance Compared to IMPROVE Monitoring Data for 2002 Monthly Bias (Positive/Negative) and Error (%) Month bias (+ or -) / % error Species Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Sulfate +/70 +/60 +/50 -/40 -/40 -/50 -/50 -/50 -/40 +/40 +/60 +/70 Nitrate +/100 +/100 +/100 +/100 -/100 -/140 -/150 -/130 -/120 +/100 +/110 +/100 Org. Carbon +/80 +/80 +/70 +/70 +/70 +/70 -/70 +/70 +/70 +/70 +/70 +/80 Elem. Carbon +/70 +/60 -/60 +/60 +/60 -/60 -/70 +/60 -/60 -/60 -/60 +/80 Soil +/130 +/110 +/90 -/90 -/90 -/90 +/80 +/80 +/70 +/80 +/100 +/120 Coarse Mass -/100 -/100 -/100 -/130 -/130 -/130 -/130 -/120 -/120 -/100 -/90 -/90 There are two unique challenges in applying model simulations for visibility SIPs and TIPs. First, because of the long time period being simulated, excessive resources would be required to fine tune the model input data as is typically done in episodic modeling of high pollution events of a few days or weeks duration. This can lead to larger errors in input data in the long term modeling compared to the episodic modeling, with errors occurring in either the magnitude or the spatial location of the model simulated data. Second, episodic modeling has traditionally focused on the most polluted days and model performance criteria have been developed for high pollution events. For example, ozone model performance evaluations routinely apply a filter to exclude all low ozone days in the model validation. For the WRAP modeling, however, the model must be used to simulate both extremes of the most polluted and cleanest days. For both of these reasons it is expected that performance for long term modeling of visibility will yield larger error and bias terms. There are also large uncertainties in the ambient monitoring data themselves. There are typically error and bias terms as large as 30% when comparing measurements from different monitors and different monitoring networks. Finally, there are errors introduced in the model simulation by the mismatch that occurs when the monitored data at a specific point in space is compared with a model simulated concentration that is averaged over the size of the model grid cell. For the WRAP continental domain the point measurement is being compared to a modeled concentration average over an area of 36x36 km. In light of these sources of error and uncertainty, WRAP is continuing to work with EPA to develop guidance for acceptable model performance for visibility simulations. It is not expected that the model will represent accurately both the spatial and temporal location of both low and high visibility events. The goal for the AoH project is for the model to represent the physical and chemical processes in a sufficiently realistic manner that the model can be used to accurately predict in general terms the direction and degree of changes in visibility that result from emissions reductions strategies. 2-21

39 SO4 Monthly Bias, WRAP region NO3 Monthly Bias, WRAP region 150 IMPROVE 150 IMPROVE 100 CASTNET 100 CASTNET 50 STN 50 STN MFB (%) 0 NADP MFB (%) 0 NADP Month Month OC Monthly Bias, WRAP region EC Monthly Bias, WRAP region IMPROVE IMPROVE 100 STN 100 STN MFB (%) 0 MFB (%) Month Month Figure 2-8a. The mean fractional bias (MFB) for the model simulated data to the ambient monitoring data for each of the PM ambient data networks. (Note that not all networks monitor all species).

40 SOIL and CM Monthly Bias, WRAP region PM25 and PM10 Monthly Bias, WRAP region MFB (%) IMPROVE- SOIL IMPROVE- CM MFB (%) IMPROVE- PM25 IMPROVE- PM10 STN- PM Month Month 2-23 Figure 2-8b. The mean fractional bias (MFB) for the model simulated data to the ambient monitoring data for each of the PM ambient data networks. (Note that not all networks monitor all species).

41 2.3.2 Back Trajectory Modeling In support of WRAP s Causes of Haze project, DRI generated meteorological back trajectories for IMPROVE monitoring sites. Back trajectory analyses use interpolated measured or modeled meteorological fields to estimate the most likely central path over geographical areas that provided air to a receptor at a given time. The method essentially follows a parcel of air backward in hourly steps for a specified length of time. Back trajectories are an oversimplification of the atmosphere in that dispersion is not accounted for and the potential source areas contributing to a receptor are underestimated for any given trajectory. A commonly used trajectory model is the Hybrid-Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model developed by the National Oceanic and Atmospheric Administration s (NOAA) Air Resources Laboratory (ARL). HYSPLIT uses archived 3-dimensional meteorological fields generated from observations and short-term forecasts. HYSPLIT can be run to generate forward or backward trajectories using several available meteorological data archives. The data archives used in this analysis were the National Weather Service's National Centers for Environmental Prediction Eta Data Assimilation System (EDAS) and NOAA s Air Resources Laboratory (ARL) FNL Archive. The EDAS fields, used for the continental U.S. sites, are archived by the ARL at a horizontal resolution of 80 km across the continental U.S., plus a buffer zone. The domain of these fields does not include Alaska or Hawaii. The FNL fields, used for Alaska and Hawaii, are archived by the ARL at a horizontal resolution of 190 km over the entire hemisphere. Detailed information regarding the trajectory model and these data sets can be found on NOAA s Web site ( In general, higher resolution fields are desirable to capture smaller scale flow features. Both the EDAS and FNL meteorological input fields represent large-scale flows and cannot accurately represent local to mesoscale phenomena such as topographically influenced flow, nocturnal jets, and seabreeze/landbreeze. Even the 80 km resolution of the EDAS data set is not sufficient to accurately represent flow over complex terrain. In some cases systematic biases may occur that could lead to invalid conclusions regarding source-receptor relationships. DRI did not conduct an evaluation of the meteorological fields used in this analysis, but did recommend an evaluation be considered in future work. This analysis assumes the errors are random and regional to large-scale transport patterns are reasonably accurate averaged over a significant number of trajectories. Eight back trajectories per day, spanning the years 2000 through 2002, were computed for each WRAP Class I area with an IMPROVE sampling site. The major model parameters selected for this analysis include those listed in Table 2-4. Table 2-4 Back Trajectory Model Parameters Selected for AoH Analysis Model Parameter Trajectory duration Top of model domain Vertical motion option Receptor height Meteorological Field Value 192 hours (8 days) backward in time 14,000 meters used model data 500 meters EDAS and FNL (location dependent)

42 The choice of these parameters affects the trajectories generated and the final attribution analyses based on them. In particular, trajectories tend to become increasingly uncertain the further back in time they are used. A duration of eight days was chosen to represent a compromise between higher certainty (shorter duration) and expected atmospheric lifetime of sulfate aerosols (1-2 weeks). Vertical motion in the model is sometimes best represented by following actual vertical motion measurements (represented by model data), surfaces of constant entropy, or surfaces of constant pressure, depending on the meteorological conditions at a given location and time. The impact of receptor height (or end height) on an individual trajectory is also important. Lowending trajectories represent air parcels nearer ground level, and consequently nearer the groundbased samplers. High-ending trajectories may represent more accurate boundary layer flow above the local terrain. Trajectory heights are not constant throughout the trajectory duration and often vary considerably from the receptor height selected. Consequently, trajectories generated for the same location and time but with different receptor heights may indicate significantly different flow patterns for part of or the entire trajectory. The choice of different meteorological fields can yield dissimilar result as well. Based on the three years of individual back trajectories, DRI generated residence time maps. Residence time analysis computes the amount of time (e.g., number of hours) or percent of time an air parcel is in a horizontal grid cell. Plotted on a map, residence time is shown as percent of total hours in each one-degree latitude by one-degree longitude grid cell across the domain. A sample residence time map is presented in Figure 2-9. The values associated with each color in the map legend are percentages (e.g., the residence time associated with the dark blue cells fall in the range 0.28 to 0.84%). Because there are so many grid cells, the percent values are low for any given cell, but the difference in fractional values from region to region is important. Residence time over an area is indicative of general flow patterns, but does not necessarily imply the area contributed significantly to haze compounds at a receptor site since it does not account for emissions and removal processes. Residence time maps should be used in conjunction with emissions information to relate emissions and transport to components of haze at the receptor sites. DRI generated other types of back trajectory summary maps which can be useful in better understanding source attribution. The set of maps are available on DRI s Causes of Haze Web site ( and include: 3-year total residence time maps (described above). Monthly total residence time maps similar to the 3-year maps, but broken out into months to display seasonality of the site s and region s meteorology. Residence time maps associated with the days showing the 20% best and worst extinction and individual species mass during the 3-year period. Residence time difference maps, depicting the difference between all days and the 20% worst extinction and individual species mass during the 3-year period. Positive (negative) values on the map indicate the residence time on the 20% worst days was greater than (less than) on all days. Conditional probability maps, depicting the ratio of the residence time for the 20% worst days to all days. A conditional probability value of 2 (0.5) means that the residence time over an area on worst sulfate days is twice (half) as frequent as on all days.

43 2-26 Figure 2-9. Back trajectory residence time map for Great Sand Dunes NP. The total residence time is presented in percent for 1 degree grid cells. The model parameters used to generate the back trajectories for this map are described in the text.

44 2.4 ATTRIBUTION ASSESSMENTS Tagged Species Source Apportionment (TSSA) Method The Regional Modeling Center performed a Tagged Species Source Apportionment analysis using CMAQ to attribute sulfate and nitrate to geographic source regions. Details beyond the summary given here can be found on the RMC Web site: A variety of modeling and data analysis methods have been used to perform source attribution. Model sensitivity simulations have been used in which a base case model simulation is performed and then a particular source is zeroed out of the emissions. The importance of that source is assessed by evaluating changes in pollutants at the receptor site, calculated as pollutant concentration in the sensitivity case minus the base case. Although sensitivity approaches are widely used and provide valuable information, they also have some disadvantages. For small sources, the model may lack sufficient numerical precision to represent the effect of the source at a receptor site. More importantly, the chemical reactions of some precursor species have a non-linear dependence on ambient species concentrations and are subject to positive and negative feedback effects when a model input parameter is changed. Thus, the change in pollutant concentration at a receptor site in the sensitivity case might not accurately reflect the contribution of that source to the receptor in the base case. Finally, large numbers of sensitivity simulations must be completed to develop source attributions for all of the emissions source categories and source regions of interest. Mass tracking, or tagged-species algorithms, represent a second approach for performing source attribution. Mass tracking approaches do not attempt to predict how the pollutant concentrations will change at a receptor site; instead, they attempt to identify the mass contributions of each source to the pollutant at the receptor site in the base case model simulation. Mass tracking source attribution algorithms have the potential to evaluate the contribution for all sources in a single model simulation. This approach has not been widely used in air quality modeling because of the complexity and difficulty of implementation. The key difference between the sensitivity and mass tracking approaches is that source attribution estimates the contribution of a source to pollutants at a receptor, while sensitivity simulations estimate the change in pollutants at receptor when an individual source is changed by a particular amount. During 2003 and 2004 RMC developed a new algorithm in CMAQ to assess source attribution. The algorithm uses a system of tracers or tagged species to track the chemical transformations, transport and removal of emissions from particular source categories or source locations. This algorithm has been implemented and tested for nitrate-sulfate-ammonia chemistry. For the final Phase I analysis the RMC implemented the TSSA algorithm in a beta release of CMAQ 4.4, using the Pre02c 2002 emissions inventory. The CMAQ TSSA methodology uses tagged species to track the temporal evolution of emissions, chemical transformations, transport and dispersion of mass across the model s spatial domain. Tracers are defined for a subset of key chemical species and for certain pre-defined sub-regions and emissions source categories. In addition, tracers are included to represent mass contributions from the initial conditions and from transport into the model domain from the 2-27

45 boundary conditions. The tracers are initialized with all zero values, except for the initial conditions tracer which is identical to the initial species concentrations. The tracer concentrations are updated at each model time-step over the entire model simulation period. The TSSA routine checks for mass conservation at each time-step and adjusts the mass if needed. Tracers are defined for selected, pre-defined geographical regions. Source regions are defined by assigning a unique numeric code to the model grid cells composing each source area. The source region can be defined to be a single grid cell or any number of grid cells up to the total number of cells used in the spatial domain. Figure 2-10 illustrates the TSSA regions for each state within the WRAP 36 km modeling domain. Emissions attributed to tribal lands are tracked at the state level. In this figure, California is assigned the number 1, Nevada is assigned 2, Utah 6, Arizona 7, etc. The entire source regions of Canada and Mexico are each assigned a unique number. The accuracy of resolving state, county or tribal boundaries is inherently limited by the spatial grid resolution. The 36-km grid might result in some cities or point source from one state being misallocated into an adjacent state. This problem can be mitigated to some extent by assigning the grid cell the numeric code of the state with the dominant contribution to emissions in that cell Figure TSSA source area mapping file. Each source region (state, country) is distinguished by a unique numeric code. Table 2-5 lists the available source category tags for the TSSA analysis. The source tags selected for the AoH Phase I work (Initial Conditions, Boundary Conditions, Mobile Sources, Point Sources, and Other) are indicated in red. The TSSA method was used to attribute particulate nitrate and sulfate, formed from NO x and SO 2 emissions. Due to computational and time limitations only point and mobile sources were tagged. Since Point and Mobile emissions account for approximately 80% of the WRAP NO x and SO 2 emissions, the Other category, 2-28

46 which is intended to represent all sources not explicitly defined by the remaining categories (e.g., Area, Fire, Biogenic, etc.), is expected to contribute approximately 20% of all modeled mass. However, as will be discussed in more detail in Section 4.1, the Other category contribution to WRAP sites ranged from approximately 10 70%. It is not currently understood why the results indicate such a large contribution from Other at many Class I areas. It is possible that errors in the model have accumulated in this category. Further investigation into this problem is required in Phase II of the AoH project (see Recommendations in Section 5.1). Table 2-5 Source Tags Available for Use in the TSSA Simulation Types Source Category Notes ICON ICON Initial Conditions BCON BCON Boundary Conditions Emissions MV_* Mobile sources from any state BG_* Biogenic sources from any state RD_* Paved + unpaved road dust from any state NR_* Non-road dust sources from any state PT_* Point sources from any state AR_* Area sources from any state WF_* Wildland fire from any state AG_* Agricultural fire from any state RX_* Prescribed fire from any state ET_* Total emissions from any state *_WRAP Any type of source category emissions from WRAP domain Others OTHER Any sources other than all of the above * Asterisks represent unique identifiers for each source region (e.g., AZ represents Arizona). RED tags are those selected for the AoH work Trajectory Regression Attribution Method Desert Research Institute performed a Trajectory Regression Analysis (TRA) to attribute sulfate and extinction to geographic source regions. Details beyond the summary given here can be found on the Causes of Haze Assessment (COHA) Web site: Back trajectory analysis is used to estimate the amount of time an air parcel spends over each source region. Trajectory regression analysis is a way to determine a simple relationship between a measured air quality parameter and the amount of time air is flowing across potential source regions on their way to the receptor location. A multiple-linear regression is developed where the dependent variable is the air quality parameter (e.g., sulfate) and the independent variables are the estimates of time air spends over each of a group of specific source regions. 2-29

47 Implicit in the TRA method is the concept that the amount of time air spends over a region determines that region s potential contribution to pollutants measured at the receptor site. This approach is too simplistic to capture the effects of varying atmospheric factors known to influence concentration (e.g., washout by precipitation, enhanced chemistry in clouds, atmospheric mixing, etc.). The amount of emissions captured by a trajectory air parcel depends also on plume rise, air parcel height, and atmospheric stability. An air parcel transported aloft over a source region during stable nighttime conditions may not pick up any surface emissions. Another source of uncertainty results from errors in the estimated movement of air over sources prior to arriving at the receptor sites. When interpreting the results of back trajectory analyses, it is important to understand the density of emissions along and near the back trajectories. Because geographic areas may be high emitters of some haze causing compounds and low for others, trajectory calculations were made and plotted for conditions in addition to high or low estimated extinction, such as high or low sulfate, nitrate, and organic carbon mass. It should also be recognized that the meteorological input fields represent large-scale flows and cannot accurately represent local to mesoscale flows. In some cases systematic biases may occur leading to invalid conclusions regarding source-receptor relationships. In spite of these sources of error, long-term average source region contributions determined by trajectory regression attribution analysis can provide a reasonable picture of potential source regions impacting a given area. The trajectory regression analyses were done two ways, with an additive intercept term and with no intercept. Graphically, the intercept of the multiple linear regression would project through a zero intercept if all the measured parameters were accounted for by the source regions. However, often this is not the case because sources of contaminants may come from outside of the source regions, such as in the case of intercontinental transport of pollutants. Thus, the intercept term is thought to account for both contributions from beyond the selected source regions (analogous to a global background value) and statistical noise from imprecise parameters and an imperfect model. Regression without the intercept forces the sources to account for some of the background contributions, and may overestimate some of the source region contributions. Regression analysis with an intercept may underestimate some source region contributions by incorporating statistical noise in the intercept term. The most reasonable attribution results by the trajectory regression method are likely within a range set by regression with and without an intercept. A feature of the regression analysis methodology is that an uncertainty level and statistical significance are estimated for each regression coefficient, or attribution result. The uncertainty is the standard error for the attribution result, meaning that there is a 67% probability that the true result is within a range of plus or minus the uncertainty around the reported attribution result. The magnitude of the attribution uncertainties are shown as error bars in the attribution bar plots on the COHA Web site. If the uncertainty bar is of a comparable magnitude to the attribution bar height (e.g., ½ or greater than the bar), then the attribution should be considered uncertain. A second method for assessing the statistical uncertainty of each attribution is with the significance or p value for each coefficient. This estimates the probability the regression coefficient value is not statistically different than zero (i.e., that it could be zero). Larger p values correspond to lower probabilities of being significantly different than zero (e.g., p = % probability; p = % probability; and p = % probability of being zero). The smaller the p value the more reliable the regression attribution method from a 2-30

48 statistical perspective. It is suggested that results with p values less than 0.05 (5%) are statistically reliable. TRA results for a given monitoring site are based entirely on the data collected at the site and the relationship of back trajectories to that data. This method does not take into account what emissions are associated with specific regions. Therefore, when a strong source of emissions is near the border or corner of a state, it is possible that air flowing over the source will pick up pollutants but the modeled back trajectories may not accrue sufficient residence time in the source state to correctly attribute the pollutant. This edge effect is shown graphically in Figure Figure Graphical representation of the edge effect. The purple back trajectories may underestimate the impacts from the strong sources within the circles in Nevada and New Mexico because they don t spend much time in those states. When selecting source regions for trajectory regression analysis it is beneficial to have relatively larger source regions at a greater distance from the site than those near the site. Monitoring sites were assigned from 9 to 19 source regions, depending on their location within the WRAP states. The approach chosen to define source regions for the TRA method is outlined below. Figure 2-12 presents a map indicating the selected regions for the Rocky Mountain National Park monitoring site. States containing the monitoring site were divided into quadrants (i.e., NE, SE, SW, and NW) with the origin at the monitoring site. 2-31

49 Each state bordering the state containing the monitoring site was considered a separate source region. All other states beyond the bordering states were combined into four quadrants (i.e., NE, SE, SW, and NW). Mexico, Canada, the Pacific Ocean, the Gulf of Mexico, and the Atlantic Ocean were all considered separate source regions. Alaska & Hawaii were divided into quadrants, and everything outside of those states was divided into quadrants. Figure Map indicating the selected source regions for the Rocky Mountain NP monitoring site. 2-32

50 3.0 ATTRIBUTION OF HAZE WEB SITE DESCRIPTION The overall objective of the Attribution of Haze project is to prepare a policy-level report describing the emission source categories and geographic source regions presently contributing to visibility impairment, at each of the over 100 tribal and mandatory federal Class I areas in the WRAP region. The Attribution of Haze Web site ( was developed to provide report products as well as other pieces of information for each of these Class I areas. The navigational toolbar along the right hand side of the home page contains these major topic headings, which are described below: Attribution of Haze Workgroup Reports Links Tools Attribution of Haze Workgroup: These links provide general background information on the workgroup that oversees the AoH project. A list of all the workgroup members and contact information for each member can be found here as well as meeting notes and various documents related to the workgroup. Reports: These links provide site specific and regional reports, under the following headings: AoH Phase I Report: State Reports: Tribal Reports: Individual Class I Reports: Select the State Reports link to view a list of each state within the WRAP region. Click on an individual state to review its state and regional report. State Reports The state and regional report products present an overview of monitoring and emissions data used in the AoH project and provide the following information: State map with Class I areas identified. Links to Class I area reports. Regional Aerosol Mass and Extinction: This link leads to maps which depict the average mass and extinction (by species), deciview, and standard visual range of the 20% worst days monitored in the WRAP region in Regional Emissions: This link provides maps which depict the 36x36 km gridded emissions summaries for the CMAQ modeling domain. Maps are available for NO x, SO 2, VOC, NH 3, PM 2.5 and PMC emissions, displayed in tons/year. State Emissions: This link provides state specific maps of the same emissions data displayed above, with the addition of Class I areas, major cities, and interstate highways for reference. 3-1

51 Emissions Inventory Information: These links lead to the emissions inventory data used for the AoH project and the most current emissions inventory data in the WRAP Emissions Data Management System (EDMS) database. Apportionment of State Emissions to Class I Areas: These links provide charts and tables displaying the apportionment of state contributions of sulfate, nitrate, and extinction to Class I areas within the WRAP region. A sample state page for North Dakota is presented in Figure Figure 3-1. State report page for North Dakota. Tribal Reports Select the Tribal Reports link to list each tribal Class I area. These Class I area reports are also listed above under the State Reports and the Individual Class I Reports. The Tribal Reports page is shown in Figure

52 Figure 3-2. Tribal report page. Individual Class I Reports The individual Class I area report products combine monitoring, modeling, and emissions data, and associated analyses, for all WRAP tribal and mandatory federal Class I areas. Each report contains the following: A brief description of the Class I area site and associated IMPROVE monitoring location. WinHaze Image: This link provides a split image simulated to represent the 20% best (left) and 20% worst (right) visibility at the Class I area for (Not available for all locations.) Aerosol Summary Products: This link opens a series of aerosol extinction timeline products. Each page contains a stacked bar timeline chart of reconstructed aerosol extinction for 2002 with the 20% best and worst days indicated with a "B" and "W" (for monitored data only). Each page also contains two pie charts showing the extinction budgets for the 20% best and worst days in Timelines are available for monitored and modeled data. Source Apportionment Comparisons: This link opens a series of source apportionment comparison products. CMAQ Tagged Species Source Apportionment (TSSA) was performed for sulfate and nitrate mass. Trajectory Regression Analysis (TRA) was performed using monitored data and HYSPLIT back trajectories for sulfate and extinction. Each of these products divides the modeling domain into distinct source regions: the state the Class I area resides in; the surrounding states; the remaining WRAP states; international and other categories. Each product presents results for either different methods or different pollutant parameters. 3-3

53 Regional Modeling Center TSSA Data Products: This link points to the Regional Modeling Center's TSSA results Web page. Causes of Haze Data Products: Selecting this link opens a Class I area-specific site and data descriptions page on the Causes of Haze Web site. The page contains a series of links to descriptive pages for aerosol data, meteorological conditions, emissions information, and trajectory regression analysis and uncertainty results. Causes of Haze Back Trajectory Map Gallery: This link opens a Class I area-specific map gallery of back trajectory products on the Causes of Haze Web site. Monitoring Site Information on VIEWS Web Site: This link opens a Class I areaspecific Monitoring Site Browser on the VIEWS Web site. Available on this page are monitoring site specifications, photos, maps, and site and network histories. Emissions Inventory Information: This links leads to the emissions inventory data used for the AoH project and the most current emissions inventory data in the WRAP Emissions Data Management System (EDMS) database. A sample Class I area page for Rocky Mountain National Park is presented in Figure Figure 3-3. Class I area report page for Rocky Mountain National Park. 3-4

54 Links: This section provides links to the following: Causes of Haze Project Regional Modeling Center AoH Emissions Inventories WRAP EDMS VIEWS visibility Web site IMPROVE Web site WinHaze Visual Air Quality Modeler APACE Website Report Broken Links Tools: This section contains links for WRAP forum guidelines, a WRAP events calendar, and a document submittal tool. 3-5

55 4.0 REGIONAL ATTRIBUTIONS AND ASSESSMENTS This section describes sulfate and nitrate source attribution results for individual Class I areas, and regional assessments that were performed to understand the impacts of fire, carbon, and dust, for which no attribution analyses were performed. 4.1 SOURCE ATTRIBUTION RESULTS Two methods of source attribution techniques were used to attribute pollutants at Class I areas to geographic source regions, as described in Section 2.0. The tagged species source apportionment (TSSA) method was adopted as the primary attribution tool because, of the two methods, it was conceptually the most rigorous: it applied chemistry, meteorology, and transport, and relied on known (though sometimes incomplete) emissions inventories. This method was applied to all Class I areas, even those without IMPROVE monitors. The trajectory regression analysis (TRA) method relied on back trajectories and monitoring data, and was quite useful as a check on, and supplement to, the TSSA method, especially where the CMAQ model performance was questionable Regional Attributions Taken together, the TSSA and TRA attribution results for WRAP Class I areas provide site-specific and regional information about the geographic source regions contributing to sulfate and nitrate in the Western U.S. Attribution results indicate that particulate sulfate and nitrate are regional in scope, and all geographic regions analyzed are responsible for some level of contribution. Distance is often, but not always a factor in determining the relationship between source regions and impacts at Class I areas. Comparisons between TSSA and TRA attribution results are often reasonable, given the differences that exist between methods. Some of the attribution results that do not seem reasonable or need further examination are: At some Class I areas the Other source category contributes more than the expected 20% to sulfate and nitrate, particularly in coastal states. The TSSA results attribute less sulfate and nitrate to Canada and Mexico than expected, particularly in border states. TRA results indicate somewhat higher attributions from Canada and Mexico. The lower contributions in TSSA results may be reflective of the state of the emissions inventories for these regions. The TRA results attribute significant sulfate to the Pacific Coastal and more distant Pacific regions. Since no emissions inventories for these regions were developed for Phase I, there is no way for the TSSA method to account for these emissions. Some sites show differences between TSSA and TRA attributions for single states. In some cases these may be explained by the edge effect described in Section 2, in which strong emissions sources near state boundaries are not correctly accounted for by the TRA method. 4-1

56 This section presents a series of tables and figures highlighting some of the attribution results. Due to the volume of information, all tables and figures have been placed after the text for this section. Class I areas were assigned to specific groups based on a review of TSSA sulfate and nitrate attribution results. The motivation for grouping sites by species attribution was to understand if specific source regions might impact multiple Class I areas in a similar way, thus allowing those Class I areas to be treated together in Phase II. The groupings introduced here are a compromise between the desire to have as few groups as possible, and the goal of highlighting significant differences. Sites within groups may have very different monitored aerosol characteristics (e.g., grouping by sulfate and nitrate attribution does not take into account the possibly different impacts of carbon or dust). Table 4-1 presents the Class I area group characteristics as the range of TSSA sulfate and nitrate contributions, rounded to the nearest 5%. To help identify regions of large impacts, contributions in the range 10 25% are highlighted blue, and those ranging above 25% are highlighted red. As an example, the Class I areas in Group 1 exhibit a strong sulfate attribution signal almost entirely split between Washington (29 71%) and Other (25 63%). In contrast, the Class I areas in Group 2 exhibit strong sulfate signals from Oregon (28 40%), Washington (16 21%), and Other (36 45%). Figure 4-1 presents a map of the Class I area groups. Together, Table 4-1 and Figure 4-1 present a concise summary of the TSSA attribution results across the WRAP region. Tables 4-2a and 4-2b present TSSA sulfate and nitrate attribution results as a percent contribution of modeled species mass for all Class I areas. The Class I areas are organized by group along the left side of the table; geographic source regions are listed along the top, including the categories of EA US (eastern U.S.), Other (those emissions not tagged in the model) and Minor (the sum of all contributions below the top 20). The Initial Conditions and Boundary Conditions contributions, not shown in these tables, were always between 0 3%. (These tables and tables which further apportion results by Point and Mobile source categories can be downloaded in Excel format from the project Web site: The relative impact of each source region on a given Class I area can be reviewed by looking across the appropriate row. The impact of a given state on all Class I areas can be reviewed by looking down the appropriate column. For example, in Table 4-2a, the first row indicates that at Mount Rainier NP the largest two source regions/categories for sulfate are the state of Washington (estimated contribution 71%) and Other (estimated contribution 25%). In the same table, looking down along the Washington column, Washington s contribution of sulfate is at least 10% for some Class I areas in Washington, Oregon, Idaho, California, and Montana. Graphical examples of Washington s sulfate and nitrate contributions to Class I areas are presented in Figures 4-2a and 4-2b. Tables 4-3a and 4-3b present TSSA sulfate and nitrate attribution results as a percent contribution to aerosol extinction for all Class I areas. These tables have the same format as the previous tables, but the distinction is important. Tables 4-3a and 4-3b combine the information in Tables 4-2a and 4-2b with aerosol extinction estimated from monitored data. By their nature, these percentages are smaller than those in the previous tables. To use the same example as 4-2

57 above, looking at the first row in Table 4-3a indicates that at Mount Rainier NP the largest two source regions/categories for sulfate contributing to aerosol extinction are the state of Washington (estimated contribution 31%) and Other (estimated contribution 11%). Washington s contribution of sulfate extinction to total aerosol extinction is at least 5% for some Class I areas in Washington, Oregon, and Montana. Note that Washington s contribution to aerosol extinction at all Class I areas in California is approximately 1% or less. (These tables are available on the project Web site with the addition of the attribution in inverse megameters, in the same Excel spreadsheet noted above: As can be seen in these tables, the contribution from the Other category is often quite large, sometimes as high as 70%. As described in Section 2, the TSSA method was used to attribute particulate nitrate and sulfate, formed from NO x and SO 2 emissions. Due to computational and time limitations only point and mobile sources were tagged. Since Point and Mobile emissions account for approximately 80% of the WRAP NO x and SO 2 emissions, the Other category, which is intended to represent all sources not explicitly defined by the remaining categories (e.g., Area, Fire, Biogenic, etc.), is expected to contribute approximately 20% of all modeled mass. It is not currently understood why the results indicate such a large contribution from Other at many Class I areas. It is possible that errors in the model have accumulated in this category. Figure 4-3 presents maps showing the spatial distribution of the TSSA sulfate and nitrate contributions at each Class I area attributed to the Other category. The largest sulfate Other contributions generally occur in coastal states, the smallest in the southeastern WRAP states. The nitrate Other contributions are generally smaller, with the largest occurring at a few sites in northern California and southern Oregon. Further investigation into this problem is required in Phase II of the AoH project (see Recommendations in Section 5). Assignment of Class I areas to groups based on TRA results for sulfate attribution was not done. Instead, these results have been used as a check on TSSA results. TRA analyses for each Class I area are accompanied by various statistics that can be used to evaluate the overall uncertainty of the results. (These statistical measurements are available for each IMPROVE monitoring site on the COHA Web site, most easily accessed from the AoH Class I Report pages, as described in Section 3.) Table 4-4 presents a summary of the TRA uncertainty and highlights the differences between TSSA and TRA. Included in the table are: R 2 values (square of the correlation coefficients) associated with each analysis. Higher R 2 values indicate a better, less uncertain regression. The non-intercept method yielded values in the range.67.90; the intercept method in the range This implies a much higher confidence in the non-intercept method. Qualitative measure of the general uncertainty of the individual source region attributions. This is based on the reported p-values, uncertainty values, and the length of aerosol data record: + indicates high certainty in the results, medium certainty, and - low certainty. Quantitative measures of uncertainty are available for each geographic source region attribution for each Class I area on the COHA Web site. 4-3

58 Calculated intercept value. This probably represents some combination of global background and method uncertainty. It is not clear how the intercept values should be interpreted in light of analysis uncertainty. Summary of the differences between TSSA and TRA results. This is presented as a list of regions where the TRA attribution percentage is either greater or less than the TSSA attribution by at least 10. (The non-intercept method was used for this comparison because it performed better statistically than the intercept method.) In many cases the largest differences exist between TSSA and TRA attributions to Canada, Mexico, and the Pacific Ocean. These differences are likely due to uncertainties in the Phase I emissions inventories for these regions. Some differences are believed to be related to the edge effect described in Section 2. (A specific example of a site likely affected by the edge effect is presented later in this section.) Tables summarizing the TRA attribution results by source region for each Class I area are not presented in this report but can be found on the project Web site in the same Excel spreadsheet noted above ( The degree to which TRA results corroborate TSSA results needs to be reviewed for each Class I area, and states and tribes are encouraged to perform these reviews for Class I areas under their jurisdiction. 4-4

59 Table 4-1 Class I Area Group Characteristics Range of TSSA Sulfate and Nitrate Contributions in Percent (Rounded to Nearest 5%) 4-5 Group Species AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10 Group 11 Group 12 Group 13 Group 14 Group 15 Group 16 Group 17 Group 18 Group 19 Group 20 SO NO SO NO SO NO SO NO SO NO SO NO SO NO SO NO SO4 SO4 SO4 SO4 SO4 SO4 SO4 SO4 SO4 SO4 SO4 SO NO3 NO3 NO3 NO3 NO3 NO3 NO3 NO3 NO3 NO3 NO3 NO BLUE text indicates ranges ~10 25% RED text indicates ranges exceeding 25%

60 Figure 4-1. Initial grouping of Class I areas by TSSA source region attribution of sulfate and nitrate. No consideration of monitoring data was made to generate these groupings. 4-6

61 Table 4-2a TSSA Analysis Attribution Results for SO4 (%) Based on 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 1 Mount Rainier National Park WA MORA1 MORA Glacier Peak Wilderness Area WA GLPE NOCA North Cascades National Park WA NOCA1 NOCA Olympic National Park WA OLYM1 OLYM Pasayten Wilderness Area WA PASA1 PASA Alpine Lakes Wilderness Area WA SNPA1 SNPA Spokane Tribe of Indians WA SPOK1 SPOK Goat Rocks Wilderness Area WA GORO WHPA Mount Adams Wilderness Area WA MOAD WHPA Group 2 Diamond Peak Wilderness Area OR DIPE CRLA Mount Hood Wilderness Area OR MOHO1 MOHO Three Sisters Wilderness Area OR THSI1 THSI Mount Jefferson Wilderness Area OR MOJE THSI Mount Washington Wilderness Area OR MOWA THSI Group 3 Mountain Lakes Wilderness Area OR MOLA CRLA Crater Lake National Park OR CRLA1 CRLA Gearheart Mountain Wilderness Area OR GEAR CRLA Marble Mountain Wilderness Area CA MAMO TRIN Group 4 Kalmiopsis Wilderness Area OR KALM1 KALM Redwood National Park CA REDW1 REDW Group 5 Mokelumne Wilderness Area CA MOKE BLIS Desolation Wilderness Area CA BLIS1 BLIS Dome Lands Wilderness Area CA DOME1 DOME Hoover Wilderness Area CA HOOV1 HOOV Ansel Adams Wilderness Area CA ANAD KAIS John Muir Wilderness Area CA JOMU KAIS Kaiser Wilderness Area CA KAIS1 KAIS Lava Beds National Monument CA LABE1 LABE South Warner Wilderness Area CA SOWA LABE Caribou Wilderness Area CA CARI LAVO Thousand Lakes Wilderness Area CA THLA LAVO Lassen Volcanic National Park CA LAVO1 LAVO Kings Canyon National Park CA KICA SEQU Sequoia National Park CA SEQU1 SEQU Yolla Bolly-Middle Eel Wilderness Area CA YOBO TRIN Yosemite National Park CA YOSE1 YOSE Emigrant Wilderness Area CA EMIG YOSE Group 6 Agua Tibia Wilderness Area CA AGTI1 AGTI Joshua Tree National Park CA JOSH1 JOSH (cont.)

62 Table 4-2a (cont.) TSSA Analysis Attribution Results for SO4 (%) Based on 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Pinnacles National Monument CA PINN1 PINN Ventana Wilderness Area CA VENT PINN Point Reyes National Seashore CA PORE1 PORE San Rafael Wilderness Area CA RAFA1 RAFA San Gabriel Wilderness Area CA SAGA1 SAGA Cucamonga Wilderness Area CA CUCA SAGA San Gorgonio Wilderness Area CA SAGO1 SAGO San Jacinto Wilderness Area CA SAJA SAGO Group 7 Hells Canyon Wilderness Area OR HECA1 HECA Jarbidge Wilderness Area NV JARB1 JARB Eagle Cap Wilderness Area OR EACA STAR Strawberry Mountain Wilderness Area OR STMO STAR Group 8 Bridger Wilderness Area WY BRID1 BRID Fitzpatrick Wilderness Area WY FITZ BRID Craters of the Moon National Monument ID CRMO1 CRMO Washakie Wilderness Area WY WASH NOAB North Absaroka Wilderness Area WY NOAB1 NOAB Sawtooth Wilderness Area ID SAWT1 SAWT Red Rocks Lakes Wilderness Area MT RERO YELL Grand Teton National Park WY GRTE YELL Yellowstone National Park WY YELL2 YELL Teton Wilderness Area WY TETO YELL Group 9 Cabinet Mountains Wilderness Area MT CABI1 CABI Confederated Salish and Kootenai Tribes MT FLAT1 FLAT Bob Marshall Wilderness Area MT BOMA MONT Mission Mountains Wilderness Area MT MIMO MONT Scapegoat Wilderness Area MT SCAP MONT Anaconda-Pintler Wilderness Area MT ANPI SULA Selway-Bitterroot Wilderness Area MT SULA1 SULA Group 10 Gates of the Mountains Wilderness Area MT GAMO1 GAMO Glacier National Park MT GLAC1 GLAC Group 11 Fort Peck Tribes MT FOPE1 FOPE Lostwood Wilderness ND LOST1 LOST Medicine Lake Wilderness Area MT MELA1 MELA Northern Cheyenne Tribe MT NOCH1 NOCH Theodore Roosevelt National Park ND THRO1 THRO U.L. Bend Wilderness Area MT ULBE1 ULBE Group 12 Badlands National Park SD BADL1 BADL Wind Cave National Park SD WICA1 WICA (cont.)

63 Table 4-2a (cont.) TSSA Analysis Attribution Results for SO4 (%) Based on 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 13 Mount Zirkel Wilderness Area CO MOZI1 MOZI Rawah Wilderness Area CO RAWA MOZI Rocky Mountain National Park CO ROMO1 ROMO Eagles Nest Wilderness Area CO EANE WHRI Group 14 Great Sand Dunes National Park CO GRSA1 GRSA Black Canyon of Gunnison National Park CO BLCA WEMI La Garita Wilderness Area CO LAGA WEMI Flat Tops Wilderness Area CO FLTO WHRI Maroon Bells-Snowmass Wilderness Area CO MABE WHRI West Elk Wilderness Area CO WEEL WHRI Group 15 Bandelier National Monument NM BAND1 BAND Bosque del Apache Wilderness Area NM BOAP1 BOAP Mesa Verde National Park CO MEVE1 MEVE San Pedro Parks Wilderness Area NM SAPE1 SAPE Weminuche Wilderness Area CO WEMI1 WEMI Wheeler Peak Wilderness Area NM WHPE1 WHPE Pecos Wilderness Area NM PECO WHPE Group 16 Carlsbad Caverns National Park NM CACA GUMO Salt Creek Wilderness Area NM SACR1 SACR White Mountain Wilderness Area NM WHIT1 WHIT Group 17 Mount Baldy Wilderness Area AZ BALD1 BALD Galiuro Wilderness Area AZ GALI CHIR Chiricahua National Monument AZ CHIR1 CHIR Chiricahua Wilderness Area AZ CHIR CHIR Gila Wilderness Area NM GICL1 GICL Mazatzal Wilderness Area AZ MAZA IKBA Pine Mountain Wilderness Area AZ PINE IKBA Yavapai-Apache Nation AZ YAAP IKBA Petrified Forest National Park AZ PEFO1 PEFO Saguaro National Park - East AZ SAGU1 SAGU Saguaro National Park - West AZ SAWE1 SAWE Sierra Ancha Wilderness Area AZ SIAN1 SIAN Superstition Wilderness Area AZ TONT1 TONT Group 18 Grand Canyon National Park AZ GRCA2 GRCA Hualapai Tribe AZ HUAL GRCA Grand Canyon National Park - In Canyon AZ INGA1 INGA Sycamore Canyon Wilderness Area AZ SYCA1 SYCA (cont.)

64 Table 4-2a (cont.) TSSA Analysis Attribution Results for SO4 (%) Based on 20% Highest Modeled Extinction Days in 2002 Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 19 Bryce Canyon National Park UT BRCA1 BRCA Capitol Reef National Park UT CAPI1 CAPI Zion National Park UT ZION1 ZION Group 20 Canyonlands National Park UT CANY1 CANY Arches National Park UT ARCH1 CANY

65 Table 4-2b TSSA Analysis Attribution Results for NO3 (%) Based on 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 1 Mount Rainier National Park WA MORA1 MORA Glacier Peak Wilderness Area WA GLPE NOCA North Cascades National Park WA NOCA1 NOCA Olympic National Park WA OLYM1 OLYM Pasayten Wilderness Area WA PASA1 PASA Alpine Lakes Wilderness Area WA SNPA1 SNPA Spokane Tribe of Indians WA SPOK1 SPOK Goat Rocks Wilderness Area WA GORO WHPA Mount Adams Wilderness Area WA MOAD WHPA Group 2 Diamond Peak Wilderness Area OR DIPE CRLA Mount Hood Wilderness Area OR MOHO1 MOHO Three Sisters Wilderness Area OR THSI1 THSI Mount Jefferson Wilderness Area OR MOJE THSI Mount Washington Wilderness Area OR MOWA THSI Group 3 Mountain Lakes Wilderness Area OR MOLA CRLA Crater Lake National Park OR CRLA1 CRLA Gearheart Mountain Wilderness Area OR GEAR CRLA Marble Mountain Wilderness Area CA MAMO TRIN Group 4 Kalmiopsis Wilderness Area OR KALM1 KALM Redwood National Park CA REDW1 REDW Group 5 Mokelumne Wilderness Area CA MOKE BLIS Desolation Wilderness Area CA BLIS1 BLIS Dome Lands Wilderness Area CA DOME1 DOME Hoover Wilderness Area CA HOOV1 HOOV Ansel Adams Wilderness Area CA ANAD KAIS John Muir Wilderness Area CA JOMU KAIS Kaiser Wilderness Area CA KAIS1 KAIS Lava Beds National Monument CA LABE1 LABE South Warner Wilderness Area CA SOWA LABE Caribou Wilderness Area CA CARI LAVO Thousand Lakes Wilderness Area CA THLA LAVO Lassen Volcanic National Park CA LAVO1 LAVO Kings Canyon National Park CA KICA SEQU Sequoia National Park CA SEQU1 SEQU Yolla Bolly-Middle Eel Wilderness Area CA YOBO TRIN Yosemite National Park CA YOSE1 YOSE Emigrant Wilderness Area CA EMIG YOSE Group 6 Agua Tibia Wilderness Area CA AGTI1 AGTI Joshua Tree National Park CA JOSH1 JOSH (cont.)

66 Table 4-2b (cont.) TSSA Analysis Attribution Results for NO3 (%) Based on 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Pinnacles National Monument CA PINN1 PINN Ventana Wilderness Area CA VENT PINN Point Reyes National Seashore CA PORE1 PORE San Rafael Wilderness Area CA RAFA1 RAFA San Gabriel Wilderness Area CA SAGA1 SAGA Cucamonga Wilderness Area CA CUCA SAGA San Gorgonio Wilderness Area CA SAGO1 SAGO San Jacinto Wilderness Area CA SAJA SAGO Group 7 Hells Canyon Wilderness Area OR HECA1 HECA Jarbidge Wilderness Area NV JARB1 JARB Eagle Cap Wilderness Area OR EACA STAR Strawberry Mountain Wilderness Area OR STMO STAR Group 8 Bridger Wilderness Area WY BRID1 BRID Fitzpatrick Wilderness Area WY FITZ BRID Craters of the Moon National Monument ID CRMO1 CRMO Washakie Wilderness Area WY WASH NOAB North Absaroka Wilderness Area WY NOAB1 NOAB Sawtooth Wilderness Area ID SAWT1 SAWT Red Rocks Lakes Wilderness Area MT RERO YELL Grand Teton National Park WY GRTE YELL Yellowstone National Park WY YELL2 YELL Teton Wilderness Area WY TETO YELL Group 9 Cabinet Mountains Wilderness Area MT CABI1 CABI Confederated Salish and Kootenai Tribes MT FLAT1 FLAT Bob Marshall Wilderness Area MT BOMA MONT Mission Mountains Wilderness Area MT MIMO MONT Scapegoat Wilderness Area MT SCAP MONT Anaconda-Pintler Wilderness Area MT ANPI SULA Selway-Bitterroot Wilderness Area MT SULA1 SULA Group 10 Gates of the Mountains Wilderness Area MT GAMO1 GAMO Glacier National Park MT GLAC1 GLAC Group 11 Fort Peck Tribes MT FOPE1 FOPE Lostwood Wilderness ND LOST1 LOST Medicine Lake Wilderness Area MT MELA1 MELA Northern Cheyenne Tribe MT NOCH1 NOCH Theodore Roosevelt National Park ND THRO1 THRO U.L. Bend Wilderness Area MT ULBE1 ULBE Group 12 Badlands National Park SD BADL1 BADL Wind Cave National Park SD WICA1 WICA (cont.)

67 Table 4-2b (cont.) TSSA Analysis Attribution Results for NO3 (%) Based on 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 13 Mount Zirkel Wilderness Area CO MOZI1 MOZI Rawah Wilderness Area CO RAWA MOZI Rocky Mountain National Park CO ROMO1 ROMO Eagles Nest Wilderness Area CO EANE WHRI Group 14 Great Sand Dunes National Park CO GRSA1 GRSA Black Canyon of Gunnison National Park CO BLCA WEMI La Garita Wilderness Area CO LAGA WEMI Flat Tops Wilderness Area CO FLTO WHRI Maroon Bells-Snowmass Wilderness Area CO MABE WHRI West Elk Wilderness Area CO WEEL WHRI Group 15 Bandelier National Monument NM BAND1 BAND Bosque del Apache Wilderness Area NM BOAP1 BOAP Mesa Verde National Park CO MEVE1 MEVE San Pedro Parks Wilderness Area NM SAPE1 SAPE Weminuche Wilderness Area CO WEMI1 WEMI Wheeler Peak Wilderness Area NM WHPE1 WHPE Pecos Wilderness Area NM PECO WHPE Group 16 Carlsbad Caverns National Park NM CACA GUMO Salt Creek Wilderness Area NM SACR1 SACR White Mountain Wilderness Area NM WHIT1 WHIT Group 17 Mount Baldy Wilderness Area AZ BALD1 BALD Galiuro Wilderness Area AZ GALI CHIR Chiricahua National Monument AZ CHIR1 CHIR Chiricahua Wilderness Area AZ CHIR CHIR Gila Wilderness Area NM GICL1 GICL Mazatzal Wilderness Area AZ MAZA IKBA Pine Mountain Wilderness Area AZ PINE IKBA Yavapai-Apache Nation AZ YAAP IKBA Petrified Forest National Park AZ PEFO1 PEFO Saguaro National Park - East AZ SAGU1 SAGU Saguaro National Park - West AZ SAWE1 SAWE Sierra Ancha Wilderness Area AZ SIAN1 SIAN Superstition Wilderness Area AZ TONT1 TONT Group 18 Grand Canyon National Park AZ GRCA2 GRCA Hualapai Tribe AZ HUAL GRCA Grand Canyon National Park - In Canyon AZ INGA1 INGA Sycamore Canyon Wilderness Area AZ SYCA1 SYCA (cont.)

68 Table 4-2b (cont.) TSSA Analysis Attribution Results for NO3 (%) Based on 20% Highest Modeled Extinction Days in 2002 Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 19 Bryce Canyon National Park UT BRCA1 BRCA Capitol Reef National Park UT CAPI1 CAPI Zion National Park UT ZION1 ZION Group 20 Canyonlands National Park UT CANY1 CANY Arches National Park UT ARCH1 CANY

69 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Washington Highest Attribution to Class I Areas SO4 CMAQ TSSA Results 20% Worst Modeled Extinction Days, 2002 Mount Rainier NP Alpine Lakes W Goat Rocks W Spokane Tribe of Indians Mount Adams W Glacier Peak W Olympic NP Pasayten W North Cascades NP Mount Jefferson W Mount Hood W Three Sisters W Diamond Peak W Mount Washington W Hells Canyon W Crater Lake NP Strawberry Mountain W Eagle Cap W Mountain Lakes W Gearheart Mountain W Kalmiopsis W Sawtooth W Marble Mountain W Redwood NP Lava Beds NM South Warner W Cabinet Mountains W Bob Marshall W Confederated Salish and Kootenai Tribes Mission Mountains W Anaconda-Pintler W Selway-Bitterroot W Scapegoat W Glacier NP Jarbidge W WA OR ID CA MT NV Figure 4-2a. TSSA sulfate contributions to Class I areas attributed to Washington. Percent of Modeled Concentration at CIA 4-15

70 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Washington Highest Attribution to Class I Areas NO3 CMAQ TSSA Results 20% Worst Modeled Extinction Days, 2002 Mount Rainier NP Alpine Lakes W Glacier Peak W Goat Rocks W Mount Adams W North Cascades NP Olympic NP Pasayten W Spokane Tribe of Indians Mount Hood W Mount Jefferson W Three Sisters W Mount Washington W Diamond Peak W Strawberry Mountain W Eagle Cap W Kalmiopsis W Hells Canyon W Crater Lake NP Gearheart Mountain W Sawtooth W Cabinet Mountains W Bob Marshall W Mission Mountains W Glacier NP Confederated Salish and Kootenai Tribes Anaconda-Pintler W Selway-Bitterroot W Scapegoat W Gates of the Mountains W Fort Peck Tribes U.L. Bend W Medicine Lake W Theodore Roosevelt NP Lostwood Wilderness Jarbidge W Washakie W Teton W North Absaroka W Yellowstone NP Bridger W Fitzpatrick W WA OR ID MT ND NV WY Figure 4-2b. TSSA nitrate contributions to Class I areas attributed to Washington. Percent of Modeled Concentration at CIA 4-16

71 Table 4-3a TSSA Analysis SO4 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 1 Mount Rainier National Park WA MORA1 MORA Glacier Peak Wilderness Area WA GLPE NOCA North Cascades National Park WA NOCA1 NOCA Olympic National Park WA OLYM1 OLYM Pasayten Wilderness Area WA PASA1 PASA Alpine Lakes Wilderness Area WA SNPA1 SNPA Spokane Tribe of Indians WA SPOK1 SPOK Goat Rocks Wilderness Area WA GORO WHPA Mount Adams Wilderness Area WA MOAD WHPA Group 2 Diamond Peak Wilderness Area OR DIPE CRLA Mount Hood Wilderness Area OR MOHO1 MOHO Three Sisters Wilderness Area OR THSI1 THSI Mount Jefferson Wilderness Area OR MOJE THSI Mount Washington Wilderness Area OR MOWA THSI Group 3 Mountain Lakes Wilderness Area OR MOLA CRLA Crater Lake National Park OR CRLA1 CRLA Gearheart Mountain Wilderness Area OR GEAR CRLA Marble Mountain Wilderness Area CA MAMO TRIN Group 4 Kalmiopsis Wilderness Area OR KALM1 KALM Redwood National Park CA REDW1 REDW Group 5 Mokelumne Wilderness Area CA MOKE BLIS Desolation Wilderness Area CA BLIS1 BLIS Dome Lands Wilderness Area CA DOME1 DOME Hoover Wilderness Area CA HOOV1 HOOV Ansel Adams Wilderness Area CA ANAD KAIS John Muir Wilderness Area CA JOMU KAIS Kaiser Wilderness Area CA KAIS1 KAIS Lava Beds National Monument CA LABE1 LABE South Warner Wilderness Area CA SOWA LABE Caribou Wilderness Area CA CARI LAVO Thousand Lakes Wilderness Area CA THLA LAVO Lassen Volcanic National Park CA LAVO1 LAVO Kings Canyon National Park CA KICA SEQU Sequoia National Park CA SEQU1 SEQU Yolla Bolly-Middle Eel Wilderness Area CA YOBO TRIN Yosemite National Park CA YOSE1 YOSE Emigrant Wilderness Area CA EMIG YOSE Group 6 Agua Tibia Wilderness Area CA AGTI1 AGTI Joshua Tree National Park CA JOSH1 JOSH (cont.)

72 Table 4-3a (cont.) TSSA Analysis SO4 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Pinnacles National Monument CA PINN1 PINN Ventana Wilderness Area CA VENT PINN Point Reyes National Seashore CA PORE1 PORE San Rafael Wilderness Area CA RAFA1 RAFA San Gabriel Wilderness Area CA SAGA1 SAGA Cucamonga Wilderness Area CA CUCA SAGA San Gorgonio Wilderness Area CA SAGO1 SAGO San Jacinto Wilderness Area CA SAJA SAGO Group 7 Hells Canyon Wilderness Area OR HECA1 HECA Jarbidge Wilderness Area NV JARB1 JARB Eagle Cap Wilderness Area OR EACA STAR Strawberry Mountain Wilderness Area OR STMO STAR Group 8 Bridger Wilderness Area WY BRID1 BRID Fitzpatrick Wilderness Area WY FITZ BRID Craters of the Moon National Monument ID CRMO1 CRMO Washakie Wilderness Area WY WASH NOAB North Absaroka Wilderness Area WY NOAB1 NOAB Sawtooth Wilderness Area ID SAWT1 SAWT Grand Teton National Park WY GRTE YELL Red Rocks Lakes Wilderness Area MT RERO YELL Yellowstone National Park WY YELL2 YELL Teton Wilderness Area WY TETO YELL Group 9 Cabinet Mountains Wilderness Area MT CABI1 CABI Confederated Salish and Kootenai Tribes MT FLAT1 FLAT Bob Marshall Wilderness Area MT BOMA MONT Mission Mountains Wilderness Area MT MIMO MONT Scapegoat Wilderness Area MT SCAP MONT Anaconda-Pintler Wilderness Area MT ANPI SULA Selway-Bitterroot Wilderness Area MT SULA1 SULA Group 10 Gates of the Mountains Wilderness Area MT GAMO1 GAMO Glacier National Park MT GLAC1 GLAC Group 11 Fort Peck Tribes MT FOPE1 FOPE Lostwood Wilderness ND LOST1 LOST Medicine Lake Wilderness Area MT MELA1 MELA Northern Cheyenne Tribe MT NOCH1 NOCH Theodore Roosevelt National Park ND THRO1 THRO U.L. Bend Wilderness Area MT ULBE1 ULBE Group 12 Badlands National Park SD BADL1 BADL Wind Cave National Park SD WICA1 WICA (cont.)

73 Table 4-3a (cont.) TSSA Analysis SO4 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 13 Mount Zirkel Wilderness Area CO MOZI1 MOZI Rawah Wilderness Area CO RAWA MOZI Rocky Mountain National Park CO ROMO1 ROMO Eagles Nest Wilderness Area CO EANE WHRI Group 14 Great Sand Dunes National Park CO GRSA1 GRSA Black Canyon of Gunnison National Park CO BLCA WEMI La Garita Wilderness Area CO LAGA WEMI Flat Tops Wilderness Area CO FLTO WHRI Maroon Bells-Snowmass Wilderness Area CO MABE WHRI West Elk Wilderness Area CO WEEL WHRI Group 15 Bandelier National Monument NM BAND1 BAND Bosque del Apache Wilderness Area NM BOAP1 BOAP Mesa Verde National Park CO MEVE1 MEVE San Pedro Parks Wilderness Area NM SAPE1 SAPE Weminuche Wilderness Area CO WEMI1 WEMI Wheeler Peak Wilderness Area NM WHPE1 WHPE Pecos Wilderness Area NM PECO WHPE Group 16 Carlsbad Caverns National Park NM CACA GUMO Salt Creek Wilderness Area NM SACR1 SACR White Mountain Wilderness Area NM WHIT1 WHIT Group 17 Mount Baldy Wilderness Area AZ BALD1 BALD Galiuro Wilderness Area AZ GALI CHIR Chiricahua National Monument AZ CHIR1 CHIR Chiricahua Wilderness Area AZ CHIR CHIR Gila Wilderness Area NM GICL1 GICL Mazatzal Wilderness Area AZ MAZA IKBA Pine Mountain Wilderness Area AZ PINE IKBA Yavapai-Apache Nation AZ YAAP IKBA Petrified Forest National Park AZ PEFO1 PEFO Saguaro National Park - East AZ SAGU1 SAGU Saguaro National Park - West AZ SAWE1 SAWE Sierra Ancha Wilderness Area AZ SIAN1 SIAN Superstition Wilderness Area AZ TONT1 TONT Group 18 Grand Canyon National Park AZ GRCA2 GRCA Hualapai Tribe AZ HUAL GRCA Grand Canyon National Park - In Canyon AZ INGA1 INGA Sycamore Canyon Wilderness Area AZ SYCA1 SYCA (cont.)

74 Table 4-3a (cont.) TSSA Analysis SO4 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in 2002 Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 19 Bryce Canyon National Park UT BRCA1 BRCA Capitol Reef National Park UT CAPI1 CAPI Zion National Park UT ZION1 ZION Group 20 Canyonlands National Park UT CANY1 CANY Arches National Park UT ARCH1 CANY

75 Table 4-3b TSSA Analysis NO3 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 1 Mount Rainier National Park WA MORA1 MORA Glacier Peak Wilderness Area WA GLPE NOCA North Cascades National Park WA NOCA1 NOCA Olympic National Park WA OLYM1 OLYM Pasayten Wilderness Area WA PASA1 PASA Alpine Lakes Wilderness Area WA SNPA1 SNPA Spokane Tribe of Indians WA SPOK1 SPOK Goat Rocks Wilderness Area WA GORO WHPA Mount Adams Wilderness Area WA MOAD WHPA Group 2 Diamond Peak Wilderness Area OR DIPE CRLA Mount Hood Wilderness Area OR MOHO1 MOHO Three Sisters Wilderness Area OR THSI1 THSI Mount Jefferson Wilderness Area OR MOJE THSI Mount Washington Wilderness Area OR MOWA THSI Group 3 Mountain Lakes Wilderness Area OR MOLA CRLA Crater Lake National Park OR CRLA1 CRLA Gearheart Mountain Wilderness Area OR GEAR CRLA Marble Mountain Wilderness Area CA MAMO TRIN Group 4 Kalmiopsis Wilderness Area OR KALM1 KALM Redwood National Park CA REDW1 REDW Group 5 Mokelumne Wilderness Area CA MOKE BLIS Desolation Wilderness Area CA BLIS1 BLIS Dome Lands Wilderness Area CA DOME1 DOME Hoover Wilderness Area CA HOOV1 HOOV Ansel Adams Wilderness Area CA ANAD KAIS John Muir Wilderness Area CA JOMU KAIS Kaiser Wilderness Area CA KAIS1 KAIS Lava Beds National Monument CA LABE1 LABE South Warner Wilderness Area CA SOWA LABE Caribou Wilderness Area CA CARI LAVO Thousand Lakes Wilderness Area CA THLA LAVO Lassen Volcanic National Park CA LAVO1 LAVO Kings Canyon National Park CA KICA SEQU Sequoia National Park CA SEQU1 SEQU Yolla Bolly-Middle Eel Wilderness Area CA YOBO TRIN Yosemite National Park CA YOSE1 YOSE Emigrant Wilderness Area CA EMIG YOSE Group 6 Agua Tibia Wilderness Area CA AGTI1 AGTI Joshua Tree National Park CA JOSH1 JOSH (cont.)

76 Table 4-3b (cont.) TSSA Analysis NO3 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Pinnacles National Monument CA PINN1 PINN Ventana Wilderness Area CA VENT PINN Point Reyes National Seashore CA PORE1 PORE San Rafael Wilderness Area CA RAFA1 RAFA San Gabriel Wilderness Area CA SAGA1 SAGA Cucamonga Wilderness Area CA CUCA SAGA San Gorgonio Wilderness Area CA SAGO1 SAGO San Jacinto Wilderness Area CA SAJA SAGO Group 7 Hells Canyon Wilderness Area OR HECA1 HECA Jarbidge Wilderness Area NV JARB1 JARB Eagle Cap Wilderness Area OR EACA STAR Strawberry Mountain Wilderness Area OR STMO STAR Group 8 Bridger Wilderness Area WY BRID1 BRID Fitzpatrick Wilderness Area WY FITZ BRID Craters of the Moon National Monument ID CRMO1 CRMO Washakie Wilderness Area WY WASH NOAB North Absaroka Wilderness Area WY NOAB1 NOAB Sawtooth Wilderness Area ID SAWT1 SAWT Grand Teton National Park WY GRTE YELL Red Rocks Lakes Wilderness Area MT RERO YELL Yellowstone National Park WY YELL2 YELL Teton Wilderness Area WY TETO YELL Group 9 Cabinet Mountains Wilderness Area MT CABI1 CABI Confederated Salish and Kootenai Tribes MT FLAT1 FLAT Bob Marshall Wilderness Area MT BOMA MONT Mission Mountains Wilderness Area MT MIMO MONT Scapegoat Wilderness Area MT SCAP MONT Anaconda-Pintler Wilderness Area MT ANPI SULA Selway-Bitterroot Wilderness Area MT SULA1 SULA Group 10 Gates of the Mountains Wilderness Area MT GAMO1 GAMO Glacier National Park MT GLAC1 GLAC Group 11 Fort Peck Tribes MT FOPE1 FOPE Lostwood Wilderness ND LOST1 LOST Medicine Lake Wilderness Area MT MELA1 MELA Northern Cheyenne Tribe MT NOCH1 NOCH Theodore Roosevelt National Park ND THRO1 THRO U.L. Bend Wilderness Area MT ULBE1 ULBE Group 12 Badlands National Park SD BADL1 BADL Wind Cave National Park SD WICA1 WICA (cont.)

77 Table 4-3b (cont.) TSSA Analysis NO3 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 13 Mount Zirkel Wilderness Area CO MOZI1 MOZI Rawah Wilderness Area CO RAWA MOZI Rocky Mountain National Park CO ROMO1 ROMO Eagles Nest Wilderness Area CO EANE WHRI Group 14 Great Sand Dunes National Park CO GRSA1 GRSA Black Canyon of Gunnison National Park CO BLCA WEMI La Garita Wilderness Area CO LAGA WEMI Flat Tops Wilderness Area CO FLTO WHRI Maroon Bells-Snowmass Wilderness Area CO MABE WHRI West Elk Wilderness Area CO WEEL WHRI Group 15 Bandelier National Monument NM BAND1 BAND Bosque del Apache Wilderness Area NM BOAP1 BOAP Mesa Verde National Park CO MEVE1 MEVE San Pedro Parks Wilderness Area NM SAPE1 SAPE Weminuche Wilderness Area CO WEMI1 WEMI Wheeler Peak Wilderness Area NM WHPE1 WHPE Pecos Wilderness Area NM PECO WHPE Group 16 Carlsbad Caverns National Park NM CACA GUMO Salt Creek Wilderness Area NM SACR1 SACR White Mountain Wilderness Area NM WHIT1 WHIT Group 17 Mount Baldy Wilderness Area AZ BALD1 BALD Galiuro Wilderness Area AZ GALI CHIR Chiricahua National Monument AZ CHIR1 CHIR Chiricahua Wilderness Area AZ CHIR CHIR Gila Wilderness Area NM GICL1 GICL Mazatzal Wilderness Area AZ MAZA IKBA Pine Mountain Wilderness Area AZ PINE IKBA Yavapai-Apache Nation AZ YAAP IKBA Petrified Forest National Park AZ PEFO1 PEFO Saguaro National Park - East AZ SAGU1 SAGU Saguaro National Park - West AZ SAWE1 SAWE Sierra Ancha Wilderness Area AZ SIAN1 SIAN Superstition Wilderness Area AZ TONT1 TONT Group 18 Grand Canyon National Park AZ GRCA2 GRCA Hualapai Tribe AZ HUAL GRCA Grand Canyon National Park - In Canyon AZ INGA1 INGA Sycamore Canyon Wilderness Area AZ SYCA1 SYCA (cont.)

78 Table 4-3b (cont.) TSSA Analysis NO3 Contribution to Aerosol Extinction (%) on the 20% Highest Modeled Extinction Days in 2002 Class I Area State Site Code IMPROVE Code AZ CA CO ID MT NV NM ND OR SD UT WA WY Mex Can EA US Other Minor Group 19 Bryce Canyon National Park UT BRCA1 BRCA Capitol Reef National Park UT CAPI1 CAPI Zion National Park UT ZION1 ZION Group 20 Canyonlands National Park UT CANY1 CANY Arches National Park UT ARCH1 CANY

79 TSSA Sulfate Contribution Attributed to the "Other" (non-point, non-mobile) Source Category TSSA Sulfate Contribution by "Other" 10% 40% 70% TSSA Nitrate Contribution Attributed to the "Other" (non-point, non-mobile) Source Category TSSA Nitrate Contribution by "Other" 10% 40% 70% Figure 4-3. TSSA Sulfate (top) and nitrate (bottom) contribution at each Class I area attributed to the Other (non-point, non-mobile) source category. 4-25

80 Table 4-4 Uncertainty Analysis of TRA Results and Comparison to TSSA Results 4-26 Class I Area Site Code IMPROVE Code State Non-Int. Method R 2 Value Confidence Differences between TSSA and TRA Analysis Intercept Method Non-Int. Method Intercept Method Intercept Value (%) TRA < TSSA by more than ~10% TRA > TSSA by more than ~10% Group 1 Glacier Peak WA GLPE NOCA1 WA WA Canada, Pacific Goat Rocks WA GORO WHPA1 WA OR Canada Mount Adams WA MOAD WHPA1 WA OR WA, Canada, Pacific Mount Rainier NP MORA1 MORA1 WA WA Canada, Pacific North Cascades NP NOCA1 NOCA1 WA Canada, Pacific Olympic NP OLYM1 OLYM1 WA WA, Canada, Pacific Pasayten WA PASA1 PASA1 WA Canada, Pacific Alpine Lakes WA SNPA1 SNPA1 WA WA, OR Canada, Pacific Spokane Tribe of Indians SPOK1 SPOK1 WA WA Canada Group 2 Diamond Peak WA DIPE CRLA1 OR WA OR, Pacific Mount Hood WA MOHO1 MOHO1 OR OR WA, Canada, Pacific Mount Jefferson WA MOJE THSI1 OR Canada, Pacific Mount Washington WA MOWA THSI1 OR OR, Canada, Pacific Three Sisters WA THSI1 THSI1 OR OR, Canada, Pacific Group 3 Marble Mountain WA MAMO TRIN1 CA CA, Pacific Crater Lake NP CRLA1 CRLA1 OR OR, Pacific Gearheart Mountain WA GEAR CRLA1 OR CA OR, Pacific Mountain Lakes WA MOLA CRLA1 OR OR, Pacific Group 4 Redwood NP REDW1 REDW1 CA CA, OR, Pacific Kalmiopsis WA KALM1 KALM1 OR OR, Pacific Group 5 Ansel Adams WA ANAD KAIS1 CA OR, Canada, Pacific Desolation WA BLIS1 BLIS1 CA Pacific Caribou WA CARI LAVO1 CA Pacific Dome Land WA DOME1 DOME1 CA Pacific Emigrant WA EMIG YOSE1 CA Pacific Hoover WA HOOV1 HOOV1 CA CA, OR, NW, East, Pacific John Muir WA JOMU KAIS1 CA OR, Canada, Pacific Kaiser WA KAIS1 KAIS1 CA OR, Canada, Pacific Kings Canyon NP KICA SEQU1 CA CA, Pacific Lava Beds NM LABE1 LABE1 CA OR, Pacific Lassen Volcanic NP LAVO1 LAVO1 CA Pacific Mokelumne WA MOKE BLIS1 CA Pacific Sequoia NP SEQU1 SEQU1 CA CA, Pacific South Warner WA SOWA LABE1 CA OR, Pacific Thousand Lakes WA THLA LAVO1 CA Pacific Yolla Bolly-Middle Eel WA YOBO TRIN1 CA OR, Pacific Yosemite NP YOSE1 YOSE1 CA Pacific (cont.)

81 Table 4-4 (cont.) Uncertainty Analysis of TRA Results and Comparison to TSSA Results 4-27 Class I Area Site Code IMPROVE Code State Non-Int. Method R 2 Value Intercept Method Confidence Non-Int. Method Intercept Method Intercept Value (%) Differences between TSSA and TRA Analysis TRA < TSSA by more than ~10% TRA > TSSA by more than ~10% Group 6 Agua Tibia WA AGTI1 AGTI1 CA CA East, Pacific Cucamonga WA CUCA SAGA1 CA CA OR, Pacific Joshua Tree NP JOSH1 JOSH1 CA CA Mexico, Pacific Pinnacles NM PINN1 PINN1 CA CA Pacific Point Reyes NS PORE1 PORE1 CA CA Pacific San Rafael WA RAFA1 RAFA1 CA CA Pacific San Gabriel WA SAGA1 SAGA1 CA CA OR, Pacific San Gorgonio WA SAGO1 SAGO1 CA CA Pacific San Jacinto WA SAJA SAGO1 CA CA Pacific Ventana WA VENT PINN1 CA CA Pacific Group 7 Jarbidge WA JARB1 JARB1 NV ID, NWUS CA, Pacific Eagle Cap WA EACA STAR1 OR ID OR, Canada, Pacific Hells Canyon WA HECA1 HECA1 OR WA OR, ID Strawberry Mountain WA STMO STAR1 OR CA OR, Canada Group 8 Craters of the Moon NM CRMO1 CRMO1 ID ID OR Sawtooth WA SAWT1 SAWT1 ID OR, Pacific Red Rocks Lakes WA RERO YELL2 MT MT, WY, SWUS NWUS, Canada Bridger WA BRID1 BRID1 WY ID Pacific Fitzpatrick WA FITZ BRID1 WY ID Pacific Grand Teton NP GRTE YELL2 WY NWUS North Absaroka WA NOAB1 NOAB1 WY MT WY, NWUS, Canada Teton WA TETO YELL2 WY NWUS Washakie WA WASH NOAB1 WY ID, MT NWUS, Canada Yellowstone NP YELL1 YELL2 WY NWUS Group 9 Anaconda-Pintler WA ANPI SULA1 MT ID MT, NWUS, Canada, Pacific Bob Marshall WA BOMA MONT1 MT MT, Canada Cabinet Mountains WA CABI1 CABI1 MT SD MT, Canada, Pacific Confederated Salish and FLAT1 FLAT1 MT N/A N/A N/A N/A N/A N/A N/A Kootenai Tribes Mission Mountains WA MIMO MONT1 MT Canada Scapegoat WA SCAP MONT1 MT NWUS Selway-Bitterroot WA SULA1 SULA1 MT ID MT, NWUS, Canada, Pacific Group 10 Gates of the Mountains WA GAMO1 GAMO1 MT MT NWUS, Canada Glacier NP GLAC1 GLAC1 MT MT Canada Group 11 Fort Peck Tribes FOPE1 FOPE1 MT N/A N/A N/A N/A N/A N/A N/A Medicine Lake WA MELA1 MELA1 MT ND Canada Northern Cheyenne Tribe NOCH1 NOCH1 MT MT, ND WY, SD, East, Canada U.L. Bend WA ULBE1 ULBE1 MT ND MT, Canada Lostwood Wilderness LOST1 LOST1 ND Canada Theodore Roosevelt NP THRO1 THRO1 ND Canada (cont.)

82 Table 4-4 (cont.) Uncertainty Analysis of TRA Results and Comparison to TSSA Results 4-28 Class I Area Site Code IMPROVE Code State Non-Int. Method R 2 Value Intercept Method Confidence Non-Int. Method Intercept Method Intercept Value (%) Differences between TSSA and TRA Analysis TRA < TSSA by more than ~10% TRA > TSSA by more than ~10% Group 12 Badlands NP BADL1 BADL1 SD WY East, Canada Wind Cave NP WICA1 WICA1 SD WY SWUS, Canada Group 13 Eagles Nest WA EANE WHRI1 CO WY, NWUS Mexico Mount Zirkel WA MOZI1 MOZI1 CO CO Mexico, Pacific Rawah WA RAWA MOZI1 CO WY Mexico, Pacific Rocky Mountain NP ROMO1 ROMO1 CO CO, WY Group 14 Black Canyon of The Gunnison NP BLCA WEMI1 CO Pacific Flat Tops WA FLTO WHRI1 CO WY Great Sand Dunes NP GRSA1 GRSA1 CO La Garita WA LAGA WEMI1 CO NM Pacific Maroon Bells-Snowmass WA MABE WHRI1 CO CO West Elk WA WEEL WHRI1 CO CO Group 15 Mesa Verde NP MEVE1 MEVE1 CO NM AZ, Mexico, Pacific Weminuche WA WEMI1 WEMI1 CO NM CO, Pacific Bandelier NM BAND1 BAND1 NM NM East Bosque del Apache WA BOAP1 BOAP1 NM NM, AZ East Pecos WA PECO WHPE1 NM NM East, Pacific, Gulf San Pedro Parks WA SAPE1 SAPE1 NM NM East Wheeler Peak WA WHPE1 WHPE1 NM NM, AZ, CO East Group 16 Carlsbad Caverns NP CACA GUMO1 NM N/A N/A N/A N/A N/A N/A N/A Salt Creek WA SACR1 SACR1 NM NM White Mountain WA WHIT1 WHIT1 NM NM East, Gulf Group 17 Mount Baldy WA BALD1 BALD1 AZ AZ Mexico, Pacific Chiricahua WA CHIR CHIR1 AZ AZ, NM East, Pacific Chiricahua NM CHIR1 CHIR1 AZ AZ, NM East, Pacific Galiuro WA GALI CHIR1 AZ AZ, NM East, Pacific Mazatzal WA MAZA IKBA1 AZ AZ Mexico, Pacific Petrified Forest NP PEFO1 PEFO1 AZ AZ, NM East, Mexico, Pacific Pine Mountain WA PINE IKBA1 AZ AZ Mexico, Pacific Saguaro NP - East SAGU1 SAGU1 AZ AZ East, Mexico, Pacific Saguaro NP - West SAWE1 SAWE1 AZ AZ East, Pacific Sierra Ancha WA SIAN1 SIAN1 AZ AZ East, Mexico, Pacific Superstition WA TONT1 TONT1 AZ AZ Mexico, Pacific Yavapai-Apache Nation YAAP SYCA1 AZ AZ, CA, NV Mexico, Pacific Gila WA GICL1 GICL1 NM AZ East, Pacific (cont.)

83 Table 4-4 (cont.) Uncertainty Analysis of TRA Results and Comparison to TSSA Results Class I Area Site Code IMPROVE Code State Non-Int. Method R 2 Value Intercept Method Confidence Non-Int. Method Intercept Method Intercept Value (%) Differences between TSSA and TRA Analysis TRA < TSSA by more than ~10% TRA > TSSA by more than ~10% Group 18 Grand Canyon NP GRCA2 GRCA2 AZ NV, UT Mexico, Pacific Hualapai Tribe HUAL GRCA2 AZ CA, NV Mexico, Pacific Grand Canyon NP - In Canyon INGA1 INGA1 AZ NV, UT Mexico, Pacific Sycamore Canyon WA SYCA1 SYCA1 AZ CA, NV Mexico, Pacific Group 19 Bryce Canyon NP BRCA1 BRCA1 UT NV, SWUS East, Mexico, Pacific Capital Reef NP CAPI1 CAPI1 UT UT AZ, Mexico, Pacific Zion NP ZION1 ZION1 UT NV Mexico, Pacific Group 20 Arches NP ARCH CANY1 UT UT AZ, East, Mexico, Pacific Canyonlands NP CANY1 CANY1 UT UT AZ, East, Mexico, Pacific 4-29

84 4.1.2 Results For Selected Sites Five sites were selected to showcase the interpretation of available AoH Phase I 2002 data for each Class I area. The methods outlined here do not represent an exhaustive assessment for any one location, but show how materials available on the AoH Web site can be applied to the question of source attribution at Class I areas. Additional information, available on the RMC Web site, the COHA Web site, work anticipated in Phase II work, and other sources, can and should be used to supplement this material. Rocky Mountain National Park According to IMPROVE monitoring data, the average aerosol extinction for the 20% worst visibility days at Rocky Mountain National Park (NP) is 43 Mm -1. The contribution from ammonium sulfate is approximately 24%, or about 10 Mm -1. The contribution from ammonium nitrate is approximately 20%, or 9 Mm -1. This can be seen in Figures 4-4a and 4-4b, which present timelines of IMPROVE monitoring data (a) and CMAQ model results (b) for A general sense of model performance at this site can be gauged by comparing the timeline plots. It is difficult to fine tune the model for an entire year, expecting good model performance during periods of both high and low extinction. The model clearly does not predict the monitoring data day-to-day, nor the does it yield similar aerosol extinction averages of the 20% worst visibility days, although the range of extinction for those days is similar. (Note that the best and worst days for each timeline are determined by monitored and modeled data, respectively.) Comparisons between the timelines should focus on whether the species seasonal trends and episodes are similar. At Rocky Mountain NP, episodes of high organic material extinction occur in both timelines in the summer and early fall months, and episodes of high ammonium nitrate occur in both timelines during spring and late fall. The magnitude of the modeled ammonium nitrate is higher than that of the monitored data, which is consistent with the model evaluation presented in Section 2. Monitored and modeled sulfate is similar year round, with the monitored data slightly higher than predicted. (Detailed model performance is available at the RMC Web site: Figure 4-5a presents the attribution results for sulfate from the TSSA (top) and TRA (bottom) methods. Both methods identify Colorado as the most significant geographic source of sulfate (TSSA estimates ~40% contribution; TRA estimates ~31% contribution). The remaining sulfate is attributed to nearby states and further U.S. regions (up to ~10% contribution for a single geographic source region), with the largest discrepancy between the contributions identified with Wyoming (TSSA ~11%; TRA ~1%). The Other contribution in the TSSA results is ~18%, which is close to the expected value of about 20%. Figure 4-5b presents the source apportionment results for sulfate (top) and nitrate (bottom) from the TSSA method. The results for sulfate and nitrate show a similar pattern of source strength from all geographic regions, with these exceptions: point sources dominate the sulfate attribution; mobile sources dominate the nitrate attribution; and the estimated contributions from Colorado differ (sulfate ~40%; nitrate ~28%). The Other contribution for both is close to the expected 20%. Review of the state (Figures 4-6a and 4-6b) and regional SO 2 and NO x emissions maps (Figures 2-2a and 2-2b) confirms that there are significant sources of both species within Colorado and nearby states, and within the area of meteorological influence suggested by the 4-30

85 residence time back trajectory map for Rocky Mountain NP (Figure 4-7). The color scaling on the residence time map indicates the fraction of the total time that back trajectory paths fell in a given grid cell. The darker blue regions indicate predominant flow patterns from west of the park. Additional trajectory maps (sulfate difference maps and conditional probability maps), available on the COHA Web site ( can be reviewed to better understand the relationship between high/low sulfate loading and historical wind patterns. Figure 4-8 presents a split image simulated to represent the 20% best (left) and 20% worst (right) visibility at Rocky Mountain NP for 2002, based on IMPROVE monitoring data. This image does not support attribution results, but can be used to illustrate the differences in haze impacts on the best and worst days. This image was created using WinHaze Visual Air Quality Modeler (Ver ). Viewing a split image generated by WinHaze will enhance the perceptibility of the modeled visual air quality differences because attention is drawn to the contrasting conditions at the dividing line between each simulated image. To mitigate this problem, a black bar has been added to the split image, effectively removing the dividing line. 4-31

86 Figure 4-4a timeline of IMPROVE monitoring data for Rocky Mountain NP.

87 Figure 4-4b timeline of CMAQ model results for Rocky Mountain NP.

88 O1.pdf Figure 4-5a. Sulfate TSSA and TRA source apportionment method comparison for Rocky Mountain NP. 4-34

89 O1.pdf Figure 4-5b. Sulfate and nitrate TSSA source apportionment results for Rocky Mountain NP. 4-35

90 Colorado SO 2 Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Sulfur oxide gases (SOx) are formed when sulfur containing fuels, such as oil or coal, are burned, when gasoline is extracted from oil or when metals are extracted from ore. In Colorado, 2002 emissions of SO 2 were dominated by point sources. SO 2 dissolves in water vapor to form acid, and contributes to the formation of sulfate compounds (e.g. (NH 4 ) 2 SO 4 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. SO 2 Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Colorado 2002 SO 2 Emissions 124 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire 0% A N Figure 4-6a. Colorado emissions map for SO 2. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-36

91 Colorado NO X Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Nitrogen oxides (NO X ) form when fuel is burned at high temperatures. In Colorado, 2002 emissions of NO X were dominated by mobile sources (on-road and off-road) and point sources (industrial, commercial, and residential sources that burn fuel). NO X emissions are highly reactive and can form nitrate compounds (e.g. NH 4 NO 3 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. NO X Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% Colorado 2002 NO x Emissions 370 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire Biogenics 10% 0% A N Figure 4-6b. Colorado emissions map for NO x. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-37

92 Figure 4-7. Residence time map for Rocky Mountain NP ( ). The location of the IMPROVE monitoring site is marked in red. Darker blue colors indicate predominant flow patterns. Figure 4-8. Simulated visibility based on the 20% best days (left) and 20% worst days (right). Extinction values taken from the monitoring data pie charts in Figure 4-4a. 4-38

93 Yellowstone National Park According to IMPROVE monitoring data, the average aerosol extinction for the 20% worst visibility days at Yellowstone NP is 22 Mm -1. The contribution from ammonium sulfate is approximately 21%, or about 5 Mm -1. The contribution from ammonium nitrate is approximately 9%, or 2 Mm -1. This can be seen in Figures 4-4a through 4-4c, which present timelines of IMPROVE monitoring data (a), raw IMPROVE monitoring data (b), which contains available data from clogged filters, and CMAQ model results (c) for Note that on the raw data plot, the average of the 20% worst aerosol extinction days is 33 Mm -1, 50% higher than shown on the previous timeline, but the total contribution from ammonium sulfate and ammonium nitrate does not change significantly. The largest difference between the plots are the two large organic samples in mid-july that drive the changes in the aerosol extinction pie charts. (Additional discussion of the impacts of clogged filters in the IMPROVE monitoring network is presented in Section ) A general sense of model performance at this site can be gauged by comparing the timeline plots. It is difficult to fine tune the model for an entire year, expecting good model performance during periods of both high and low extinction. The model clearly does not predict the monitoring data day-to-day, nor does it yield similar aerosol extinction averages of the 20% worst visibility days, although the range of extinction for those days is similar for the raw data timeline and the model results timeline. (Note that the best and worst days for each timeline are determined by monitored and modeled data, respectively.) Comparisons between the timelines should focus on whether the species seasonal trends and episodes are similar. At Yellowstone NP, episodes of high organic material extinction occur in both timelines in the summer and early fall months, with the large July organic event captured on the raw plot predicted in the model. Episodes of high ammonium nitrate occur in both timelines during spring and late fall. The magnitude of the modeled ammonium nitrate is higher than that of the monitored data, which is consistent with the model evaluation presented in Section 2. Monitored and modeled sulfate is similar year round, except for the somewhat higher model predictions in November and December. (Detailed model performance is available at the RMC Web site: Figure 4-10a presents the attribution results for sulfate from the TSSA (top) and TRA (bottom) methods. Both methods identify Idaho as the most significant geographic source of sulfate (TSSA estimates ~22% contribution; TRA estimates ~28% contribution). The remaining sulfate is attributed to the state of Wyoming, nearby states and further U.S. regions (up to ~21% contribution for a single geographic source region), with the largest discrepancy between the contributions identified with the northwest U.S. region (TSSA ~9%; TRA ~21%). The TRA results indicate a contribution from the Pacific Ocean of approximately 13%. This source region was not evaluated by TSSA. The Other contribution in the TSSA results is ~30%, which is somewhat higher than the expected value of about 20%. Figure 4-10b presents the source apportionment results for sulfate (top) and nitrate (bottom) from the TSSA method. The results for sulfate and nitrate show a similar pattern of source strength from all geographic regions except that point sources dominate the sulfate attribution and mobile sources dominate the nitrate attribution. There are differences in the estimated contributions from all boundary states: Wyoming (sulfate ~11%; nitrate ~5%), Utah 4-39

94 (sulfate ~8%; nitrate ~15%), Idaho (sulfate ~21%; nitrate ~13%), and Montana (sulfate ~9%; nitrate ~5%). The Other contribution for both is close to the expected 20%. Review of the state (Figures 4-11a and 4-11b) and regional SO 2 and NO x emissions maps (Figures 2-2a and 2-2b) confirms that there are significant sources of both species within Wyoming and nearby states, and within the area of meteorological influence suggested by the residence time back trajectory map for Yellowstone NP (Figure 4-12). The color scaling on the residence time map indicates the fraction of the total time that back trajectory paths fell in a given grid cell. The darker blue regions indicate predominant flow patterns from west and southwest of the park. Additional trajectory maps (sulfate difference maps and conditional probability maps), available on the COHA Web site ( can be reviewed to better understand the relationship between high/low sulfate loading and historical wind patterns. An image simulating various aerosol conditions using WinHaze Visual Air Quality Modeler (Ver ) is not available for this site. 4-40

95 Figure 4-9a timeline of IMPROVE monitoring data for Yellowstone NP.

96 Figure 4-9b timeline of raw IMPROVE monitoring data for Yellowstone NP. Data set includes available data from days with clogged filters.

97 Figure 4-9c timeline of CMAQ model results for Yellowstone NP.

98 2.pdf Figure 4-10a. Sulfate TSSA and TRA source apportionment method comparison for Yellowstone NP. 4-44

99 2.pdf Figure 4-10b. Sulfate and nitrate TSSA source apportionment results for Yellowstone NP. 4-45

100 Wyoming SO 2 Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Sulfur oxide gases (SOx) are formed when sulfur containing fuels, such as oil or coal, are burned, when gasoline is extracted from oil or when metals are extracted from ore. In Wyoming, 2002 emissions of SO 2 were dominated by point sources. SO 2 dissolves in water vapor to form acid, and contributes to the formation of sulfate compounds (e.g. (NH 4 ) 2 SO 4 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. SO 2 Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Wyoming 2002 SO 2 Emissions 149 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire 0% A N Figure 4-11a. Wyoming emissions map for SO 2. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-46

101 Wyoming NO X Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Nitrogen oxides (NO X ) form when fuel is burned at high temperatures. In Wyoming, 2002 emissions of NO X were dominated by mobile sources (on-road and off-road) and point (industrial, commercial, and residential sources that burn fuel) and area sources. NO X emissions are highly reactive and can form nitrate compounds (e.g. NH 4 NO 3 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. NO X Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% Wyoming 2002 NO x Emissions 285 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire Biogenics 10% 0% A N Figure 4-11b. Wyoming emissions map for NO x. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-47

102 Figure Residence time map for Yellowstone NP ( ). The location of the IMPROVE monitoring site is marked in red. Darker blue colors indicate predominant flow patterns. 4-48

103 Mount Rainier National Park According to IMPROVE monitoring data, the average aerosol extinction for the 20% worst visibility days at Mount Rainier NP is 47 Mm -1. The contribution from ammonium sulfate is approximately 45%, or about 21 Mm -1. The contribution from ammonium nitrate is approximately 12%, or 6 Mm -1. This can be seen in Figures 4-13a and 4-13b, which present timelines of IMPROVE monitoring data (a) and CMAQ model results (b) for A general sense of model performance at this site can be gauged by comparing the timeline plots. It is difficult to fine tune the model for an entire year, expecting good model performance during periods of both high and low extinction. The model clearly does not predict the monitoring data day-to-day, nor does it yield similar aerosol extinction averages of the 20% worst visibility days. (Note that the best and worst days for each timeline are determined by monitored and modeled data, respectively.) Comparisons between the timelines should focus on whether the species seasonal trends and episodes are similar. At Mount Rainier NP, organic material and ammonium sulfate may be reasonably predicted by the model in terms of there seasonal magnitudes. However, modeled ammonium nitrate is over predicted on average by about a factor of 8. This raises concerns about the TSSA method to attribute nitrate at Mount Rainier. As discussed below, the TSSA method attributes sulfate and nitrate nearly identically, and this may indicate that while the model performance for nitrate is poor, the attribution results could be reasonable. (Detailed model performance is available at the RMC Web site: Figure 4-14a presents the attribution results for sulfate from the TSSA (top) and TRA (bottom) methods. Both methods identify Washington as the most significant geographic source of sulfate (TSSA estimates ~71% contribution; TRA estimates ~51% contribution). The largest discrepancy is between the contributions attributed to Canada (TSSA ~1%; TRA ~21%). The lower contribution in TSSA results may be reflective of the state of the emissions inventories for Canada. The TRA results indicate a contribution from the Pacific Ocean of approximately 28%. This source region was not evaluated by TSSA. The Other contribution in the TSSA results is ~25%, which is close to the expected value of about 20%. Figure 4-14b presents the source apportionment results for sulfate (top) and nitrate (bottom) from the TSSA method. The results for sulfate and nitrate show a similar pattern of source strength from all geographic regions with a much larger fraction of mobile sources in the nitrate attribution. Review of the state (Figures 4-15a and 4-15b) and regional SO 2 and NO x emissions maps (Figures 2-2a and 2-2b) confirms that there are significant sources of both species within Washington and nearby states, and within the area of meteorological influence suggested by the residence time back trajectory map for Mount Rainier NP (Figure 4-16). The color scaling on the residence time map indicates the fraction of the total time that back trajectory paths fell in a given grid cell. The darker blue regions indicate predominant flow patterns from northeast, northwest, west, and southwest of the park. Additional trajectory maps (sulfate difference maps and conditional probability maps), available on the COHA Web site ( can be reviewed to better understand the relationship between high/low sulfate loading and historical wind patterns. An image simulating various aerosol conditions using WinHaze Visual Air Quality Modeler (Ver ) is not available for this site. 4-49

104 Figure 4-13a timeline of IMPROVE monitoring data for Mount Rainier NP.

105 Figure 4-13b timeline of CMAQ model results for Mount Rainier NP.

106 A1.pdf Figure 4-14a. Sulfate TSSA and TRA source apportionment method comparison for Mount Rainier NP. 4-52

107 A1.pdf Figure 4-14b. Sulfate and nitrate TSSA source apportionment results for Mount Rainier NP. 4-53

108 Washington SO 2 Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Sulfur oxide gases (SOx) are formed when sulfur containing fuels, such as oil or coal, are burned, when gasoline is extracted from oil or when metals are extracted from ore. In Washington, 2002 emissions of SO 2 were dominated by point sources. SO 2 dissolves in water vapor to form acid, and contributes to the formation of sulfate compounds (e.g. (NH 4 ) 2 SO 4 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. SO 2 Emissions Washington 2002 SO 2 Emissions 87 thousand tons/year 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire 0% A N Figure 4-15a. Washington emissions map for SO 2. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-54

109 Washington NO X Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Nitrogen oxides (NO X ) form when fuel is burned at high temperatures. In Washington, 2002 emissions of NO X were dominated by mobile sources (onroad and off-road) and point sources (industrial, commercial, and residential sources that burn fuel). NO X emissions are highly reactive and can form nitrate compounds (e.g. NH 4 NO 3 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. NO X Emissions Washington 2002 NO x Emissions 355 thousand tons/year 100% 90% 80% 70% 60% 50% 40% 30% 20% Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire Biogenics 10% 0% A N Figure 4-15b. Washington emissions map for NO x. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-55

110 Figure Residence time map for Mount Rainier NP ( ). The location of the IMPROVE monitoring site is marked in red. Darker blue colors indicate predominant flow patterns. 4-56

111 Yosemite National Park According to IMPROVE monitoring data, the average aerosol extinction for the 20% worst visibility days at Yosemite NP is 57 Mm -1. The contribution from ammonium sulfate is approximately 15%, or about 9 Mm -1. The contribution from ammonium nitrate is approximately 17%, or 10 Mm -1. This can be seen in Figures 4-17a and 4-17b, which present timelines of IMPROVE monitoring data (a) and CMAQ model results (b) for A general sense of model performance at this site can be gauged by comparing the timeline plots. It is difficult to fine tune the model for an entire year, expecting good model performance during periods of both high and low extinction. The model clearly does not predict the monitoring data day-to-day, nor does it yield similar aerosol extinction averages of the 20% worst visibility days. (Note that the best and worst days for each timeline are determined by monitored and modeled data, respectively.) Comparisons between the timelines should focus on whether the species seasonal trends and episodes are similar. At Yosemite NP, episodes of high organic material extinction occur in both timelines in the summer and early fall months, although the monitored organic material is somewhat higher than predicted by the model. Episodes of high ammonium nitrate occur in both timelines during spring and late fall, with monitored values often lower than model predictions. Monitored and modeled sulfate is similar year round, with the monitored data somewhat higher than predicted. (Detailed model performance is available at the RMC Web site: Figure 4-18a presents the attribution results for sulfate from the TSSA (top) and TRA (bottom) methods. Both methods identify California as the most significant geographic source of sulfate (TSSA estimates ~42% contribution; TRA estimates ~38% contribution). The remaining sulfate is attributed to nearby states and further U.S. regions (generally low contributions, as high as ~10% for a single geographic source), with the largest discrepancy between the contributions identified with Oregon (TSSA ~2%; TRA ~9%). The TRA results indicate a contribution from the Pacific Ocean of approximately 41%. This source region was not evaluated by TSSA. The Other contribution in the TSSA results is ~42%, which is significantly higher than the expected value of about 20%. Figure 4-18b presents the source apportionment results for sulfate (top) and nitrate (bottom) from the TSSA method. The results for sulfate and nitrate show a similar pattern of source strength from all geographic regions with a much larger fraction of mobile sources in the nitrate attribution. The Other contribution to nitrate is ~24%, which is close to the expected value of about 20%. Review of the state (Figures 4-19a and 4-19b) and regional SO 2 and NO x emissions maps (Figures 2-2a and 2-2b) confirms that there are significant sources of both species within California and nearby states, and within the area of meteorological influence suggested by the residence time back trajectory map for Yosemite NP (Figure 4-20). The color scaling on the residence time map indicates the fraction of the total time that back trajectory paths fell in a given grid cell. The darker blue regions indicate predominant flow patterns from northwest, west, and southwest of the park. Additional trajectory maps (sulfate difference maps and conditional probability maps), available on the COHA Web site ( can be reviewed to better understand the relationship between high/low sulfate loading and historical wind patterns. 4-57

112 Figure 4-21 presents a split image simulated to represent the 20% best (left) and 20% worst (right) visibility at Yosemite NP for 2002, based on IMPROVE monitoring data. This image does not support attribution results, but can be used to illustrate the differences in haze impacts on the best and worst days. This image was created using WinHaze Visual Air Quality Modeler (Ver ). Viewing a split image generated by WinHaze will enhance the perceptibility of the modeled visual air quality differences because attention is drawn to the contrasting conditions at the dividing line between each simulated image. To mitigate this problem, a black bar has been added to the split image, effectively removing the dividing line. 4-58

113 Figure 4-17a timeline of IMPROVE monitoring data for Yosemite NP.

114 Figure 4-17b timeline of CMAQ model results for Yosemite NP.

115 1.pdf Figure 4-18a. Sulfate TSSA and TRA source apportionment method comparison for Yosemite NP. 4-61

116 1.pdf Figure 4-18b. Sulfate and nitrate TSSA source apportionment results for Yosemite NP. 4-62

117 California SO 2 Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Sulfur oxide gases (SOx) are formed when sulfur containing fuels, such as oil or coal, are burned, when gasoline is extracted from oil or when metals are extracted from ore. In California, 2002 emissions of SO 2 were dominated by point and off-road sources. SO 2 dissolves in water vapor to form acid, and contributes to the formation of sulfate compounds (e.g. (NH 4 ) 2 SO 4 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. SO 2 Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% California 2002 SO 2 Emissions 123 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Offshore Natural (N) Rx Fires Wildland Fire 0% A N Figure 4-19a. California emissions map for SO 2. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-63

118 California NO X Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Nitrogen oxides (NO X ) form when fuel is burned at high temperatures. In California, 2002 emissions of NO X were dominated by mobile sources (on-road and off-road). NO X emissions are highly reactive and can form nitrate compounds (e.g. NH 4 NO 3 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. NO X Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% California 2002 NO x Emissions 1251 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Offshore Natural (N) Rx Fires Wildland Fire Biogenics 10% 0% A N Figure 4-19b. California emissions map for NO x. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-64

119 Figure Residence time map for Yosemite NP ( ). The location of the IMPROVE monitoring site is marked in red. Darker blue colors indicate predominant flow patterns. Figure Simulated visibility based on the 20% best days (left) and 20% worst days (right). Extinction values taken from the pie charts in Figure 4-17a. 4-65

120 Zion National Park According to IMPROVE monitoring data, the average aerosol extinction for the 20% worst visibility days at Zion NP is 31 Mm -1. The contribution from ammonium sulfate is approximately 16%, or about 5 Mm -1. The contribution from ammonium nitrate is approximately 19%, or 6 Mm -1. This can be seen in Figures 4-22a and 4-22b, which present timelines of IMPROVE monitoring data (a) and CMAQ model results (b) for A general sense of model performance at this site can be gauged by comparing the timeline plots. It is difficult to fine tune the model for an entire year, expecting good model performance during periods of both high and low extinction. The model clearly does not predict the monitoring data day-to-day, but it does it yield similar aerosol extinction averages of the 20% worst visibility days. (Note that the best and worst days for each timeline are determined by monitored and modeled data, respectively.) Comparisons between the timelines should focus on whether the species seasonal trends and episodes are similar. At Zion NP, episodes of high organic material extinction occur in both timelines in the summer and early fall months, although there is a large event predicted in late July that is not in the monitored data timeline. Episodes of high ammonium nitrate occur in both timelines during spring and late fall, with monitored values often somewhat lower than model predictions. Monitored and modeled sulfate is similar year round. (Detailed model performance is available at the RMC Web site: Figure 4-23a presents the attribution results for sulfate from the TSSA (top) and TRA (bottom) methods. The TSSA method identifies Nevada and Utah as the highest contributing source regions (~22% and ~18%, respectively) with the southwest U.S. (in this case, California) also as an important contributor (~12%). The TRA method shows the greatest sulfate contributions from the Pacific Coast and the Pacific Ocean (~20% and ~12%, respectively) and minor contributions (<10%) from all other geographic regions. The Pacific source regions were not evaluated by TSSA. The Other contribution in the TSSA results is ~31%, which is higher than the expected value of about 20%. The large difference between methods in the contribution from Nevada may be due to the edge effect described in Section 2. The TRA method does not take into account what emissions are associated with specific regions, so for a strong source of emissions near the border of a state, it is possible that air flowing over the source will pick up pollutants but the modeled back trajectories may not accrue sufficient residence time in the source state to correctly attribute the pollutant. In the case of Zion NP, there is a significant SO 2 source upwind of the site in the southern-most corner of Nevada that may not be accurately accounted for in the TRA results. Figure 4-23b presents the source apportionment results for sulfate (top) and nitrate (bottom) from the TSSA method. The results for sulfate and nitrate show somewhat different source region strengths for sulfate and nitrate, particularly from Utah, Nevada, and the southwest U.S. (California). There is a larger fraction of mobile source attribution for nitrate than for sulfate. The Other contribution to nitrate is ~17%, which is close to the expected value of about 20%. Review of the state (Figures 4-24a and 4-24b) and regional SO 2 and NO x emissions maps (Figures 2-2a and 2-2b) confirms that there are significant sources of both species within Utah, Nevada, and nearby states, and within the area of meteorological influence suggested by the residence time back trajectory map for Zion NP (Figure 4-25). The color scaling on the residence time map indicates the fraction of the total time that back trajectory paths fell in a 4-66

121 given grid cell. The darker blue regions indicate predominant flow patterns from west of the park. Additional trajectory maps (sulfate difference maps and conditional probability maps), available on the COHA Web site ( can be reviewed to better understand the relationship between high/low sulfate loading and historical wind patterns. An image simulating various aerosol conditions using WinHaze Visual Air Quality Modeler (Ver ) is not available for this site. 4-67

122 Figure 4-22a timeline of IMPROVE monitoring data for Zion NP.

123 Figure 4-22b timeline of CMAQ model results for Zion NP.

124 1.pdf Figure 4-23a. Sulfate TSSA and TRA source apportionment method comparison for Zion NP. 4-70

125 1.pdf Figure 4-23b. Sulfate and nitrate TSSA source apportionment results for Zion NP. 4-71

126 Utah SO 2 Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Sulfur oxide gases (SOx) are formed when sulfur containing fuels, such as oil or coal, are burned, when gasoline is extracted from oil or when metals are extracted from ore. In Utah, 2002 emissions of SO 2 were dominated by point sources. SO 2 dissolves in water vapor to form acid, and contributes to the formation of sulfate compounds (e.g. (NH 4 ) 2 SO 4 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. SO 2 Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Utah 2002 SO 2 Emissions 71 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire 0% A N Figure 4-24a. Utah emissions map for SO 2. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-72

127 Utah NO X Emissions WRAP Interim 2002 Inventory Emissions from all available source categories are indicated according to magnitude in 36x36km squares. Class I Areas are indicated in green, dotted lines indicate interstate highways and black dots indicate major cities. Nitrogen oxides (NO X ) form when fuel is burned at high temperatures. In Utah, 2002 emissions of NO X were dominated by mobile sources (on-road and offroad) and point sources (industrial, commercial, and residential sources that burn fuel). NO X emissions are highly reactive and can form nitrate compounds (e.g. NH 4 NO 3 ). These compounds can block the transmission of light, contributing to visibility reduction on a regional scale in our Class I Areas. NO X Emissions 100% 90% 80% 70% 60% 50% 40% 30% 20% Utah 2002 NO x Emissions 267 thousand tons/year Anthropogenic (A) Ag Fires Rx Fires Off-Road Mobile On-Road Mobile Area Point Natural (N) Rx Fires Wildland Fire Biogenics 10% 0% A N Figure 4-24b. Utah emissions map for NO x. Emissions from all available source categories are indicated according to magnitude in 36x36km grid cells. Class I areas are indicated in green, dotted lines indicate interstate highways, and black dots indicate major cities. 4-73

128 Figure Residence time map for Zion NP ( ). The location of the IMPROVE monitoring site is marked in red. Darker blue colors indicate predominant flow patterns. 4-74

129 4.1.3 Fire Episode Impacts on IMPROVE Samples In section 2.2 it was noted that the Regional Haze Rule guidance on using IMPROVE monitoring data allows for data substitution of missing species mass under certain conditions. It is possible, for example, to lose the nitrate or coarse mass data for a sampling day (due to a variety of errors) yet not lose that day in the annual data set. Review of the IMPROVE data on the VIEWS Web site shows that data substitutions were made at some WRAP sites during However, some sites experience clogged filters in 2002 and the days associated with these samples could not be remedied with the data substitution rules. Table 4-5 presents a summary of sample days with clogged filters at WRAP sites in Most of these events affected the A module filter, which is used to derive soil, coarse mass (in combination with the D module PM 10 measurement), and sulfate. The B module filter is analyzed for sulfate, thus providing a backup of this last measurement. Some of the clogged filter events affected the D module, which is principally used to derive coarse mass (in combination with the A module PM 2.5 measurement). These clogged filter events are believed to be primarily the result of heavy fire impacts at the sites because most of them occurred during the traditional fire season and the C module filters for most of the days yielded high carbon mass. Table 4-5 Summary of Clogged Filters at WRAP IMPROVE Sites by Module in 2002 Site Date A B C D Site Date A B C D Site Date A B C D KALM1 5/5/2002 X CRLA1 7/31/2002 X THRO1 8/6/2002 X KALM1 5/8/2002 X CRLA1 8/3/2002 X THRO1 8/24/2002 X KALM1 5/11/2002 X CRLA1 8/6/2002 X WEMI1 6/22/2002 X KALM1 5/14/2002 X CRLA1 8/9/2002 X WEMI1 6/25/2002 X KALM1 5/17/2002 X CRLA1 8/12/2002 X BOAP1 12/7/2002 X KALM1 8/9/2002 X CRLA1 8/15/2002 X BRCA1 4/5/2002 X KALM1 8/12/2002 X CRLA1 8/18/2002 X PASA1 10/17/2002 X KALM1 8/15/2002 X X TRIN1 7/28/2002 X KALM1 8/18/2002 X TRIN1 8/9/2002 X KALM1 8/21/2002 X TRIN1 8/15/2002 X KALM1 9/2/2002 X TRIN1 8/18/2002 X KALM1 9/5/2002 X TRIN1 8/21/2002 X OKEF1 6/16/2002 X X xxxxxxxxxxxxxxxxxxxxxx Non-WRAP sites xxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxx KALM1 9/8/2002 X REDW1 6/28/2002 X VILA1 6/19/2002 X X KALM1 9/11/2002 X REDW1 7/28/2002 X LABE1 6/7/2002 X REDW1 8/9/2002 X LABE1 6/10/2002 X YELL2 7/4/2002 X LABE1 7/25/2002 X YELL2 7/10/2002 X LABE1 7/28/2002 X YELL2 7/13/2002 X LABE1 7/31/2002 X PEFO1 6/22/2002 X LABE1 8/15/2002 X X PEFO1 6/25/2002 X LABE1 8/18/2002 X X SYCA1 7/19/2002 X LABE1 8/30/2002 X X SYCA1 8/24/2002 X 4-75

130 Removing sample days associated with clogged A or D module filters from the RHRdefined data set can significantly affect the identification and average of the 20% highest and lowest daily extinction values for a calendar year. Two examples are presented here, one for Crater Lake National Park, which lost 7 sample days to clogged filters, and one for Petrified Forest National Park, which lost only 2 sample days to clogged filters. (An example was already presented in the previous section in the discussion about Yellowstone National Park). Figures 4-26a and 4-26b are IMPROVE data timeline plots for Crater Lake. The first presents 2002 data based on the RHR guidelines for data substitutions. The second presents raw 2002 data, including available data from clogged filters, but without RHR data substitutions. Note that the scales on the two timelines are not the same. The clogged filters occurred predominantly in August, and all were associated with high carbon (organic and elemental) and somewhat elevated nitrate. The average of the 20% highest aerosol extinction samples increased from 68 Mm -1, based on the RHR data set, to 149 Mm -1, based on the raw data set. The fraction of aerosol extinction due to organic material increased from 64% to 72%. The 20% lowest aerosol extinction days were not significantly affected. Figures 4-27a and 4-27b are IMPROVE data timeline plots for Petrified Forest. The first presents 2002 data based on the RHR guidelines for data substitutions. The second presents raw 2002 data, including available data from clogged filters, but without RHR data substitutions. Again, note that the scales on the two timelines are not the same. The clogged filters occurred on two days in June, both associated with high carbon (organic and elemental). The average of the 20% highest aerosol extinction samples increased from 40 Mm -1, based on the RHR data set, to 67 Mm -1, based on the raw data set. The fraction of aerosol extinction due to organic material increased from 38% to 54%. The 20% lowest aerosol extinction days were not significantly affected. Similar timeline plots are available on the AoH project Web site for all of the sites listed in Table 4-5. See Section 3 for instructions on how to access these reports from the Web site. 4-76

131 Figure 4-26a timeline of IMPROVE monitoring data according to the RHR guidance for Crater Lake NP.

132 Figure 4-26b timeline of raw IMPROVE monitoring data, including data from clogged filter events, for Crater Lake NP.

133 Figure 4-27a timeline of IMPROVE monitoring data according to the RHR guidance for Petrified Forest NP.

134 Figure 4-27b timeline of raw IMPROVE monitoring data, including data from clogged filter events, for Petrified Forest NP.

135 4.2 REGIONAL ASSESSMENTS Attribution of visibility impacts due to fire activity, carbonaceous materials, and dust emissions to geographic source regions were not investigated with the TSSA or TRA attribution methods. However, regional impacts for fire emissions impacts, the state of knowledge for carbonaceous aerosols, and the magnitude of dust emissions source categories were investigated and assessed. This section describes these assessments and their results Fire Assessment The Regional Modeling Center completed 3 modeling sensitivity runs focused on fire emissions impacts for the WRAP region. These runs used the Fire Emissions Joint Forum (FEJF) Phase 1, 2002 emissions inventory developed for prescribed and wildfire emissions in the entire WRAP region, and the base case agricultural fire inventory developed for the Section 309 planning process. Natural and anthropogenic fire definitions and rules for applying those definitions to fire activity data were developed by the FEJF. In general, wildfires, wildland fires used for resource management, and prescribed burning used for specific aspects of resource management are considered natural, while agricultural burning and some prescribed burning are considered anthropogenic. (For more information regarding the definitions of specific fire activities see: and The sensitivity runs consisted of removing one or more categories of fire emissions from the modeling inputs. Each of the sensitivity run was then compared to the complete modeling run that included all emissions, and the differences were interpreted as the contribution to overall visibility by the missing emissions category. The three sensitivity runs included: Removal of all fire emissions The difference between this run and the complete run characterizes the net impact of all fire emissions. Removal of natural fire emissions The difference between this run and the complete run characterizes the impact of natural fire emissions. Removal of anthropogenic fire emissions The difference between this run and the complete run characterizes the impact of anthropogenic fire emissions. During 2002 there were a number of large wildfires in the WRAP region, including the Biscuit fire in southwestern Oregon, the Rodeo-Chediski fire in central-eastern Arizona, and the Hayman fire in central Colorado. Fire activity and associated emissions are very spatially and temporally episodic by their nature, but across the WRAP region, annual state totals of natural and anthropogenic fire activity data in acres and tons of PM 2.5 emissions show that natural fire emissions were dominant during PM 2.5 fire emissions include organic and elemental carbon, fine soil and metals, and small amounts of sulfate. Alaska emissions were not modeled. Table 4-6 presents the state by state results (where available) of the WRAP 2002 Phase I fire emissions inventory used in this assessment, broken down by natural and anthropogenic emissions. Natural fires accounted for 97% of the fire-related PM 2.5 emissions in

136 Table Phase I Fire Emissions Inventory Data Fire Categorization & Fire Type State Activity (Acres) PM 2.5 Emissions (tons) Prescribed Burning Natural Emissions Alaska Arizona 53,609 13,453 California 56,430 3,897 Colorado 5, Idaho 63,270 3,257 Montana 27,361 2,201 Nevada 13, New Mexico 20,327 1,651 North Dakota 27, Oregon 16,625 2,192 South Dakota 8, Utah 11,178 1,603 Washington 6, Wyoming 16,055 1,363 Prescribed Fire Total 326,414 32,363 Wildland Fire Use Alaska 159,554 72,907 Arizona N/A N/A California 4,051 2,484 Colorado 22,844 8,787 Idaho 3,543 2,027 Montana Nevada N/A N/A New Mexico 4, North Dakota N/A N/A Oregon N/A N/A South Dakota N/A N/A Utah Washington N/A N/A Wyoming 6,220 1,221 Wildland Fire Use Total 201,341 87,781 Wildfire Alaska 1,790, ,701 Arizona 555, ,920 California 510, ,759 Colorado 412, ,514 Idaho 77,645 7,960 Montana 112,413 10,453 Nevada 56,218 5,541 New Mexico 247,243 19,872 North Dakota 83,599 3,665 Oregon 933, ,411 South Dakota 78,384 2,393 Utah 212,088 30,533 Washington 89,854 11,854 Wyoming 154,857 28,521 Wildfire Total 5,314,774 1,489,097 Natural Fire Total 5,842,528 1,609,

137 Table 4-6 (Cont.) 2002 Phase I Fire Emissions Inventory Data Fire Categorization & Fire Type State Activity (Acres) PM 2.5 Emissions (tons) Agricultural Burning Anthropogenic Emissions Alaska N/A N/A Arizona 16, California 881,951 7,000 Colorado Idaho 405,924 5,603 Montana 2, Nevada N/A N/A New Mexico 3, North Dakota 215,883 2,231 Oregon 205,115 2,584 South Dakota 52, Utah 16, Washington 176,249 2,352 Wyoming 14, Agricultural Burning Total 1,992,043 21,020 Prescribed Burning Alaska Arizona California 30,954 6,207 Colorado 11, Idaho 26,204 8,241 Montana 50,232 4,717 Nevada 1, New Mexico 5, North Dakota N/A N/A Oregon 105,623 9,717 South Dakota N/A N/A Utah 3, Washington 27,488 1,492 Wyoming 5, Prescribed Fire Total 267,974 32,495 Anthropogenic Fire Total 2,260,017 53,516 Grand Total All Fire 8,102,545 1,662,

138 Figure 4-28 presents the modeled annual average contribution to light extinction by all fire emissions categories for The extinction values indicated for each grid cell on this and subsequent maps represent the modeled extinction due to fire activity only, as the extinction due to other species has been subtracted out. Visibility impacts due to fires are shown to have been generally less than 10 Mm -1 across WRAP, although some locations were impacted by as much as 25, 50, or >100 Mm -1. Geographically, the largest impacts due to fire occur in southern Oregon, much of California, and isolated locations in Utah, Arizona, and Colorado. Figure 4-29 presents the modeled annual average contribution to light extinction by all natural fires for This map is not significantly different from the previous map, indicating that natural fires contribute a large percentage of the impact of both fire categories combined. Figure 4-30 presents the modeled annual average contribution to light extinction by all anthropogenic fires for This map indicates that the most significant contributions by anthropogenic fires during 2002 occurred in the region around the pan handle of Idaho and California s Central Valley. The maximum modeled impact of anthropogenic fires is less than 5 Mm -1. Figure Modeled annual average contribution to light extinction by all fire categories for The extinction values indicated in each grid cell represent the modeled extinction due to all fire activity only; extinction due to other species is not represented. 4-84

139 Figure Modeled annual average contribution to light extinction by natural fires for The extinction values indicated in each grid cell represent the modeled extinction due to natural fire activity only; extinction due to other species is not represented. Figure Modeled annual average contribution to light extinction by anthropogenic fires for The extinction values indicated in each grid cell represent the modeled extinction due to anthropogenic fire activity only; extinction due to other species is not represented. 4-85

140 4.2.2 Carbon Assessment Particulate carbon occurs naturally and through human activities in two forms: elemental carbon (EC) and organic carbon (OC). Carbon particulate species are produced through both internal combustion (cars) and external combustion (fire), and there are other sources as well. Elemental carbon is produced primarily through combustion of diesel and gasoline fuels, wood burning, and vegetative burning. Organic carbon consists of primary and secondary components. Primary OC is emitted directly into the atmosphere. Its major sources include biogenic emissions, vehicle emissions, meat cooking, wood burning, and vegetative burning. Secondary OC is formed by the condensation of low-volatility organic gases, often as a result of complex photochemical reactions. Carbon aerosols are monitored routinely across the U.S. in the IMPROVE Network and the EPA Speciation Trends Network (STN). Analysis methods and results differ between these networks. This section will present an assessment of carbon aerosols in the U.S. based on IMPROVE data. Figure 4-31 presents the fractional contribution at Class I areas by organic carbon to total aerosol extinction for the 20% worst visibility days in Organic carbon can be seen to account for a higher fraction of aerosol extinction in the western U.S. than in the eastern U.S. For much of the WRAP region, the organic carbon contribution to extinction is in the range 25-50%. Figure 4-32 presents the same information for elemental carbon. Elemental carbon is shown to contribute less than organic carbon to aerosol extinction. For much of the WRAP region, the elemental carbon contribution to extinction is in the range 7-15%. Figure 4-33 presents the ratio of organic carbon mass to elemental carbon mass in the IMPROVE Network for Summer The ratio of organic to elemental carbon for vegetative burning tends to be much higher than for the other combined sources of carbon. Typical OC/EC ratios due to urban sources are in the range of 3 5. OC/EC ratios for fire-impacted samples can reach 10 and above due to the predominance of organic carbon in fire smoke. The site to site differences in the OC/EC ratios shown in the map are believed to be due to regional differences in mixtures of source types. While these maps demonstrate important contributions to WRAP region Class I area light extinction by carbonaceous aerosols, much work remains to be done to characterize the emissions, transport, atmospheric chemical processes, and deposition of these species. The WRAP has co-sponsored research workshops to identify the state of knowledge about carbonaceous species; these workshops and their findings are discussed next. 4-86

141 Figure IMPROVE organic carbon extinction as a fraction of total aerosol extinction based on the average of the 20% highest values for Some monitoring sites in the WRAP region show approximately 50% contribution to aerosol extinction from organic carbon. Figure IMPROVE elemental carbon extinction as a fraction of total aerosol extinction based on the average of the 20% highest values for Some monitoring sites in the WRAP region show up to 15% contribution to aerosol extinction from elemental carbon. 4-87