SAMI Air Quality Modeling Final Report

SAMI Air Quality Modeling Final Report

M. Talat Odman, James W. Boylan, James G. Wilkinson and Armistead G. Russell School of Civil and Environmental Engineering Georgia Institute of Technology 200 Bobby Dodd Way Atlanta GA 30332-0512

Stephen F. Mueller and Robert E. Imhoff Tennessee Valley Authority, PO Box 1010, Muscle Shoals AL 35662

Kevin G. Doty, William B. Norris and Richard T. McNider University of Alabama in Huntsville, Huntsville, AL 35899

July 2002

ODMAN,

ET AL.

SAMI Air Quality Modeling Report

TABLE OF CONTENTS 1

EXECUTIVE SUMMARY............................................................................................................................... 1-1 1.1 INTRODUCTION..............................................................................................................................................1-1 1.2 A IR QUALITY MODEL ...................................................................................................................................1-1 1.3 M ODEL SETUP AND INPUTS ..........................................................................................................................1-2 1.4 M ODEL PERFORMANCE ................................................................................................................................1-3 1.5 M ODEL INPUTS FOR FUTURE YEAR SIMULATIONS......................................................................................1-4 1.6 RESPONSE TO FUTURE YEAR EMISSION STRATEGIES .................................................................................1-5 1.7 SENSITIVITY ANALYSIS.................................................................................................................................1-6 1.7.1 PM Sensitivities................................................................................................................................... 1-7 1.7.2 Wet Deposition Sensitivities............................................................................................................... 1-7 1.7.3 Ozone Sensitivities.............................................................................................................................. 1-8 1.8 CONCLUSION .................................................................................................................................................1-8

2

INTRODUCTION.............................................................................................................................................. 2-1 2.1 2.2 2.3 2.4

3

BACKGROUND ...............................................................................................................................................2-1 SAMI OVERVIEW .........................................................................................................................................2-2 A IR QUALITY MODELING OBJECTIVES........................................................................................................2-4 REPORT STRUCTURE .....................................................................................................................................2-4

AIR QUALITY MODEL.................................................................................................................................. 3-1 3.1 INTRODUCTION..............................................................................................................................................3-1 3.2 URBAN TO REGIONAL M ULTISCALE - ONE ATMOSPHERE (URM-1ATM) MODEL .................................3-1 3.2.1 Overview.............................................................................................................................................. 3-1 3.2.2 Transport ............................................................................................................................................. 3-2 3.2.3 Chemistry ............................................................................................................................................ 3-3 3.2.4 Aerosols............................................................................................................................................... 3-4 3.2.5 Wet Deposition.................................................................................................................................... 3-5 3.2.6 Dry Deposition.................................................................................................................................... 3-7

4

MODEL SETUP AND INPUT DATA PREPARATION........................................................................... 4-1 4.1 M ODELING DOMAIN AND GRID....................................................................................................................4-1 4.2 M ODELED EPISODES .....................................................................................................................................4-3 4.3 M ETEOROLOGICAL INPUTS...........................................................................................................................4-4 4.4 EMISSION INPUTS..........................................................................................................................................4-7 4.4.1 Emission Summary Tables................................................................................................................. 4-8 4.4.2 Day Specific Emissions...................................................................................................................... 4-8 4.5 INITIAL AND BOUNDARY CONDITIONS.......................................................................................................4-12 4.6 M ODEL OUTPUTS........................................................................................................................................4-15

5

OZONE MODEL PERFORMANCE............................................................................................................. 5-1 5.1 AIRS OBSERVATION DATABASE .................................................................................................................5-1 5.2 DAILY M AXIMUM OZONE SPATIAL PLOTS..................................................................................................5-1 5.3 DIURNAL STATION PLOTS............................................................................................................................5-2 5.3.1 Comparison to AIRS Data for Selected Episodes and Sites ............................................................ 5-2 5.3.2 Comparison to SOS Surface Data for the July 11-19, 1995 Episode ............................................. 5-4 5.4 BIAS AND ERROR CALCULATIONS................................................................................................................5-4

6

AEROSOL MODEL PERFORMANCE....................................................................................................... 6-1 6.1 6.2 6.3 6.4

IMPROVE OBSERVATION DATABASE .......................................................................................................6-1 SPATIAL PLOTS OF DAILY A VERAGE CONCENTRATIONS ...........................................................................6-1 DAILY A VERAGE CONCENTRATIONS AT IMPROVE STATIONS................................................................6-1 SCATTER PLOTS.............................................................................................................................................6-4 I

ODMAN,

6.5 6.6 7

ET AL.


BIAS AND ERROR CALCULATIONS................................................................................................................6-8 DISCUSSION OF PERFORMANCE RESULTS....................................................................................................6-9

WET DEPOSITION MODEL PERFORMANCE...................................................................................... 7-1 7.1 7.2 7.3 7.4 7.5 7.6 7.7

8

NADP OBSERVATION DATABASE ...............................................................................................................7-1 SPATIAL PLOTS OF 7-DAY CUMULATIVE DEPOSITION ...............................................................................7-1 7-DAY CUMULATIVE DEPOSITION AT NADP STATIONS............................................................................7-1 SCATTER PLOTS OF SEVEN -DAY CUMULATIVE W ET DEPOSITION.............................................................7-3 BIAS AND ERROR CALCULATIONS................................................................................................................7-8 DISCUSSION OF PERFORMANCE RESULTS....................................................................................................7-8 COMPARISON TO OTHER STUDIES..............................................................................................................7-10

SEASONAL AND ANNUAL AIR QUALITY.............................................................................................. 8-1 8.1 8.2 8.3

9

CALCULATION OF SEASONAL AND A NNUAL METRICS...............................................................................8-1 M ODEL PERFORMANCE IN TERMS OF SEASONAL AND ANNUAL METRICS................................................8-5 SOURCES OF M ODELING UNCERTAINTY......................................................................................................8-6

MODEL INPUTS FOR FUTURE YEAR SIMULATIONS...................................................................... 9-1 9.1 M ETEOROLOGICAL INPUTS...........................................................................................................................9-1 9.2 EMISSION INPUTS..........................................................................................................................................9-1 9.2.1 Overview of Future Year Emission Strategies.................................................................................. 9-1 9.2.2 Emission Summary Tables................................................................................................................. 9-6 9.2.3 Spatial Plots for Selected Species and Days..................................................................................... 9-7 9.3 INITIAL AND BOUNDARY CONDITIONS.........................................................................................................9-9 9.3.1 Initial Conditions ................................................................................................................................ 9-9 9.3.2 Boundary Conditions........................................................................................................................ 9-13

10

RESPONSES TO FUTURE YEAR EMISSION STRATEGIES ....................................................... 10-1

10.1 OZONE RESPONSE TO FUTURE YEAR EMISSION STRATEGIES ..................................................................10-1 10.1.1 Spatial Plots of Daily Maximum Ozone .......................................................................................... 10-1 10.1.2 Diurnal Plots of Hourly Ozone at Selected Stations ...................................................................... 10-2 10.1.3 Seasonal Cumulative Ozone W126 Response ................................................................................ 10-3 10.2 AEROSOL RESPONSE TO FUTURE YEAR EMISSION STRATEGIES ..............................................................10-4 10.2.1 Spatial Plots of Daily Average PM Concentrations....................................................................... 10-4 10.2.2 Charts of Daily Average PM Concentrations for Selected Stations............................................. 10-5 10.2.3 Annual Average PM Response ........................................................................................................ 10-6 10.3 A CID DEPOSITION RESPONSE TO FUTURE YEAR EMISSION STRATEGIES ................................................10-8 10.3.1 Spatial Plots of Weekly Cumulative Deposition Fluxes ................................................................. 10-8 10.3.2 Charts of Weekly Cumulative Deposition for Selected Stations.................................................. 10-11 10.3.3 Annual Acid Deposition Response ................................................................................................ 10-12 11 11.1 11.2 11.3 11.4 12

MODELING OF SENSITIVITIES .......................................................................................................... 11-1 INTRODUCTION............................................................................................................................................11-1 DIRECT SENSITIVITY A NALYSIS.................................................................................................................11-1 COMPARISON OF DDM-3D AND BRUTE FORCE METHODS......................................................................11-5 IMPORTANT REMARKS ABOUT DDM-3D ..................................................................................................11-7 SENSITIVITY ANALYSIS ....................................................................................................................... 12-1

12.1 DEFINITION OF REGIONS FOR SENSITIVITY A NALYSIS..............................................................................12-1 12.2 A NNUAL AND SEASONAL SENSITIVITIES ...................................................................................................12-3 12.3 AEROSOL SENSITIVITIES.............................................................................................................................12-3 12.3.1 PM Sensitivities to SO2 Emissions................................................................................................... 12-3 12.3.2 PM Sensitivities to Elevated NOx Emissions................................................................................... 12-9 12.3.3 PM Sensitivities to Ground-Level NOx Emissions........................................................................ 12-11 12.3.4 PM Sensitivities to NH3 Emissions................................................................................................ 12-12 12.4 WET DEPOSITION SENSITIVITIES..............................................................................................................12-15 II

ODMAN,

ET AL.


12.4.1 Wet Deposition Sensitivities to SO2 Emissions............................................................................. 12-15 12.4.2 Wet Deposition Sensitivities to Elevated NOx Emissions............................................................. 12-17 12.4.3 Wet Deposition Sensitivities to Ground-Level NOx Emissions.................................................... 12-19 12.4.4 Wet Deposition Sensitivities to NH3 Emissions............................................................................ 12-21 12.5 OZONE SENSITIVITIES ...............................................................................................................................12-22 12.5.1 Ozone Sensitivities to Elevated NOx Emissions............................................................................ 12-23 12.5.2 Ozone Sensitivities to Ground-Level NOx Emissions................................................................... 12-25 12.6 SUMMARY OF SENSITIVITY ANALYSIS.....................................................................................................12-26 13

CONCLUSION............................................................................................................................................. 13-1

14

REFERENCES ............................................................................................................................................. 14-1

15

APPENDIX A: COMPACT DISK (CD).................................................................................................. 15-1

16

APPENDIX B: CD FILE STRUCTURE AND NAMING CONVENTIONS .................................. 16-1

16.1 16.2 16.3 16.4 17 17.1 17.2 17.3 17.4 18

DIRECTORY STRUCTURE OF "EMISSIONS" .................................................................................................16-1 DIRECTORY STRUCTURE OF "M ODEL RESULTS" ......................................................................................16-3 DIRECTORY STRUCTURE OF "FUTURE YEARS" .........................................................................................16-6 DIRECTORY STRUCTURE OF "REGIONAL SENSITIVITIES" .........................................................................16-8 APPENDIX C: DRY DEPOSITION MODEL PERFORMANCE.................................................... 17-1 AIRM ON OBSERVATION DATABASE ........................................................................................................17-1 SPATIAL PLOTS OF 7-DAY CUMULATIVE DRY DEPOSITION .....................................................................17-2 7-DAY CUMULATIVE DRY DEPOSITION AT AIRMON STATIONS............................................................17-3 DISCUSSION OF PERFORMANCE RESULTS..................................................................................................17-3 APPENDIX D: COMPARISON OF DDM-3D AND BRUTE-FORCE METHODS..................... 18-1

III

ODMAN,

ET AL.


List of Tables Table Table Table Table Table Table

3-1 3-2 3-3 3-4 3-5 3-6

Table 3-7 Table 4-1 Table 4-2 Table 4-3 Table 4-4 Table 4-5 Table 4-6 Table 4-7 Table 4-8 Table 4-9 Table 4-10 Table 4-11 Table 4-12 Table 4-13 Table 5-1 Table 6-1 Table 6-2 Table 6-3 Table 6-4 Table 6-5 Table 7-1 Table 7-2 Table 7-3 Table 7-4 Table 7-5 Table 8-1 Table 8-2

Transported Gas-Phase Species ........................................................................ 3-3 Transported PM Species .................................................................................... 3-3 Steady State Species ........................................................................................ 3-4 Constant Species .............................................................................................. 3-4 Possible gas and PM species in ISORROPIA ....................................................... 3-5 Equilibrium relations and equilibrium constant expressions, where γ is the activity coefficient (Nenes et al., 1998)................................................................. 3-6 Aerosol Particle Deposition Velocities .................................................................. 3-7 Vertical structure of the URM-1ATM model. .......................................................... 4-2 Modeled episodes and pollution levels observed at Great Smoky Mountains and Shenandoah National Parks. ............................................................................. 4-3 RAMS Meteorological parameters used by URM-1ATM.......................................... 4-5 Summary of National Weather Service surface observations, RAMS standard deviations, and performance statistics for wind speeds and wind direction................ 4-6 Summary of National Weather Service surface observations, RAMS standard deviations, and performance statistics for temperatures and mixing ratios. ............... 4-6 Four-cell contingency table used to calculated the equitable threat score (Schaefer, 1990). ............................................................................................. 4-7 Emission species generated by EMS-95 .............................................................. 4-8 Daily Average Emissions for July 1995 Episode .................................................... 4-8 Daily average total emissions from the 8 SAMI states for the July-95 episode (in tons per day). .................................................................................................. 4-9 Gas phase species in SAPRC mechanism and the methodology used for setting their surface layer initial and boundary conditions. .............................................. 4-13 Particulate Species Modeled by URM-1ATM and IMPROVE Observations ............. 4-15 The estimated size distribution of particulate species used in BC/ICs...................... 4-15 The IC/BC concentrations (in µg/m3) for particulate species in July 1995 episode. ....................................................................................................... 4-16 List of sites for which diurnal plots were prepared. .................................................. 5-2 IMPROVE monitoring stations and URM-1ATM grid resolution at station location........................................................................................................... 6-1 Daily mean concentration, bias, normalized bias, error and normalized error of PM2.5 by IMPROVE day. ................................................................................... 6-9 Performance statistics for speciated PM2.5 (all episodes). ...................................... 6-11 Average concentration and classification bins used for each particulate species....... 6-12 Mean and normalized bias as a function of classification for each particulate species. ........................................................................................................ 6-13 NADP stations falling into the 12-km grid cells....................................................... 7-1 Weekly mean deposition flux and model bias, normalized bias, error and normalized error of sulfate by episode. ................................................................ 7-9 Performance statistics for the wet deposition species (all episodes). ......................... 7-9 Average observed (NADP) mass flux and classification bin for each wet deposition species at the 14 monitoring stations in the 12-km grid. ........................ 7-10 Mean and normalized bias as a function of classification bin for each wet deposition species at the 14 monitoring stations in the 12-km grid. ........................ 7-11 Percent of days falling into each class based on severity of pollutant levels (Deuel and Douglas, 1998)................................................................................ 8-1 Ozone and PM classes and their contribution (weight) to the seasonal cumulative ozone (W126) and annual average visibility metrics at Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks. ............................... 8-3

IV

ODMAN,

ET AL.

Table 8-3

Table 9-1 Table 9-2 Table 9-3 Table 9-4 Table 9-5 Table 9-6 Table 9-7 Table 10-1 Table 10-2 Table 10-3 Table 10-4 Table 11-1 Table 11-2 Table 12-1 Table 12-2 Table 12-3 Table 12-4 Table 12-5

Table 12-6 Table 16-1 Table 16-2 Table 17-1


Wet and dry deposition classes and their contributions (weight) to the annual metrics at Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks. ............................................................................................................ 8-4 Daily average total 2010-OTW emissions from the 8 SAMI states for the July-95 episode (in tons per day). .................................................................................. 9-6 Daily average total 2010-BWC emissions from the 8 SAMI states for the July-95 episode (in tons per day). .................................................................................. 9-6 Daily average total 2010-BB emissions from the 8 SAMI states for the July-95 episode (in tons per day). .................................................................................. 9-6 Daily average total 2040-OTW emissions from the 8 SAMI states for the July-95 episode (in tons per day). .................................................................................. 9-7 Daily average total 2040-BWC emissions from the 8 SAMI states for the July-95 episode (in tons per day). .................................................................................. 9-7 Daily average total 2040-BB emissions from the 8 SAMI states for the July-95 episode (in tons per day). .................................................................................. 9-7 Percent reductions in initial conditions of NOx , SO2, sulfate and ammonium............. 9-10 List of stations for which diurnal plots of future year ozone concentrations were prepared. ...................................................................................................... 10-3 List of stations for which charts of daily PM concentrations were prepared. .............. 10-6 Equations used to evaluate the sulfur and nitrogen wet and dry deposition. ............. 10-8 List of stations for which charts of weekly cumulative acid deposition fluxes were prepared. .....................................................................................................10-11 a Mass conservation and charge balance expressions. .......................................... 11-3 Comparison of DDM-3D sensitivity results to brute force results for a 30% reduction in emissions .................................................................................... 11-6 List of states in each source region used for sensitivity analysis. ............................ 12-2 Emission types and species list for which the sensitivities to the emissions were analyzed. ...................................................................................................... 12-2 Labels, and locations for receptors used in aerosol and wet deposition sensitivity analysis. ....................................................................................................... 12-5 Labels, and locations for receptors used in ozone sensitivity analysis. ...................12-24 Average sensitivity (absolute) of ozone W126, PM, and wet deposition at Class I areas to a 10% reduction in SO2, elevated NOx, ground level NOx, and NH3 emissions. ...................................................................................................12-26 Average sensitivity (%) of ozone W126, PM, and wet deposition at Class I areas to a 10% reduction in SO2, elevated NOx, ground level NOx, and NH3 emissions. ....12-27 IMPROVE monitoring station identifiers used in the file names. ............................. 16-4 AIRS monitoring station identifiers used in the file names. ..................................... 16-5 Performance statistics for the dry deposition species (all episodes). ....................... 17-3

V

ODMAN,

ET AL.


List of Figures Figure 1-1

Figure 2-1 Figure 4-1 Figure 4-2 Figure 4-3

Figure 4-4 Figure 4-5 Figure 4-6 Figure 4-7 Figure 4-8 Figure 4-9 Figure 5-1 Figure 5-2 Figure 5-3 Figure 5-4 Figure 5-5 Figure 5-6 Figure Figure Figure Figure

5-7 5-8 5-9 5-10

Figure 5-11 Figure 5-12 Figure 6-1 Figure 6-2 Figure 6-3 Figure 6-4 Figure 6-5 Figure 6-6 Figure 6-7

Total annual emissions from the eight SAMI states in 1990 and projections for 2010 and 2040 under the OTW, BWC, and BB strategies. VOC emissions reported here do not include biogenic emissions................................................... 1-5 SAMI integrated assessment framework focusing on the atmospheric modeling aspects........................................................................................................... 2-3 The SAMI modeling domain and grid. .................................................................. 4-1 Surface elevation (above sea level) mapped onto the grid....................................... 4-2 Horizontal grid structures used for the SAMI meteorological modeling. (a) Grids used for the July 1995 episode. (b) Grids used for the other eight episodes. Only every other grid point is plotted. .................................................................. 4-5 Gridded daily total emissions of elevated SO2 over the SAMI region for July 15, 1995............................................................................................................... 4-9 Gridded daily total emissions of ground level SO2 over the SAMI region for July 15, 1995. ...................................................................................................... 4-10 Gridded daily total emissions of elevated NOx over the SAMI region for July 15, 1995............................................................................................................. 4-10 Gridded daily total emissions of ground-level NOx over the SAMI region for July 15, 1995. ...................................................................................................... 4-11 Gridded daily total emissions of ground level NH3 over the SAMI region for July 15, 1995. ...................................................................................................... 4-11 Initial surface layer SO2 concentrations used for the July 9-19, 1995 episode........... 4-14 Daily maximum ozone on July 12, 1995. .............................................................. 5-1 Observed (* ) and simulated (-) ozone levels at Knoxville, TN for May 11-18, 1993............................................................................................................... 5-3 Observed (* ) and simulated (-) ozone levels at Great Smoky Mountains, TN for August 3-11, 1993............................................................................................ 5-3 Observed (* ) and simulated (-) ozone levels at Shenandoah, VA August 3-11, 1993............................................................................................................... 5-4 Nitrogen oxide (NO) concentrations at Giles County, TN during July 11-19, 1995............................................................................................................... 5-5 Nitrogen dioxide (NO2) concentrations at Giles County, TN during July 11-19, 1995............................................................................................................... 5-5 Ozone (O3) concentrations at Giles County, TN during July 11-19, 1995 ................... 5-6 Hourly mean normalized bias (12-km grid only) for the July 9-19, 1995 episode. ........ 5-7 Hourly mean normalized error (12-km grid only) for the July 9-19, 1995 episode. ....... 5-7 Ozone daily mean normalized bias for the seventy four AIRS sites in the 12-km grid during the seven ozone episodes. ................................................................ 5-8 Ozone daily mean normalized error for the seventy four AIRS sites in the 12-km grid during the seven ozone episodes. ................................................................ 5-8 Model estimated error versus observed daily maximum 8-hour average ozone. ........ 5-9 Daily average PM2.5 concentrations over the SAMI region on July 15, 1995. .............. 6-2 Simulated and observed PM2.5 concentrations at Great Smoky Mountains: July1995 episode. .................................................................................................. 6-2 Simulated and observed PM2.5 concentrations at all IMPROVE stations on July 15, 1995. ........................................................................................................ 6-3 Simulated (left) and Observed (right) speciated PM2.5 on July 15, 1995. .................... 6-3 IMPROVE measurements vs. simulated concentrations for sulfate. Also shown are the 1:1 and ±50% bias lines. ........................................................................ 6-5 IMPROVE measurements vs. simulated concentrations for nitrate. Also shown are the 1:1 and ±50% bias lines. ........................................................................ 6-5 IMPROVE measurements vs. simulated concentrations for ammonium. Also shown are the 1:1 and ±50% bias lines. .............................................................. 6-6

VI

ODMAN,

ET AL.


Figure 6-8

IMPROVE measurements vs. simulated concentrations for organics. Also shown are the 1:1 and ±50% bias lines. .............................................................. 6-6 IMPROVE measurements vs. simulated concentrations for elemental carbon. Also shown are the 1:1 and ±50% bias lines. ....................................................... 6-7 IMPROVE measurements vs. simulated concentrations for soils. Also shown are the 1:1 and ±50% bias lines. ........................................................................ 6-7 IMPROVE measurements vs. simulated concentrations for PM2.5. Also shown are the 1:1 and ±50% bias lines. ........................................................................ 6-8 Simulated and Observed (gravimetric) PM2.5 at Great Smoky Mountains National Park................................................................................................. 6-10 Simulated and Observed (gravimetric) PM2.5 at Shenandoah National Park............ 6-10 Weekly cumulative sulfate wet deposition for July 11-18, 1995. ............................... 7-2 Simulated and observed (NADP) sulfate wet deposition fluxes at 14 monitoring sites in the 12-km grid for the week of March 23-30, 1993. ..................................... 7-2 Sulfate deposition mass flux at Great Smoky Mountains National Park ..................... 7-3 Simulated and observed wet-deposited sulfate concentrations for the week of March 23-30, 1993. .......................................................................................... 7-3 Nitrate deposition mass flux at Great Smoky Mountains National Park ...................... 7-4 Precipitation at Great Smoky Mountains National Park ........................................... 7-4 National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated sulfate flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines. .............................................. 7-5 National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated nitrate flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines. ....................................................... 7-5 National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated ammonium flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines. .............................................. 7-6 National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated calcium flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines. .............................................. 7-6 National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated magnesium flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines. .............................................. 7-7 National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM model input precipitation using the "best" cell value. Also shown are the 1:1 and ±50% bias lines. .............................................. 7-7 Comparison of modeled and observed seasonal ozone W126 at Great Smoky Mountains (Look Rock) and Shenandoah (Big Meadows) National Parks................. 8-5 Comparison of modeled and observed annual averaged fine PM concentrations at Great Smoky Mountains (GRSM—Look Rock) and Shenandoah (SHEN— Big Meadows) National Parks. ........................................................................... 8-6 Comparison of modeled and observed annual averaged wet deposition mass fluxes at Great Smoky Mountains (GRSM—Elkmont) and Shenandoah (SHEN—Big Meadows) National Parks. .............................................................. 8-7 Emissions from the eight SAMI states in 1990 and projections for 2010 and 2040 under the OTW, BWC, and BB strategies. VOC emissions reported here are anthropogenic and do not include biogenic emissions...................................... 9-2 SO2 emissions in 1990 and projected for 2010 and 2040 under the OTW, BWC, and BB strategies............................................................................................. 9-3 NOx emissions in 1990 and projected for 2010 and 2040 under the OTW, BWC, and BB strategies............................................................................................. 9-3 Summer day NOx emissions in 1990 and projected for 2010 and 2040 under the OTW, BWC, and BB strategies. ......................................................................... 9-4 NOx emissions in 1990 and projected for 2010 and 2040 under the OTW (A2), BWC (B1), and BB (B3) strategies. ..................................................................... 9-5

Figure 6-9 Figure 6-10 Figure 6-11 Figure 6-12 Figure 6-13 Figure 7-1 Figure 7-2 Figure 7-3 Figure 7-4 Figure 7-5 Figure 7-6 Figure 7-7

Figure 7-8

Figure 7-9

Figure 7-10

Figure 7-11

Figure 7-12

Figure 8-1 Figure 8-2

Figure 8-3

Figure 9-1


VII

ODMAN,

ET AL.


Figure 9-6

Nitrogen emissions in 1990 and projected for 2010 and 2040 under the OTW (A2), BWC (B1), and BB (B3) strategies .............................................................. 9-5 Gridded daily total emissions of ground-level NOx over the SAMI region for July 15, 2010 under the OTW strategy....................................................................... 9-8 Gridded change in daily total emissions of ground-level NOx for July 15, from 1995 to 2010 under the OTW strategy. ............................................................... 9-8 Change in daily maximum ozone for July 11 from 1995 to 2010 (left) and further change due to reducing NOx initial conditions by 35% (right). ................................ 9-11 2Fractional change in daily average SO4 for July 11 from 1995 to 2010 (left) and further change due to reducing SO2 initial conditions by 30% (right). ................ 9-11 Change in daily average NO3 for July 11 from 1995 to 2010 (left) and further change due to increasing NH3 initial conditions from zero to 1 ppb (right). .............. 9-12 Fractional change in cumulative NO3 wet deposition for the week of July 11-18 from 1995 to 2010 (left) and further change due to decreasing NOx initial conditions by 35% (right). ................................................................................ 9-13 Sensitivity of O3 at Look Rock, TN to boundary conditions during the July 1995 episode: baseyear (black), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (green).......................................................................................................... 9-14 Sensitivity of O3 at Big Meadows, VA to boundary conditions during the July 1995 episode: baseyear (black), 2040-BWC emissions and 1995 to boundary conditions (red) and 2040-BWC emissions and scaled to boundary conditions (green).......................................................................................................... 9-14 2Sensitivity of fine SO4 at Great Smoky Mountains, TN to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow). ......................................................................................... 9-16 2Sensitivity of fine SO4 at Shenandoah National Park, VA to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow). ................................................................. 9-16 Sensitivity of fine NO3 at Great Smoky Mountains, TN to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow). ......................................................................................... 9-17 Sensitivity of fine NO3 at Shenandoah National Park, VA to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow). ................................................................. 9-17 + Sensitivity of fine NH4 at Great Smoky Mountains, TN to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow). ......................................................................................... 9-18 + Sensitivity of fine NH4 at Shenandoah National Park, VA to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow). ................................................................. 9-18 2Sensitivity of SO4 wet deposition at Great Smoky Mountains and Shenandoah National Parks to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040BWC emissions and scaled boundary conditions (yellow). ................................... 9-19 Sensitivity of NO3 wet deposition at Great Smoky Mountains and Shenandoah National Parks to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040BWC emissions and scaled boundary conditions (yellow). ................................... 9-19

Figure 9-7 Figure 9-8 Figure 9-9 Figure 9-10 Figure 9-11 Figure 9-12

Figure 9-13

Figure 9-14

Figure 9-15

Figure 9-16

Figure 9-17

Figure 9-18

Figure 9-19

Figure 9-20

Figure 9-21

Figure 9-22

VIII

ODMAN,

ET AL.

Figure 9-23

Figure 10-1 Figure 10-2 Figure 10-3 Figure 10-4 Figure 10-5 Figure 10-6 Figure 10-7 Figure 10-8

Figure 10-9

Figure 10-10 Figure 10-11 Figure 10-12 Figure 10-13 Figure 10-14

Figure 10-15

Figure 10-16 Figure 10-17 Figure 10-18

Figure 10-19



+

Sensitivity of NH4 wet deposition at Great Smoky Mountains and Shenandoah National Parks to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040BWC emissions and scaled boundary conditions (yellow). ................................... 9-20 Daily maximum ozone concentrations on July 12 when 2010-OTW emissions are used. ...................................................................................................... 10-1 Estimated change in daily maximum ozone concentrations from July 12, 1995 to 2010 under the OTW strategy. ..................................................................... 10-2 Diurnal variations of ozone using OTW, BWC, and BB emission strategies in the years 2010 and 2040 for the July 11-19, 1995 episode at Look Rock, TN. ........ 10-2 Simulated seasonal cumulative ozone (W126) for the years 2010 and 2040 under the OTW, BWC and BB strategies. .......................................................... 10-3 Daily average PM2.5 concentrations over the SAMI region on July 15, 1995 using the 2010-OTW emission strategy. ............................................................ 10-4 Change in daily average PM2.5 concentrations from July 15, 1995 to 2010 under the OTW emission strategy. ............................................................................ 10-5 Modeled sulfate concentrations at Great Smoky Mountains for the basecase and the OTW, BWC, and BB emission strategies in 2010. ................................... 10-6 Model estimates for annual average PM2.5 concentrations at Shenandoah National Park for the basecase and OTW, BWC, and BB emission strategies for 2010 and 2040. ......................................................................................... 10-7 Model estimates for annual average PM2.5 concentrations at Great Smoky Mountains National Park for the basecase and OTW, BWC, and BB emission strategies for 2010 and 2040. .......................................................................... 10-7 Estimated weekly cumulative sulfate wet deposition for July 11-18, 2010 under the OTW strategy........................................................................................... 10-9 Estimated change in weekly cumulative sulfate wet deposition for July 11-18 from 1995 to 2010 under the OTW strategy. ...................................................... 10-9 Estimated weekly cumulative SO2 dry deposition for July 11-18, 2010 under the OTW strategy. ..............................................................................................10-10 Estimated change in weekly cumulative SO2 dry deposition for July 11-18 from 1995 to 2010 under the OTW strategy. ............................................................10-10 Modeled weekly cumulative sulfate wet deposition fluxes at Great Smoky Mountains for the basecase and the OTW, BWC, and BB emission strategies in 2010 and 2040. ............................................................................................10-11 Modeled weekly cumulative nitric acid dry deposition fluxes at Shenandoah for the basecase and the OTW, BWC, and BB emission strategies in 2010 and 2040............................................................................................................10-12 Annual average wet and dry deposition mass fluxes of sulfur at Great Smoky Mountains for the basecase and the three strategies in 2010 and 2040. ................10-13 Annual average wet and dry deposition mass fluxes of sulfur at Shenandoah for the basecase and the three strategies in 2010 and 2040. ....................................10-13 Annual average wet and dry deposition mass fluxes of oxidized and reduced nitrogen at Shenandoah for the basecase and the three strategies in 2010 and 2040............................................................................................................10-14 : Annual average wet and dry deposition mass fluxes of oxidized and reduced nitrogen at Great Smoky Mountains for the basecase and the three strategies in 2010 and 2040. ............................................................................................10-14 Change in particulate sulfate concentrations due to 30% reduction in SO2 emissions using DDM-3D (left) and brute force (right).......................................... 11-5 Change in sulfate wet deposition concentrations due to 30% reduction in SO2 emissions using DDM-3D (left) and brute force (right).......................................... 11-6 Source regions for which emission sensitivities were calculated. ........................... 12-1 Daily average PM2.5 and its change on July 15, 1995 for a 10% reduction of the 2010-OTW SO2 emissions from SAMI states. .................................................... 12-4

IX

ODMAN,

ET AL.


Figure 12-5


Figure 12-8

Figure 12-9



Figure 12-14 Figure 12-15 Figure 12-16


Figure 12-21

Figure 12-22

Figure 12-23


Geographic location of the Class I receptor sites where aerosol and wet deposition sensitivity analysis was performed (reproduced from SAMI, 2001). ........ 12-5 Daily average fine sulfate concentrations (* ) and absolute sensitivities for each classified day at Great Smoky Mountains to a 10% reduction in SO2 emissions from each geographic sub-domain. ................................................................... 12-6 Daily average fine sulfate concentrations (* ) and sensitivities for each classified day at Great Smoky Mountains to a 10% reduction in SO2 emissions from each geographic sub-domain. ................................................................................. 12-6 Annual average fine sulfate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic sub-domain. ...... 12-7 Annual average fine sulfate concentrations (* ) and sensitivities for each class at Great Smoky Mountains to a 10% reduction in SO2 emissions from each geographic sub-domain. ................................................................................. 12-7 Annual average fine nitrate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic sub-domain. Note, in this case nitrate increases as SO2 is decreased. ..................................... 12-8 Annual average fine ammonium concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic subdomain. ........................................................................................................ 12-8 Annual average fine PM concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic sub-domain. ............... 12-9 Annual average fine PM concentrations (* ) and normalized sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic subdomain. .......................................................................................................12-10 Percent change in daily average PM2.5 on July 15, 1995 for a 10% reduction of the 2010-OTW elevated NOx emissions from eight SAMI states...........................12-10 Annual average fine PM2.5 concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domain. .................................................................................................12-11 Percent change in daily average PM2.5 on July 15, 1995 for a 10% reduction of the 2010-OTW ground-level NOx emissions from eight SAMI states. ....................12-12 Daily average fine sulfate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domain. .....12-13 Annual average fine PM concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in ground level NOx emissions from each geographic subdomain. .......................................................................................................12-13 Daily average fine nitrate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domain. .....12-14 Daily average fine ammonium concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domain. .....12-14 Daily average fine PM concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domain. ..............12-15 2Cumulative wet deposition of SO4 and its change for the week of July 11-18, 1995 for a 10% reduction of the 2010-OTW total SO2 emissions from SAMI states. .........................................................................................................12-16 Annual weekly average sulfate wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic subdomains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park.............................................................12-17 Weekly cumulative sulfate wet deposition fluxes (* ) and sensitivities for each classified episode at Elkmont (Great Smoky Mountains) to a 10% reduction in SO2 emissions from each geographic sub-domain. ...........................................12-17 Cumulative wet deposition of NO3 and its change for the week of July 11-18, 1995 for a 10% reduction of the 2010-OTW elevated NOx emissions from SAMI states. .........................................................................................................12-18

X

ODMAN,

ET AL.


Figure 12-24 Annual weekly average nitrate wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park........................................12-19 Figure 12-25 Cumulative wet deposition of NO3 and its change for the week of July 11-18, 1995 for a 10% reduction of the 2010-OTW ground-level NOx emissions from SAMI states. ................................................................................................12-20 Figure 12-26 Annual weekly average nitrate wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in ground level NOx emissions from each geographic sub-domains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park........................................12-21 Figure 12-27 Annual weekly average ammonium wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park. ......................................................12-22 Figure 12-28 O3 (W126) and its change on July 12, 1995 for a 10% reduction of the 2010OTW elevated NOx emissions from SAMI states................................................12-23 Figure 12-29 Seasonal average ozone W126 (* ) and sensitivities for 18 Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domain. .......12-24 Figure 12-30 O3 (W126) and its change on July 12, 1995 for a 10% reduction of the 2010OTW ground-level NOx emissions from SAMI states ..........................................12-25 Figure 12-31 Seasonal average ozone W126 (* ) and sensitivities for 18 Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domain. .......12-26 Figure 12-32 Sensitivity of various pollutants to a 10% reduction in SO2 and NOx emissions from Tennessee. ...........................................................................................12-28 Figure 12-33 Fraction of domain wide sensitivities attributed to 10% reductions in SO2 and NOx emissions from Tennessee. .....................................................................12-28 2 Figure 17-1 Dry deposition fluxes (mg/m ) of SO2 for the week of July 11-18, 1995. .................. 17-1 Figure 17-2 Nitric acid dry deposition mass flux at Oak Ridge, TN.......................................... 17-2 Figure 17-3 Sulfur dioxide dry deposition mass flux at Oak Ridge, TN..................................... 17-2

XI

ODMAN,

ET AL.


This page is left blank intentionally.

XII

ODMAN,

1

ET AL.


EXECUTIVE SUMMARY

1.1 Introduction Studies that have been conducted in national parks, forests and wilderness areas of the southern Appalachian Mountains have documented adverse effects to visibility, streams, soil, and vegetation. Poor air quality in the region has been implicated as a major source of these adverse effects. Beginning in 1990, the Federal Land Managers for Shenandoah National Park, Great Smoky Mountains National Park, and Jefferson National Forest/James River Face Wilderness Area made several adverse impact determinations in their review of proposed air permits for major new sources. Although it is known that the air pollution levels, which currently affect park and wilderness resources, come from existing sources of pollution — large and small, mobile and stationary, near and distant — the relative contribution of each source type to the regional air pollution problem is not well quantified. The 1990 Clean Air Act Amendments (CAAA) require major emissions reductions for primary airborne pollutants, including sulfur oxides (SOx ), nitrogen oxides (NOx ) and volatile organic compounds (VOCs). Although the reductions are expected to produce air quality improvements, it is uncertain whether the results will be enough to protect and preserve the ecosystems and natural resources of the Southern Appalachians, especially in Class I areas. As part of the Southern Appalachian Mountains Initiative (SAMI), an Integrated Assessment is conducted which will model and assess the environmental and socioeconomic responses to changes in atmospheric emissions, which result from various emissions management strategies. One component of the Integrated Assessment is the atmospheric modeling of air quality responses to emissions controls. The goals are to characterize the air pollution processes that affect air quality in the southern Appalachian Mountains, assess the sensitivities to changes in emissions, and to model the impact of future emission changes on air quality SAMI's approach calls for the modeling of a limited number of historic episodes that best characterize recent (1991-95) levels of ozone, visibility and acid deposition in Class I areas of the region. Using data classification and optimization techniques, nine episodes, each 6 to 10 days long, were selected to represent the seasonal and annual air quality metrics most relevant to visibility, stream, and forest effects. Weights were assigned to each one of these episodes so that seasonal and annual averages can be calculated from episodic modeling results. The air quality modeling reported here provided aerosol concentrations needed for visibility calculations, but it did not directly calculate visibility such as extinction coefficients or deciviews.

1.2 Air Quality Model The first objective was to develop a "one-atmosphere" modeling system for the collective study of ozone, particulate matter (PM) and acid deposition. The Urban-to-Regional Multiscale (URM) model, a photochemical air quality model that could only address ozone, was upgraded to include aerosol dynamics, wet deposition scavenging processes, and heterogeneous sulfate chemistry. The new version is called the URM “one-atmosphere” model or URM-1ATM. The original version of URM featured a finite-element advection scheme, which allows multi-scale grids without nesting, a gradient transport (K-theory) formulation for turbulent mixing, a gas-phase chemistry mechanism with detailed reactions for organic compounds (SAPRC), and a resistance-based dry deposition formulation. URM1ATM features two new modules: 1) the aerosol module and, 2) the wet deposition module. The aerosol module uses a sectional approach to characterize the continuous PM size distribution. All particles within a section or bin are assumed to have the same composition. In this 1-1

ODMAN,

ET AL.


study, four size bins were used but this number can be increased if more resources are available. The size bins were defined according to particle diameters as follows: 1) less than 0.156 µm, 2) 0.156-0.625 µm, 3) 0.625-2.5 µm, and 4) 2.5-10.0 µm. The first three size bins correspond to PM2.5 while the combination of all four bins corresponds to PM10. Inorganic aerosols are treated as an equilibrium system of sulfate, nitrate, ammonium, chloride, sodium and water (ISORROPIA). Organic PM is lumped into a single condensed specie that forms through VOC oxidation in the gas-phase chemistry mechanism. Gravitational settling of PM is treated with a resistance approach similar to the dry deposition of gas-phase species. The wet deposition module (Reactive Scavenging Module) treats the clouds as either stratiform or convective. Transport in convective clouds allows large updrafts to carry low-level air to the upper layers. Scavenging processes within the module include gas, aerosol, and microphysical scavenging. Scavenging of gas-phase species by cloud water is modeled via an equilibrium process that is based on species solubilities, while scavenging by rain water is modeled for most species via mass transfer. Aerosol scavenging is treated by nucleation and by inertial impaction processes. The scavenged species are sulfur dioxide, particulate sulfate, ozone, nitric acid, particulate nitrate, hydrogen 2+ 2+ peroxide, ammonia, particulate ammonium, and soluble crustals (Mg and Ca ).

1.3 Model Setup and Inputs SAMI's modeling approach consisted of simulating nine episodes (one winter, four spring and four summer episodes) and weighting the results to generate seasonal (for ozone) and annual (for particulate matter and acid deposition) averages. The modeling domain covered the eastern half of the United States. A multiscale grid placed the finest horizontal resolution of 12 km over the SAMI region and coarsened gradually towards the boundaries of the domain. Seven unequally spaced vertical layers with finer resolution near the surface extended to a height of approximately 13 km. Meteorological and emission inputs to URM-1ATM were generated using the Regional Atmospheric Modeling System (RAMS) and Emission Modeling System (EMS-95), respectively. A modified version of RAMS was run in the non-hydrostatic mode with cloud and rainwater microphysics activated. A system of three nested grids provided data at the resolution required by the URM-1ATM grid. The National Centers for Environmental Prediction (NCEP) reanalysis data was used for model initialization as well as data assimilation. Performance statistics were calculated based on all available surface National Weather Service observations within the 12-km nest of RAMS for each episode. This performance evaluation was reported separately (Doty et al., 2001). A summary of biases and root mean square errors for 2-m temperatures, 2-m mixing ratios, wind directions at 10-m, -1 10-m wind speeds (m s ) and precipitation can be found in Section 4.3. Overall, meteorological performance was considered to be adequate for the SAMI modeling requirements and consistent with the current capabilities of meteorological models. EMS-95 was used to generate speciated hourly gridded emission inputs to URM-1ATM. Emission sources are separated into two categories: elevated and ground-level sources. Ground-level sources include low-level point, mobile, anthropogenic area, non-road mobile, and biogenic sources. Point-source and area-source emissions estimates were based on data developed by the Pechan/Avanti Group, as were on-road mobile source data. EMS-95's Motor Vehicle Emissions Model (MoVEM) uses U.S. EPA's MOBILE-5b to compute vehicle-dependent emissions factors of CO, NOx, and total organic gases (TOG). Biogenic emissions were estimated using U.S. EPA’s Biogenic Emissions Inventory System (BEIS-2). The point source emissions estimates were enriched with day specific emissions data obtained from major utility companies in the modeling region. Meteorological model results were used to estimate the temperature and radiation dependent biogenic emissions and the temperature dependent on-road mobile source emissions. This report includes summary tables of emission inputs by general category as well as spatial plots (maps) of elevated and ground-level NOx , SO2 and ground-level NH3 emissions. 1-2

ODMAN,

ET AL.


Initial and boundary conditions are the other important inputs to URM-1ATM. To derive these conditions for the gaseous species, data from the Aerometric Information Retrieval System (AIRS) and the North American Research Study on Tropospheric Ozone for the Northeast (NARSTO-NE) archives were used. Considering that the measurements are made closer to urban locations, special interpolation techniques were developed for SO2, NOx and ozone to minimize potential biases in rural areas. A two-day ramp-up period was used before each episode to dampen the effect of initial conditions. The initial and boundary conditions for PM were set using data from the Interagency Monitoring of Protected Visual Environments (IMPROVE) network. Unfortunately, this network was not as dense as the AIRS network for the 1991-1995 period. In general, particulate species have longer lifetimes than the gaseous species; therefore, the uncertainty in their initial and boundary conditions may have a larger impact on modeled concentrations.

1.4 Model Performance The model estimates of ozone were compared to observations from the AIRS network at urban, rural and high-elevation sites. The agreement between the model estimates and observations was good for eight of the modeled episodes but poor for one episode. The model usually underestimated ozone peaks and overestimated nighttime ozone; however, the daytime variations and the timing of the peaks were in good agreement with the observations. For the July 1995 episode, comparisons were made to the ozone and NOx data collected during the Southern Oxidant Study (SOS) and the simulated concentrations were found to be in good agreement. Statistical performance measures such as the normalized bias and error were calculated. The model's ability to estimate ozone was within EPA’s guidance criteria for urban-scale modeling with normalized biases generally within ±15% and normalized errors less than 35%. This was encouraging given that the finest scale used in this application (12 km) is much larger than those typically used for urban-scale modeling. Finally, ozone modeling performance was also analyzed based on the maximum daily 8-hour average ozone concentration. The model consistently overestimated the maximum 8-hr concentrations below 40 ppb and underestimated ozone above 80 ppb for all episodes. The levels and composition of PM2.5 were evaluated by using IMPROVE measurements taken at Class I areas of the region. The sulfate, ammonium, elemental and organic carbon components that form 75% of PM2.5 were all estimated with normalized mean errors around 40%. The errors for less abundant components such as soils and nitrate were larger. Modeled soils consist of more species than those measured by IMPROVE and their emissions are highly uncertain. Nitrate constitutes a very small fraction of PM 2.5 in the SAMI region and its concentration is sensitive to small errors in other constituents such as sulfate and ammonium. In general, low concentrations of any component are overestimated and high concentrations are underestimated. The underestimation in PM2.5 may be, in part, due to water, which is believed to be the primary unidentified component of IMPROVE measurements, not being included in the model estimates. The errors in PM 2.5 and PM10 are similar to those obtained by other models. This is encouraging since SAMI episodes are typically longer than those of previous studies and they cover a wide range of meteorological conditions. Also, the SAMI modeling domain covered a much larger area than most other studies conducted in well-defined airsheds such as the Los Angeles basin. Wet deposition performance was evaluated using weekly cumulative observations from the National Atmospheric Deposition Program (NADP) monitoring network. Since missing the location of the precipitation by a couple of grid cells can lead to large errors, the “best” (i.e., closest in magnitude to the observation) model result within a 30 km radius of the monitoring site was used instead of the “cell” value. Simulated wet sulfate and nitrate deposition fluxes were within 25% of the observations. The amount of precipitation (calculated by RAMS and input to the URM-1ATM) was often overestimated when the observed rainfall was less than 14 mm and underestimated as the observed precipitation increased. The trends in sulfate and nitrate biases were very similar to this trend. Therefore, if the bias in precipitation could be reduced, the performance of many deposition species would likely improve. 1-3

ODMAN,

ET AL.


Ammonium wet deposition flux, which is an order of magnitude smaller than the sulfate and nitrate fluxes, was biased high. Hydrogen ion deposition was biased low because ammonium, calcium, and magnesium were biased high. Ambient pollutants were generally overestimated when observed levels were low and underestimated when they were high. Ozone clearly followed this trend, and it presents a challenge for subsequent SAMI analyses that rely on model outputs. Many PM species (including sulfate and organics) and wet deposition species also exhibited similar behavior. Modeling limitations likely contributing to such behavior include coarse vertical and horizontal grid resolution, reduced temporal and spatial variance in emissions compared to actual values, spatial and temporal smoothing of turbulent mixing, extremes in point measurements usually having a greater range than concentrations averaged over the modeling grid cell, and the inability to reproduce some meteorological and chemical processes occurring at sub-grid scales. However, some positive bias for low pollutant levels, especially involving PM, may also be caused by boundary conditions that were too high for specific episodes. Overall, performance was judged to be acceptable for all pollutants. The results of the performance evaluation suggest that the atmospheric modeling component will likely introduce a nonnegligible, but not dominating, uncertainty to SAMI’s integrated assessment. The primary sources of this uncertainty include inadequate grid resolution, the inability to model accurately some important processes such as precipitation and uncertain emission inventories. The integrative nature of the atmospheric modeling system propagates the uncertainties in emissions inventories, which have not been separately quantified. Related studies of the region using similar inventories suggest that there are likely significant uncertainties in emissions.

1.5 Model Inputs for Future Year Simulations Assuming that the selected nine episodes are climatologically representative, the same meteorological inputs used for the basecase simulations were also used for future year simulations. It is also assumed that any potential feedback of future landuse and emissions on the meteorology would be negligible. Three emission strategies were developed by the Policy Committee to reflect SAMI’s assumptions about future growth and the implementation of regulations and incentives. Each strategy proposes progressively more stringent emission controls in utility, industrial, highway vehicle, non-road engines, and area source categories for the years 2010 and 2040. From least stringent to most stringent, the three strategies are “on the way” (OTW), “bold with constraints” (BWC), and “beyond bold” (BB). The total annual emissions of sulfur SO2, NOx, VOC, PM2.5 and NH3 from the eight SAMI states under each strategy for the years 2010 and 2040 are illustrated in Figure 1-1. The levels of these emissions in 1990, which was the base year from which episodic emissions were grown, are also shown for comparison. The implementation of certain regulations leads to seasonal variations in emissions. For example, ozone regulations require larger NOx emission reductions during the ozone season (May-September). Sensitivity analyses were conducted to decide how to set the initial and boundary conditions for future year simulations. It was found that the model estimated future year pollutant levels would be sensitive to both initial and boundary conditions. The initial conditions for SO2, particulate sulfate, and particulate ammonium were reduced for all future year simulations in proportion to SO2 emission reductions. Similarly, the initial conditions for NOx were scaled by the reductions in NOx emissions. However, the boundary conditions were kept the same as those used for basecase simulations. This decision was made based on the uncertainty in determining future pollutant levels for Canada, Mexico and the western U.S.

1-4

ET AL.

Million Tons/Year

ODMAN,

7.0


Emissions in the SAMI States

6.0 5.0 4.0 3.0 2.0 1.0 0.0

SO2 1990 Figure 1-1

2010 OTW

NOx 2010 BWC

VOC 2010 BB

PM2.5

2040 OTW

2040 BWC

NH3 2040 BB

Total annual emissions from the eight SAMI states in 1990 and projections for 2010 and 2040 under the OTW, BWC, and BB strategies. VOC emissions reported here do not include biogenic emissions.

1.6 Response to Future Year Emission Strategies The responses to future year emission strategies were analyzed at selected receptors within the SAMI region. Responses were different in magnitude and sometimes in direction depending on the location of the receptor. During the July 1995 episode, which had the highest ozone concentrations, the daily maximum ozone at Great Smoky Mountains National Park (GRSM) is estimated to decrease by 10-15 ppb in 2010 and 10-25 ppb in 2040 from the baseyear levels, depending on the strategy. The BB strategy leads to the largest decreases followed by the BWC and OTW strategies. Overall, significant decreases in cumulative ozone W126 are estimated to occur by 2010 (OTW strategy) due to expected NOx emission reductions. Without additional action (e.g. BWC or BB emission control strategies), ozone W126 is estimated to remain nearly unchanged from 2010 to 2040. Fine particulate sulfate is estimated to decrease in 2010 and drop further in 2040. As the stringency of the strategy increases so does the reduction in sulfate levels at several Class I areas. Nitrate levels, on the other hand, are estimated to increase in the southeastern United States (e.g., GRSM). The increase in particulate nitrate is most likely due to additional ammonia becoming available in response to reductions in SO2 and increases in NH3 emissions. The Class I areas in northern SAMI states (e.g., SHEN) are not expected to see this nitrate increase in 2010 but perhaps in 2040. The model estimated an increase in annual average PM2.5 at GRSM in 2010 under the OTW strategy; other strategies are estimated to lead to decreases. In 2040, the model estimated decreases between 8% (OTW strategy) and 40% (BB strategy). At SHEN future PM2.5 levels are estimated to decrease between 10% (2010-OTW) to 44% (2040-OTW). The episode selection process considered wet deposition but not the dry deposition. For this reason, the ratio of estimated annual average dry deposition to wet deposition may be biased low. Nevertheless, both wet and dry sulfate deposition estimates show significant reductions for each of the emission strategies. The relative magnitudes of these reductions between different strategies are similar to the relative reductions in SO2 emissions for each strategy. Based on model results, total sulfur 1-5

ODMAN,

ET AL.


deposition fluxes are expected to decrease more in the northern SAMI states (e.g., SHEN) than in the Southeast (e.g., GRSM). The model estimated an increase in the dry deposition of reduced nitrogen with all future year strategies. This is due to the increase in ammonia emissions. Wet deposition of reduced nitrogen is also estimated to increase under the 2010-OTW, 2010-BWC and 2040-OTW strategies. Dry deposition of oxidized nitrogen is estimated to decrease but the wet deposition of oxidized nitrogen is estimated to remain unchanged. The latter was a surprising result given the level of NOx reductions in the SAMI states. However, NOx reductions are not as extensive outside the SAMI states and the increases in NH3 emissions are probably compensating for the NOx reductions. This result also suggests that the source of wet oxidized nitrogen deposition at Class I areas of the SAMI region is outside the SAMI states. The increases in reduced nitrogen deposition may compensate or even exceed the decreases in oxidized nitrogen resulting from NOx emission reductions.

1.7 Sensitivity Analysis URM-1ATM is equipped with a direct sensitivity analysis module that can calculate sensitivities of concentrations and deposition fluxes to various emissions simultaneously. In this study, the module was extended to calculate particulate matter and wet deposition sensitivities in addition to gas-phase sensitivities. Traditional sensitivity analysis involves running the model by perturbing one type of emissions at a time and comparing the results to the original run with unperturbed emissions. URM1ATM's sensitivity analysis module is based on the decoupled direct method (DDM-3D). After the governing equations of the model are solved, a second set of equations is solved for sensitivities at each time step. This makes DDM-3D much more efficient than traditional sensitivity analysis. The sensitivities are defined as local derivatives of concentrations with respect to emissions. As such, they are only first-order accurate approximations to how the model would respond to actual emission perturbations. If the perturbation is large, the error in the approximation grows. The error would grow faster for more non-linear relationships between the concentrations and emissions. A comparison between traditional and DDM-3D sensitivity analyses showed that DDM-3D sensitivities are reliable up to about a 30% change in emissions for most of the relationships that would be discussed below. Extreme caution should be used in superposing the sensitivities to emissions from different geographic region or sensitivities to emissions from different source categories. This information can be used to guide the development of control strategies, since it gives some directional sense to where to look for emission reductions. However, DDM-3D should not be substituted for full-scale modeling to demonstrate the effectiveness of a given strategy. Using DDM-3D, the sensitivities of ozone, particulate matter and acid deposition to SO2, NOx and NH3 emissions were calculated. These sensitivities were scaled and reported as responses to 10% reductions in emissions. Note that this is not the same thing as the actual response of the model to 10% emission reductions but a first-order approximation to that response. Each individual SAMI state 1 2 3 4 and five surrounding regions (Central ; Midwest ; Northeast ; Southeast regions and the rest of the modeling domain) were targeted as different geographic source areas for the emission reductions. The NOx sources were further discriminated as elevated and ground-level sources. Emission reductions from different geographic areas displayed very different levels of impact on various sites within the SAMI region. In general, sites showed the greatest response to emission reductions in the nearest areas. Based on the source area they responded to the most, Class I areas generally fell under three

1

Louisiana, Arkansas, Missouri, Iowa and the portions of Minnesota, Texas, Oklahoma, Kansas and Nebraska that are in the modeling domain 2 Ohio, Michigan, Indiana, Illinois and Wisconsin 3 Maryland, Delaware, District of Columbia, Pennsylvania, New Jersey, New York, Connecticut, Massachusetts, Rhode Island, Vermont, New Hampshire and Maine 4 Florida and Mississippi

1-6

ODMAN,

ET AL.


categories: 1) Alabama and Georgia sites, 2) Tennessee and North Carolina sites, and 3) Virginia and West Virginia sites. 1.7.1

PM Sensitivities

Annual average concentrations of fine sulfate PM were reduced between 4.0% and 7.3% in response to 10% reductions in SO2 from each source area. Sipsey, AL and Cohutta, GA show the greatest response to reductions from Alabama and Georgia. The sites in North Carolina and Tennessee show the greatest response to emission reductions in Tennessee. The sites in Virginia and West Virginia show the greatest response to emission reductions in the Midwest region and West Virginia. Annual average concentrations of particulate nitrate increased between 1.5% and 4.0%, ammonium decreased between 2.0% and 4.0%, and PM2.5 decreased between 1.7% and 2.7% in response to 10% reductions in SO2 from the entire domain. The fraction of the reductions in PM2.5 connected to different regions (or sub-domains) is almost identical to the contributions to the reductions in sulfate PM. Annual average PM2.5 reductions at Class I areas in the SAMI region were less than 0.5% in response to a 10% domain-wide reduction of elevated as well as ground level NOx emissions. PM2.5 reductions were due to reductions in nitrate (largest decrease), ammonium and organic PM. The responses to local (i.e., from SAMI states) emissions of ground-level NOx were at least 5 times larger than the responses to local elevated NOx emissions. The largest reductions of PM2.5 at Sipsey and Cohutta were attributed to ground level and elevated NOx emissions in Alabama and Georgia and elevated NOx emissions in the Central region. The PM2.5 reductions in Tennessee and North Carolina sites were, to a large extent, due to ground-level NOx reductions from Tennessee. The largest fraction of PM2.5 reductions at the Virginia and West Virginia sites were attributed to reductions of elevated NOx emissions from West Virginia and both elevated and ground-level emissions in the Midwest region. The responses to 10% reductions in NH3 emissions were mostly local. Class I areas in Virginia and West Virginia were the only ones that showed any significant response to reductions of non-local NH3 emissions, namely from the Midwest and Northeast regions. Annual average PM concentrations showed a decrease of up to 1.4% for sulfate, between 2.0% to 3.5% for nitrate, between 1.0% and 2.0% for ammonium, and between 0.4% and 1.1% for PM 2.5. 1.7.2

Wet Deposition Sensitivities

Annual sulfate wet deposition fluxes decreased between 4.0% and 9.2% at eleven Class I areas in response to a 10% reduction in SO2 emissions. In comparison to sulfate PM sensitivities to SO2 emissions, the sulfate wet deposition sensitivities were more localized. The impact of SO2 emissions from Alabama and Georgia on the sulfate wet deposition at Tennessee and North Carolina sites were substantially larger than the impact of SO2 emissions from Tennessee. This suggests that most of the rain at these sites can be attributed to air masses coming from the Gulf of Mexico. Annual nitrate wet deposition fluxes decreased between 0.6% and 1.7% at the Class I sites of the SAMI region in response to a 10% reduction in elevated NOx emissions domainwide. At the Tennessee and North Carolina sites, the portion of the reduction in nitrate wet deposition that is attributed to the emission reductions in Alabama and Georgia was the largest. At Sipsey, AL and Cohutta, GA, emission reductions from the Central region played a significant role. At the Virginia and West Virginia sites, the reductions were mostly due to elevated NOx emission reductions in West Virginia and the Midwest region. Reduction of ground level NOx emissions by the same percentage amount (i.e., 10%) resulted in a larger decrease in nitrate wet deposition at all of the eleven Class I sites. The contribution of the Central region to the reductions at the Alabama, Georgia, Tennessee and North Carolina sites was smaller and the contributions of Tennessee and Georgia were larger than they were for the elevated NOx emission reductions. At the Virginia and West Virginia sites, the contribution 1-7

ODMAN,

ET AL.


of West Virginia was larger for a 10% reduction of elevated NOx emissions than a reduction, by the same percentage amount, of ground level NOx emissions. This is because of the elevated NOx emissions as tons of NOx being much larger in this state than ground level emissions. Annual average wet deposition fluxes of sulfate and ammonium showed an insignificant response to both elevated and ground level NOx emission reductions. Annual wet deposition of ammonium is estimated to decrease between 1.0% and 1.3% at Class I areas of the SAMI region as a result of a 10% reduction in NH3 emissions domainwide. The sensitivity of nitrate wet deposition to NH3 emissions were poorly quantified by the DDM-3D method. Limited analysis has shown that the actual response of nitrate wet deposition to increases in NH3 emission may reduce the benefits of NOx emission reductions. 1.7.3

Ozone Sensitivities

The seasonal average sensitivity of daily cumulative ozone W126 to 10% reductions in elevated and ground level NOx emissions were calculated at 18 Class I areas. The sensitivities to elevated NOx emissions were between 3.0% and 3.5%. Approximately one half of this amount was connected to emissions from the five regions outside the SAMI states. The sensitivities to ground level NOx emissions ranged from 7.0% to 8.0%. These sensitivities were generally more local than sensitivities to elevated NOx emissions except in West Virginia where the NOx emissions from the Midwest region had the largest contribution to ozone W126. For the other Class I areas, ground level NOx emissions from Georgia or Tennessee had the largest contribution. The average sensitivity to a 10% reduction of total (elevated plus ground level) NOx emissions was a 10% decrease in ozone W126.

1.8 Conclusion An air quality model (URM-1ATM) has been developed for an integrated study of ozone, PM and acid deposition. In combination with a meteorological model (RAMS) and an emissions model (EMS-95), this air quality model was applied to eastern U.S. to simulate winter, spring and summer episodes in the Southern Appalachian Mountains. The model performance either met EPA's criteria (for ozone) or, in the absence of guidance (for PM and wet deposition), was comparable or better than the performance of other models in similar studies. The model was used to assess how various emission strategies would affect the seasonal ozone, annual PM2.5, and annual acid deposition levels at Class I areas of the Southern Appalachian Mountains in the years 2010 and 2040. It was found that the air quality of the region would generally improve in the future due to regulations mandated under the Clean Air Act Amendments of 1990 and other recently promulgated regulations. In general, more stringent emission controls should provide additional improvements. In response to SO2 emission controls, annual average sulfate PM concentrations and sulfate wet deposition fluxes are expected to decrease significantly in Class I areas. However, SO2 controls and increasing ammonia emissions may result in an increase of nitrate PM levels due to more ammonia becoming available to react with nitric acid to form particulate nitrate. The estimated decrease in oxidized nitrogen deposition is approximately equal to the estimated increase in reduced nitrogen deposition. Therefore, changes in total nitrogen deposition are expected to be minimal, unless ammonia emissions are controlled. Using the direct sensitivity analysis feature of the model (DDM) important source-receptor relationships were quantified. The sensitivities of ozone, PM, and wet deposition levels in the year 2010 to 10% reductions (beyond the OTW strategy) of SO2, NOx , and NH3 emissions from eight SAMI states and five surrounding regions were estimated. It was found that sulfate wet deposition would decrease by an additional 7% and sulfate PM by an additional 5% in response to 10% SO2 emission reductions domain wide. On the other hand, particulate nitrate would increase by an additional 2.5%. Additional 1-8

ODMAN,

ET AL.


NOx emission reductions would be most beneficial to seasonal cumulative ozone: a 10% decrease would result from a 10% reduction. A large fraction of this decrease can be attributed to reductions in elevated NOx emissions from regions outside the SAMI states. Reductions of NOx emissions would also affect nitrate, ammonium and organic PM, and nitrate wet deposition but to a lesser proportion. The DDM sensitivity results indicate that the benefits of an additional 10% reduction in NH3 emissions would be mostly local and marginal. However, the brute force sensitivity results indicate a larger response to reductions in NH3 emissions. According to the DDM sensitivity analysis, reductions of SO2 emissions would yield a larger reduction in PM2.5 concentrations than the reductions in NOx and NH3 emissions. These results, especially specific sensitivities to emission reductions from geographic subdomains can provide guidance for the design of emission control strategies. However, they should not be used as a substitute for full-scale modeling to determine the effectiveness of a control strategy.

1-9

ODMAN,

ET AL.



1-10

ODMAN,

2

ET AL.


INTRODUCTION

2.1 Background Elevated levels of particulate matter, acid deposition, and ozone adversely impact human health, materials, and sensitive ecosystems. Designing control strategies to reduce these pollutants is difficult because the complex chemical and physical processes that govern their formation, transport, and removal result in highly non-linear responses. Atmospheric modeling can accurately simulate these complex processes and can be used to assess the effectiveness of emission control strategies. In the past, air quality model applications focused either on ozone (Russell and Dennis, 2000), aerosols (Seigneur et al., 1999), or acid deposition (NAPAP, 1990) individually and not concurrently under a “one-atmosphere” approach. Until recently, the concept of “one-atmosphere” modeling did not seem practical due to the large computational resources required for particulate matter modeling (Zhang et al., 1999). Now that computational power is more readily available, “one-atmosphere” models are beginning to emerge (Byun and Ching, 1999). Results from these “one-atmosphere” models can be used by policy-makers to make more fully-informed decisions taking into account responses of multiple endpoints to emission changes. Ozone concentrations in the troposphere typically fall into the range of 10 to 100 ppb (part per billion) (Warnek, 1988). However, elevated ozone concentrations can occur in (and downwind of) urban areas and other source regions due to high emissions of nitrogen oxides (NOx ) and volatile organic compounds (VOCs) that react in the presence of sunlight to form ozone. These reactions can lead to ozone levels of greater than 120 ppb, the current one-hour standard in the United States. Ozone is a nose, eye, and throat irritant and can cause chest constriction and accelerated lung aging (Lippmann, 1989; USEPA, 1996). Ozone can also cause chemical deterioration in materials (Seinfeld, 1986) and damage to vegetation (Heck et al., 1982). An aerosol is a dispersion of microscopic solid and/or liquid particles in a gaseous media. Particulate matter (PM) refers to these suspended microscopic solid and/or liquid particles, except uncombined water, at standard conditions. Particles with an effective aerodynamic diameter of less than 2.5 µm are referred to as fine PM (or PM 2.5). Recently, the U.S. Environmental Protection Agency (EPA) promulgated National Ambient Air Quality Standards (NAAQS) for PM2.5 that include long-term 3 3 and 24-hour average standards of 15 µg/m and 65 µg/m , respectively. Fine particles can be directly emitted into the atmosphere (primary) or formed in the atmosphere due to gas-to-particle transformations (secondary). Primary particulate matter includes, but is not limited to sulfate, elemental carbon, organic carbon, crustal materials (calcium, sodium, magnesium, potassium, etc.), and metals (aluminum, iron, lead, etc.). The major contributors to secondary particulate matter include inorganic (sulfate, nitrate, and ammonium) and organic (semi-volatile oxidation products of VOCs) compounds. Fine particulate matter is of concern because it can cause damage to the respiratory system (Dockery et al., 1993; USEPA, 2002), can affect materials by soiling (NAPAP, 1990), and reduces visibility (NRC, 1993). Gas and particulate species can be deposited from the atmosphere to the ground by wet and/or dry deposition processes. Wet deposition of pollutants occurs when gas and particulate species are scavenged (absorbed into water droplets) followed by the droplet removal by precipitation or impaction. Gas phase scavenging by cloud water is dependent on the solubility of the individual species. Particulate scavenging is due to inertial impaction and diffusion. Dry deposition refers to the removal of species not in a water droplet at the earth’s surface by soil, water, or vegetation. Dry deposition removal rates are dependent on meteorological conditions and individual properties of the gas and particulate species. Dry deposition of gas-phase species is highly dependent on the ability of the species to “stick” to the surface, while particulate deposition is highly dependent on the particle size. All species undergo wet and dry deposition to some degree, but the major species of concern are the 2-1

ODMAN,

ET AL.


ones that contain sulfur, nitrogen, and/or hydrogen. Deposition of sulfur and nitrogen species can acidify natural water sources, leach nutrients in soil, damage vegetation directly, and corrode materials (Wark and Warner, 1981). Adverse air pollution effects on visibility, streams, soil, vegetation have been documented (Sisler and Malm, 2000; Heck et al., 1998; Cowling, 1989) in the national parks, forests, and wilderness areas of the Southern Appalachian Mountains in the eastern United States. Although, emission reductions mandated by the 1990 Clean Air Act Amendments should improve the air quality in this region, it is uncertain whether the results will be enough to protect and preserve the ecosystems and natural resources of the Southern Appalachians. The Southern Appalachian Mountains Initiative (SAMI) is currently assessing the impacts of emission controls on air quality in the region. Atmospheric modeling lies at the heart of this assessment.

2.2 SAMI Overview The Southern Appalachian Mountains are famous for their beautiful scenery, natural resources, and numerous recreational activities. Protecting these attributes can be both challenging and expensive. The southern states have been experiencing large population growth and significant economic development that has led to an increased demand for transportation, manufacturing, and energy. As a result, increased pollutant loadings from the urban areas of the Southeast are being transported to the Class I areas (national parks that are greater than 6,000 acres and wilderness areas greater than 5,000 acres) of the Southern Appalachians. The Southern Appalachian Mountains Initiative (SAMI) was formed in 1992 to study these challenging issues and assess the benefits of emission control strategies that could be used to mitigate human-induced air quality related impacts from ozone, PM, and acid deposition on forests, streams, and vistas of the Southern Appalachians, weighing the environmental and socioeconomic implications of recommendations. Ozone affects vegetation primarily through impacts on gas exchange and photosynthesis in leaf cells, reducing growth of foliage, stems, and roots. Aerosols in the atmosphere reduce visibility by the absorption and scattering of light resulting in decreased clarity and color of images and vistas. Wet and dry acid deposition can cause watershed acidification when the buffering capacity is exceeded, can cause leaching of nutrients in the soil, and can directly damage vegetation. Stream acidification reduces the viability of the watershed to support certain types of fish. SAMI is a voluntary partnership that is led by eight southeastern states: Alabama, Georgia, Kentucky, North Carolina, South Carolina, Tennessee, Virginia, and West Virginia. Other participants include the U.S. Environmental Protection Agency, National Park Service, U.S. Forest Service, industries, academia, environmental organizations, and interested citizens. SAMI uses a consensusbased approach to regional strategy development, which provides a forum for stakeholders with diverse interests and viewpoints to work together constructively to establish the scientific basis necessary for regional policy assessments (SAMI, 2001). SAMI is using an “integrated assessment” approach (Figure 2-1) that estimates the environmental effects and selected socioeconomic costs and benefits of SAMI-designed emissions reduction strategies. First, a set of nine multi-day episodes was selected to represent the full spectrum of ozone and PM2.5 concentrations and deposition fluxes between the years 1991 and 1995 (Deuel and Douglas, 1998). Next, the assessment used computer models to follow emissions from their sources through the complex chemical and physical processes that occur in the atmosphere. The resulting pollutant distributions are used to simultaneously estimate environmental and socioeconomic impacts of ozone, fine particles, and acid deposition. Below is a description of each of the assessment areas particularly related to the atmospheric modeling (SAMI, 2001):

2-2

ODMAN,

ET AL.


SAMI Goals Episode Selection Emission Inventories (EMS-95)

Air Quality Modeling (URM-1ATM)

Meteorological Modeling (RAMS)

Pollutant Distribution Environmental Effects

Socioeconomic Impacts

Policy Development Figure 2-1

SAMI integrated assessment framework focusing on the atmospheric modeling aspects.

•

Emission inventories estimate the magnitude and location of pollutants emitted into the atmosphere that contribute to ozone, fine particles and acid deposition in the eastern United States for current and future years to 2040, and for different emission reduction strategies. Direct cost of emissions reduction strategies are also assessed.

•

Atmospheric modeling simulates air quality conditions for nine weeklong episodes during 1991 to 1995. Each episode consists of selected contiguous days, chosen to represent a range of meteorological, emissions, and atmospheric chemistry conditions that contribute to air quality in the SAMI region. These episodes are aggregated to construct a synthetic year. Simulations are conducted using the emissions representing the base year, and alternative emission scenarios for the years 2010 and 2040 to identify air quality responses of SAMI emission reduction strategies.

•

Environmental effects modeling evaluates changes to forests, streams, and visibility, in response to changes in acid deposition, ozone, and fine particle levels. Both the location and sensitivity of these responses are being mapped and modeled, enabling SAMI to describe how air quality and natural resources respond to changes in emissions.

•

Socioeconomic analysis considers the social and economic implications of SAMI emissions reduction strategies. Of the large number of socioeconomic indicators possible, 2-3

ODMAN,

ET AL.


SAMI is focusing on: (1) fishing; (2) hiking/enjoying scenery; (3) stewardship/sense of place; (4) human mortality; (5) regional competition and jobs; and (6) lifestyle changes. Although SAMI’s assessment of socioeconomic indicators is not comprehensive, it provides the best available regional information for the topic area analyzed. Assessment results are used by SAMI to evaluate the need for further emissions reductions, beyond those required by the Clean Air Act and other regulations in place. The assessment tools are also useful to SAMI states in their decisions to implement federal air regulatory requirements, such as the regional haze regulations.

2.3 Air Quality Modeling Objectives This report is about the atmospheric modeling component of SAMI's integrated assessment. The objectives of atmospheric modeling were: •

Update the Urban-to-Regional Multiscale (URM) three-dimensional Eulerian photochemical model (Odman and Russell, 1991; Kumar et al., 1994) to simulate all the chemical and physical processes that govern the formation, transport and removal of gas and particulate pollutants in the atmosphere including aerosol dynamics, aqueous sulfate formation, and wet deposition scavenging processes.

•

Evaluate the ability of the model to simulate ozone, PM, and deposition levels in the Southern Appalachian Mountains by comparing model results to observations during the nine episodes selected to characterize a typical year in the region.

•

Apply the updated model to nine SAMI episodes to simulate air quality in the Southern Appalachians in the years 2010 and 2040 under a number of emission strategies.

•

Update the URM model to allow direct sensitivity analysis for PM and wet deposition.

•

Evaluate the confidence in the direct sensitivity analysis tool for PM and wet deposition by comparing the model results to the results obtained from the traditional brute force method.

•

Apply the updated model equipped with direct sensitivity analysis to the nine SAMI episodes to determine the source/receptor relationship between emission sources and pollutant levels in the Southern Appalachian Mountains.

2.4 Report Structure The rest of this report consists of 11 sections. •

Section 3 describes the air quality modeling system.

•

Section 4 describes the model setup and input data preparation.

•

Sections 5, 6, and 7 discuss the model performance in terms of ozone, PM, and wet deposition, respectively.

•

Section 8 describes how seasonal and annual air quality metrics were calculated and discusses model performance in terms of these metrics.

•

Section 9 describes the future year emission strategies. 2-4

ODMAN,

ET AL.


•

Section 10 discusses the responses to future year emission strategies.

•

Section 11 describes how the sensitivities were modeled.

•

Section 12 discusses the sensitivities to further emission reductions.

•

The report is summarized and conclusions are drawn in Section 13.

The information in this report is being disseminated in terms of publications in leading international peer-reviewed journals. The following papers are at various stages of publication: Boylan J.W., Odman M.T., Wilkinson J.W., Russell A.G., Doty K., Norris W., and McNider R. (2002) Development of a Comprehensive, Multiscale “One Atmosphere” Modeling System: Application to the Southern Appalachian Mountains. Atmospheric Environment, in press. Boylan J.W., Odman M.T., Wilkinson J.W., Russell A.G., Doty K., Norris W., McNider R., Mueller S.F., and Imhoff R. (2002) Performance Evaluation of the URM-1ATM Modeling System in the Southern Appalachian Mountains. In preparation. Boylan J.W., Odman M.T., Wilkinson J.W., Russell A.G., Mueller S.F., Imhoff R.E., and Brewer P.F. (2002) Response of ozone, PM2.5, and acid deposition in the Southern Appalachian Mountains to future year emission scenarios. In preparation. Boylan J.W., Odman M.T., Wilkinson J.W., Russell A.G., Yang Y.-J., Mueller S.F., and Imhoff R.E. (2002) Extension of direct sensitivity analysis to particulate matter dynamics and wet deposition: Application to the Southern Appalachians. In preparation.

2-5

ODMAN,

ET AL.



2-6

ODMAN,

3

ET AL.


AIR QUALITY MODEL

3.1 Introduction Photochemical air quality models are used extensively in both scientific studies and regulatory applications. They integrate our understanding of the complex chemical and physical processes that govern the formation, transport and removal of gas and particulate pollutants in the atmosphere. These mathematical models use land-use, emissions and meteorological inputs, and a description of the atmospheric transformations to predict pollutant concentrations. They are critical in developing optimal emissions control strategies to reduce atmospheric pollutants in urban and rural areas. In the past, air quality model applications focused either on acid deposition or ozone impacts individually. The concept of “one-atmosphere” modeling did not seem practical until recently (Russell and Dennis, 2000). This is, in part, due to the large computational resources required for particulate matter modeling (Zhang et al., 1999). Now that computational power is more readily available, “one-atmosphere” models are beginning to emerge (Byun and Ching, 1999). The Urban-to-Regional Multiscale (URM) model and its monoscale predecessor, the California/Carnegie Institute of Technology (CIT) model, have been widely used for simulating photochemical air pollutant dynamics. The URM model (Kumar et al, 1994; Kumar and Russell, 1996; Odman and Russell, 1991a) is a three-dimensional Eulerian photochemical model that uses a finite element, variable mesh transport scheme (Odman and Russell, 1991b) along with the SAPRC chemical mechanism (Carter, 1990 and 1995) for calculating the gas-phase reaction kinetics. URM uses variable size grids in its horizontal domain to effectively capture the details of pollution dynamics without being computationally intensive. URM has been enhanced to include aerosol dynamics through an equilibrium based aerosol module (Nenes et al., 1998), wet deposition scavenging processes through the Reactive Scavenging Module (Berkowitz et al., 1989), and aqueous sulfate chemistry. The enhanced version of URM, called URM-1ATM, is an integrated “one atmosphere” air quality model. As an integrated multi-pollutant model, the results from URM-1ATM can be used as inputs to assess the effects of ozone, aerosols, and wet deposition on forests, streams, visibility, and human health.

3.2 Urban to Regional Multiscale - One Atmosphere (URM-1ATM) Model 3.2.1

Overview

The URM-1ATM model accounts for transport and chemistry of pollutants by solving the atmospheric diffusion equation:

∂ ci + ∇ ⋅ (u c i ) = ∇ ⋅ (K ∇ ci ) + f i + S i ∂t

(3-1)

where c i is the concentration of the ith pollutant among p species, i.e., i=1,...,p, u describes the velocity field, K is the diffusivity tensor, fi(c 1,...,cp) is the chemical reaction term and Si is the net source term. Elevated emissions and removal processes other than wet and dry deposition are included in Si. The equations and assumptions describing the horizontal and vertical boundary conditions can be found in Kumar and Russell (1996).

3-1

ODMAN,

ET AL.


In solving the atmospheric diffusion equation, URM-1ATM uses operator splitting and decouples various processes. The operator splitting approach advances the solution in time as:

c n +1 = L xy LEmis L Aqueous L Advec LChem L Aero LWet L xy c n

(3-2)

where, Lxy = the horizontal transport (advection and diffusion) operator, LEmis = the elevated point source emissions operator, LAqueous = the S(IV) to S(VI) aqueous-phase chemistry operator, LAdvec = the vertical advection operator, LChem = the coupled vertical diffusion, gas-phase chemistry, area source emissions, and dry deposition operator, LAero = the aerosol condensation/evaporation and growth operator, LWet = the wet deposition and scavenging operator. The time step for each process is determined dynamically based on stability and accuracy considerations. A more detailed treatment of each of these processes will be discussed in the following sections. 3.2.2

Transport

3.2.2.1 Advection URM-1ATM uses the two-dimensional Streamline Upwind Petrov-Galerkin (SUPG) finite element method for solving the horizontal advection equations (Odman and Russell, 1991a and 1991b). The SUPG is a high-order accurate scheme, but is not monotonic or positive definite. To avoid negative concentrations, the SUPG finite element solution is followed by application of a mass conservative isotropic diffusion filter (Odman and Russell, 1993). URM-1ATM treats vertical advection by first-order upwind differencing. To avoid mass conservation problems, the vertical velocities are adjusted by solving the continuity equation using the same numerical techniques (Odman, 2000).

3.2.2.2 Diffusion Vertical and horizontal diffusion are treated using K-theory. The values of the vertical and horizontal diffusion coefficients are obtained from the meteorological inputs. Vertical diffusion is solved using an implicit finite difference scheme. Horizontal diffusion is solved together with horizontal advection using the SUPG finite element method.

3.2.2.3 Convective Transport To account for convective cloud processes and pollutant scavenging (discussed later), the Reactive Scavenging Module (Berkowitz et al., 1989; Scott, 1987) has been incorporated into URM-1ATM. Convective precipitation is simulated with a two-cell (stratiform and convective), steadystate model. This approach allows the definition of several characteristics of convective clouds including large updrafts and vertical transport of low-level air to upper levels. Table 3-1 and Table 3-2 contain the gas and particulate species that are transported, respectively.

3-2

ODMAN,

ET AL.

Table 3-1


Transported Gas-Phase Species

NO - Nitric Oxide NO2 - Nitrogen Dioxide O3 - Ozone

RRP - RO2-RO2-Product RHP - RO2-HO2-Product OLRI - OLD-RI, O Atom Reactions

M2BT - 2- Methyl -2- Butene AAR1 - General Alkane and Aromatics

HONO - Nitrous Acid HNO3 - Nitric Acid HNO4 - Peroxynitric Acid

with Olefins O3SB - O3OL-SB, Represents Conversion of SO2 TO SO3

AAR2 - General Alkane and Aromatics AAR3 - General Alkane and

N2O5 - Nitrogen Pentoxide NO3 - Nitrate Radical HO2 - Hydroperoxy Radical CO - Carbon Monoxide

MEOH - Methanol ETOH - Ethanol GLY - Glyoxal RNO3 - Organic Nitrates

Aromatics OLE1 - General Alkenes OLE2 - General Alkenes NH3 - Ammonia

HCHO - Formaldehyde MEK - Methlyethyl Ketone MGLY - Methyl Glyoxyl PAN - Peroxyacetyl Nitrate

GPAN - Glyoxyl Developed PAN PHEN - Phenol TOLU - Toluene BALD – Benzaldehyde

SO2 - Sulfur Dioxide SO3 - Sulfur Trioxide, Rapidly forms H2SO4

MPAN - Methly Peroxyacetyl Nitrate RO2 - Alkyl Peroxy Radicals RCO3 - Peroxyacyl Radical ETHE - Ethene

PBZN - Peroxy Benzoyl Nitrate AFG1 - Aromatic Ring Fragments 1 AFG2 - Aromatic Ring Fragments 2 CCHO - Acetaldehyde

CRES - Cresols and Other Alkyl Phenols NPHE - Nitrophenols

RCHO - Propionaldehyde and all higher Aldehydes ACET - Acetone

HO2H - Hydrogen Peroxide C - Carbon Atoms LN - Lost Nitrogen Atoms OOH - Lumped Hydroperoxy

PPN - Peroxy Propionyl Nitrate PRPE – Propene M1BT - 2-Methyl-1-Butene ISOP - Isoprene

Species

3.2.3

APNE - α-Pinene UNKN - Unknown PRPA - Propane MARC - Methracloin MVK - Methyl Vinyl Ketone IPRD - Isoprene Reaction Prods. MRC3 - Methly Peroxyacetyl Radical AIR - Air INRT - Inert HCL - Hydrochloric Acid ORGG - Gas Phase Condensable Organics

Chemistry

3.2.3.1 Gas-phase Chemistry The gas-phase reaction kinetics are simulated using the SAPRC chemical mechanism (Carter, 1990), which has been updated with a more accurate treatment of isoprene (Carter, 1995). This mechanism accounts for the atmospheric oxidation of over 100 reactive organic compounds (e.g., alkanes, alkenes, aromatics, alcohols, ethers) as well as a number of reactive oxygenated and organic nitrate products. Gas phase reactions produce secondary organic PM by using lumped experimental

Table 3-2

Transported PM Species

SODX - Sodium HYDX - Hydrogen AMNX - Ammonium

CARX - Elemental Carbon ORGX - Organic compounds CRMX - Magnesium

NITX - Nitrate CHLX - Chloride SULX - Sulfate WATX - Water

CRKX - Potassium CRCX - Calcium PMX - Other PM

X = 1,...,4 represents different size bins according to Stokes diameter: X = 1 for particle diameters < 0.156 µm, X = 2 for particle diameters between 0.156 and 0.625 µm, X = 3 for particle diameters between 0.625 and 2.5 µm, and X = 4 for particle diameters between 2.5 and 10.0 µm.

3-3

ODMAN,

ET AL.

Table 3-1


Steady State Species

OSD - O*1D2, O Singlet D O - Oxygen Atom HO - Hydroxyl Radical

RO2P - RO2-NP, Phenol RO2 Radical RO2R - General RO2 #1 Radical R2O2 - General RO2 #2 Radical

CCO - CCO-O2 Radical C2CO - C2CO-O2 Radical BCO2 - BZ-CO-O2 Radical

COCO - HCOCO-O2 Radical HCO3 - HOCOO Radical BZO - Phenoxy Radical

RO2N - Alkyl Nitrate RO2 Radical RO2X - RO2-XN Radical

BZNO - BZ(NO2)-O

and estimated organic aerosol yields (Pandis et al., 1992). Table 3-1 and Table 3-2 contain the steadystate species, and the constant species.

3.2.3.2 Aqueous Chemistry Aqueous-phase chemistry is based on the reactions implemented in the RSM. It includes the heterogeneous reactions of S(IV) with peroxides and ozone (when the droplet is neutral or basic) to form S(VI). The oxidation of SO2 by O2 catalyzed by trace metals and the hydrolysis of N2O5 in the presence of clouds are not included. Hydrogen ion concentrations in cloud water and rain are calculated from an electroneutrality equation, based on the concentrations of anions such as sulfate and nitrate, and cations such as ammonium and other positive ions associated with crustal material. 3.2.4

Aerosols

The aerosol module is capable of simulating concentrations of all major primary and secondary components of atmospheric particulate matter (PM). There are three groups of PM species that are considered in the aerosol routine: inert species, inorganic equilibrium species, and organic species. The inert species include magnesium, potassium, calcium, elemental carbon, and an “other PM” group which includes all other inert PM species. The inorganic equilibrium species include sulfate, nitrate, ammonium, sodium, chloride, and hydrogen ion. Organic PM is represented by a single lumped condensed specie that contains a sum of numerous condensed organic species resulting from the oxidation of organic gases and directly emitted organic particles. A sectional approach is used for characterization of the continuous PM size distribution by using four size bins: particle diameters < 0.156 µm, 0.156 - 0.625 µm, 0.625 - 2.5 µm, and 2.5 - 10.0 µm. The module simulates mass transfer and particle growth occurring between the gaseous and particulate species during condensation and evaporation (Pandis et al., 1993). Nucleation and coagulation are assumed to be small compared to condensation and are neglected (Wexler et al., 1993). Two size sections would have sufficed for the final SAMI product. However, four size bins were used in case more details were needed for the visibility calculations and to reduce the numerical diffusion in the model.

3.2.4.1 Inorganic Aerosols ISORROPIA (Nenes et al., 1998) is used to model inorganic atmospheric aerosols. ISORROPIA is a computationally efficient and rigorous thermodynamic model that predicts the physical Table 3-2

Constant Species

O2 - Oxygen H2O - Water Vapor CH4 - Methane

CO2 - Carbon Dioxide H2 - Hydrogen M - Third Body

3-4

ODMAN,

ET AL.


state and composition of the sodium-ammonium-chloride-sulfate-nitrate-water aerosol system. The aerosol particles are assumed to be internally mixed, meaning that all particles of the same size have the same composition. The possible species for each phase are contained in Table 3-1. Table 3-1 contains the fifteen equilibrium reactions that are computed in the ISORROPIA mechanism in conjunction with their equilibrium constants (Nenes et al., 1998). The PM mass that is condensed or evaporated in the ISORROPIA routine is partitioned among each size bin based on the original size distribution. Then, the movement of these sections in the size coordinate as a result of particle growth is calculated using the moving section technique (Gelbard, 1990; Kim and Seinfeld, 1990). Species such as ammonium nitrate and ammonium sulfate will preferentially form in the lower three sections, but most of sodium nitrate will be found in the fourth section. Furthermore, it should be noted that the diameters of particles used in the model differ from those used in sampling. Models use the Stokes diameter whereas specifications of sampling equipment refer to the aerodynamic diameter. They are related by the square root of the particle density. Therefore, the first three size bins do not exactly correspond to PM2.5, but are only an approximation to PM2.5.

3.2.4.2 Organic Aerosols The production of condensable organic species from the oxidation of gaseous organic compounds is based on experimental and estimated organic aerosol yields (Pandis et al., 1992). The formation of condensable organic PM is performed in the chemistry module and is assumed to be irreversible. The distribution and growth of condensed organic PM to the four size bins is simulated in the aerosol routine using the same technique as the inorganic PM. Secondary particulate organic compounds will most likely form in the lower three size bins. Also, the possible interactions between inorganic and organic species are not taken into account. Also, an algorithm to simulate particle deposition and gravitational settling for particles of various sizes has been added. Inputs to the aerosol module include temperature, relative humidity, air density, and gas and PM concentrations. Outputs from the module are the updated equilibrium concentrations for the gas-phase and particulate species. 3.2.5

Wet Deposition

The Reactive Scavenging Module (Berkowitz et al., 1989) uses synoptic scale temperature and precipitation rates to simulate a field of representative clouds that are defined by scavenging rates, water profiles, and wind fields. The module simulates the time dependent chemical kinetic interaction of these clouds with the gas and particulate species and the vertical convective transport within a column of air. Scavenging processes within the module include gas, aerosol, and microphysical scavenging. Gas-phase species scavenging by cloud water is modeled via an equilibrium process that is based on species solubilities. Scavenging of gas-phase species by rain water is modeled for most species via mass transfer. Scavenging by snow is limited to nitric acid. Aerosol scavenging is treated by nucleation and by inertial impaction processes. The scavenged species are sulfur dioxide, particulate sulfate, ozone, nitric acid, particulate nitrate, hydrogen peroxide, ammonia, particulate ammonium, and soluble 2+ 2+ crustals (Mg and Ca ). Other gas and particulate species are passed into the RSM module where vertical convective transport is simulated. Output from the module includes updated concentration

Table 3-1

Possible gas and PM species in ISORROPIA

Phase

Species

Gas Liquid Solid

NH3, HNO3, HCl, H2O NH4+, Na+, H+, Cl-, NO3-, SO42-, HSO4-, OH-, H2O (NH4)2SO4, NH4HSO4, (NH4)3H(SO4)2, NH4NO3, NH4Cl, NaCl, NaNO3, NaHSO4, Na2SO4, H2SO4

3-5

ODMAN,

ET AL.

Table 3-1


Equilibrium relations and equilibrium constant expressions, where γ is the activity coefficient (Nenes et al., 1998).

Reaction

Equilibrium Constant (K)

[ H + ][ SO42 − ] γ H + γ SO 42− [ HSO4− ] γ HSO−

K1 HSO4−( aq ) ←→ H (+aq ) + SO42(−aq )

4

NH 3 ( g ) ←→  NH3 ( aq ) K 21

[ NH3 ( aq ) ] [ PNH 3 ]

γ NH 3

[ NH 4+ ][ OH − ] γ NH 4+ γ OH −

K 22 NH 3 (aq ) + H 2 O( aq ) ← → NH 4+( aq ) + OH (−aq )

[ NH3 ( aq ) ]aw K4 HNO3 ( g ) ←→ H (+aq ) + NO3−(aq )

[ H + ][ NO3− ] [ PHNO3 ]

K3 HCl( g ) ←→ H (+aq ) + Cl(−aq )

[ H + ][ Cl − ] [ PHCl ]

Kw H 2 O( aq ) ← → H (+aq ) + OH (−aq )

γ H + γ NO − 3

γ H +γ Cl −

[ H + ][ OH − ] [ aw ]

γ NH 3

γ H + γ OH −

K5 Na2 SO4 ( s ) ←→ 2 Na(+aq ) + SO42(−aq )

2 [ Na + ]2 [ SO42 − ]γ Na +γ SO 2 −

K7 ( NH4 ) 2 SO4 ( s ) ←→ 2 NH4+(aq ) + SO42(−aq )

2 [ NH4+ ]2 [ SO42 − ]γ NH +γ SO 24 − 4

K6 NH 4 Cl( s ) ←→ NH3 ( g ) + HCl( g )

PNH 3 PHCl

K9 NaNO3 ( s ) ←→ Na(+aq ) + NO3−( aq )

[ Na+ ][ NO3− ]γ Na + γ NO 3−

K8 NaCl( s ) ←→ Na(+aq ) + Cl(−aq )

[ Na+ ][Cl − ]γ Na + γ Cl −

K11 NaHSO4 ( s ) ←→  Na(+aq ) + HSO4−( aq )

[ Na + ][ HSO4− ]γ Na +γ HSO4−

K 10 NH 4 NO3 ( s ) ←→  NH 3 ( g ) + HNO3 ( g )

PNH 3 PHNO3

K12 NH 4 HSO4 ( s ) ←→  NH4+(aq ) + HSO4−( aq )

[ NH4+ ][ HSO4− ]γ NH + γ HSO −

( NH4 )3 H ( SO4 ) 2 ( s ) ←→  3 NH K13

+ 4 (aq )

4

4

+ HSO

− 4 ( aq )

+ SO

2− 4 ( aq )

+ 3 4

2− 4

4

− 4

[ NH ] [SO ][ HSO ]γ

3 NH 4+

γ SO γ HSO 2− 4

− 4

Nomenclature: aw is water’s activity, γ i is the activity coefficient of species i, [i] is the concentration of species i, Km is the equilibrium constant for reaction m, Pi is the partial pressure of species i.

profiles for all the species affected by scavenging and convective cloud transport, in addition to wet 2+ 2+ 2+ + deposition mass fluxes for SO2, SO4 , NO3 , H2O2, NH4 , Mg and Ca and H .

3-6

ODMAN,

ET AL.

Table 3-1


Aerosol Particle Deposition Velocities

Bin

Geometric Mean Diameter (µm)

Deposition Resistance (sec cm -1 )

1 2 3 4

0.078 0.312 1.25 5.0

78.3 163.2 266.4 1.34

3.2.6

Dry Deposition

For dry deposition, URM-1ATM uses the three-resistance approach based on the formulation i of Wesely (1989). Total resistance to deposition of species i, rt , is composed of three resistances:

rt i = ra + rb + rsi

(3-3)

where ra is the resistance to deposition due to turbulent transport through the atmosphere, rb is the i resistance due to diffusion through a laminar sub-layer, and r s is the resistance due to chemical i interaction between the surface and the pollutant of interest. The deposition velocity, v g, for a species i then becomes:

v ig =

1 rt i

(3-4)

A detailed description of the calculation of the various resistances is given in Russell et al. (1990) and Harley et al. (1992). For the aerosol particles, size-dependent deposition resistances are calculated using experimental data from the National Center for Atmospheric Research (1982). The deposition resistances located in Table 3-1 are estimated using the log-mean diameter for each size bin and are independent of the species composition.

3-7

ODMAN,

ET AL.



3-8

ODMAN,

4

ET AL.


MODEL SETUP AND INPUT DATA PREPARATION

4.1 Modeling Domain and Grid URM-1ATM differs from other air quality models in the way it provides a multiscale modeling capability. While other models resort to grid nesting techniques, URM-1ATM provides a single grid with variable resolution. The finest grids are placed over the target area of interest and the adjacent areas that are expected to most directly influence the air quality in that region. Coarser cells are placed in areas that are not expected to significantly contribute to the air quality in the region of interest with the coarsest cells typically near the boundary of the domain. The SAMI modeling domain covers the eastern half of the United States and the southeastern part of Canada as shown in Figure 4-1. This domain is discretized by using a multiscale grid with variable resolution. The resolution ranges from 12 km (finest) to 192 km (coarsest) with intermediate values of 24, 48 and 96 km. By using finite element refinements, the major sources and Class I receptors of the SAMI region were covered with fine resolution; at the same time, the domain boundaries were placed far enough from the region of interest. This gridding technique is more efficient than conventional nesting techniques that require rectangular domains. The surface elevation data that was mapped onto the grid is shown in Figure 4-2. The modeling domain extends from the surface to a height of 12,867 meters, consistent with the vertical extent of the meteorological modeling domain, and is divided into seven layers (Table 4-1). The use of finer resolution near the surface, as compared to

Figure 4-1

The SAMI modeling domain and grid. 4-1

ODMAN,

ET AL.

Figure 4-2


Surface elevation (above sea level) mapped onto the grid.

the coarser resolution aloft, allows the steeper concentration gradients in the boundary layer and the evolution of the mixing depths during the day to be captured with greater detail.

Table 4-1

a

Vertical structure of the URM-1ATM model.

Layer

Coverage (m)a

Thickness (m)

1 2 3 4 5 6 7

0-19 19-62 62-494 494-1493 1493-3272 3272-6860 6860-12867

19 43 432 999 1779 3588 6007

Above ground level

4-2

ODMAN,

ET AL.


4.2 Modeled Episodes The nine episodes listed in Table 4-1 were chosen for detailed modeling. A general description of the severity of the ozone, PM, and acid deposition levels is also included for each episode that will be used to calculate the corresponding seasonal and annual air quality metrics. The metrics are seasonal cumulative ozone W126 (Lefohn, 1987), annual average PM concentrations, and annual average deposition fluxes, respectively. A brief summary of the important weather systems for each episode is presented below. •

February 8-13, 1994 - A major incursion of Arctic air occurred during the period of February 9-11 for the eastern half of the United States and was accompanied by a major ice storm from Texas to Ohio. The heaviest precipitation was in a band from northern Mississippi to West Virginia.

•

March 23-31, 1993 - This period was an active one across the Southeast with parts of Tennessee, Alabama, Georgia, North Carolina, and South Carolina having total precipitation over 70 mm.

•

April 26 - May 3, 1995 - A progressive weather pattern allowed several systems to move rapidly across the eastern United States. Precipitation amounts were generally moderate but covered a large area of the Midwest and South.

•

May 11-17, 1993 - Several storm and frontal systems affected the eastern United States during this episode with the heaviest precipitation falling across an area from Missouri and Arkansas southeastward across much of the Southeast.

•

May 24-29, 1995 - An active southwest to northeast storm track brought several storm systems to the eastern United States with the strongest being during the period of May 27-29. Precipitation fell over a large part of the eastern U.S. but with the heaviest amounts over an area extending from Missouri and Iowa eastward to the Ohio River valley and then southwestward to the Gulf coast.

•

June 24-29, 1992 - The eastern United States was dominated by surface high pressure at the beginning and end of this episode. Otherwise the main systems included a slow moving frontal system that moved from the northern Plains southeastward and then stalled over the Ohio River Valley, and a persistent stationary front over the Gulf coast. Total precipitation was generally light except for locally heavier amounts across parts of the coastal areas of the Southeast, the southern Appalachians, and the Tennessee Valley.

Table 4-1

Modeled episodes and pollution levels observed at Great Smoky Mountains and Shenandoah National Parks.

Episode February 8-13, 1994

Ozone No Observations

PM2.5 Low

Acid Deposition Moderate

March 23-31, 1993

No Observations

Low

Moderate

April 26-May 3,1995

Low

Low to Moderate

Low to Moderate

May 11-17,1993

Low to Moderate

Moderate

Moderate to High

May 24-29, 1995

Moderate

Moderate

Low to Moderate

June 24-29, 1992

Moderate

Moderate to High

Low to Moderate

July 23-31, 1991

Low to Moderate

High

High

July 11-19, 1995

High

High

Low

August 3-11, 1993

Low

Moderate

Low to Moderate

4-3

ODMAN,

ET AL.


•

July 23-31, 1991 - Several frontal systems moved across the Midwest and then stalled over the Southeast. This resulted in an active and very wet pattern for areas along and east of the Appalachians with locally heavy rains and flooding.

•

July 11-19, 1995 - For much of the United States east of the Mississippi River and north of the Gulf coast the weather was dominated by high pressure at the surface and aloft with light winds, little o precipitation, and daily maximum temperatures of 30 C and above. After July 17-18, a frontal passage for much of the same area brought lower temperatures and humidity. For the SAMI region, only light and scattered precipitation occurred during the episode.

•

August 3-11, 1993 - Several weak frontal systems moved across the eastern United States with several days of a stationary front across the Southeast. Along and south of this front precipitation was locally heavy.

In order to simulate the various episodes, URM-1ATM requires estimates of the meteorology, emissions, and air quality fields. For purposes of this study, meteorological predictions are taken from the Regional Atmospheric Modeling System (RAMS) (Pielke et al., 1992), and emissions estimates are prepared using the Emissions Modeling System (EMS-95) (Wilkinson et al., 1994). The observed air quality fields are used for setting the initial and boundary conditions. Other inputs include landuse, dry deposition resistances, and surface roughness. The following sections discuss these issues in more detail.

4.3 Meteorological Inputs A modified version of the Regional Atmospheric Modeling System (Pielke et al., 1992) version 3a has been used to produce meteorological input fields for URM-1ATM. It was run with a system of three nested grids in non-hydrostatic mode with cloud and rainwater microphysics activated. The National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) reanalysis data (Kalnay et al., 1996) was chosen as the main data source for the meteorological simulations. Figure 4-1(a) shows the nested horizontal grid structure used by RAMS for the July 1995 episode. The coarse, intermediate, and fine grids had resolutions of 48-, 24-, and 12-km, respectively. The RAMS nested grid arrangement used for the other eight episodes is shown in Figure 4-1(b). The coarse, intermediate, and fine grids for this grid setup had resolutions of 96-, 24-, and 12-km, respectively. Before these fields can be used, they must be converted from the RAMS nested grid structure to the URM-1ATM multiscale grid structure (Figure 4-1). The grid nests in RAMS are arranged such that there is always meteorological data available at the resolution of the URM-1ATM grid. The RAMS and URM-1ATM vertical grids are also different. Though the two grids do match in the vertical extent (i.e. 12,867 meters), the RAMS vertical structure has thirty-one layers, but the URM-1ATM vertical structure has seven layers. Hence, it is necessary to aggregate the higher resolution RAMS layers into the more coarse URM-1ATM layers. A distance-weighting scheme is used to interpolate the scalar fields in the vertical. The meteorological variables that are used by URM-1ATM are shown in Table 4-1. Specific details of the meteorological modeling can be found in Doty et al., 2001. Meteorological performance statistics were calculated based on all available surface National Weather Service observations within the RAMS 12-km domain (typically on the order of 100 stations) for each episode. A summary of National Weather Service mean observations, standard deviations, o biases, and root mean square errors (RMSE) for 2-m temperatures ( C), 2-m water vapor mixing ratios -1 -1 (g kg ), wind directions at 10-m (degrees clockwise from North), and 10-m wind speeds (m s ) within the 12-km domain for each episode is contained in Table 4-3 and Table 4-2. Specific details and assumptions for these calculations can be found in Doty et al., 2001.

4-4

ODMAN,

ET AL.


(a)

Figure 4-1

(b)

Horizontal grid structures used for the SAMI meteorological modeling. (a) Grids used for the July 1995 episode. (b) Grids used for the other eight episodes. Only every other grid point is plotted.

The coldest and driest (with regard to the 2-m absolute humidity) episode was February 1994 while the warmest and most moist was July 1995. The February 1994 and March 1993 cases were dominated by northeasterly winds and the April 1995 episode by northwesterly winds. All other episodes had predominantly southwesterly winds. As expected, the lightest average wind speeds were for the summer episodes which also had a tendency to have the largest standard deviation of wind direction. The three wettest cases with respect to the mean total precipitation were the March 1993, July 1991, and February 1994 episodes with values of 57, 52, and 46 mm, respectively. The four driest cases with respect to the same statistic were the June 1992, July 1995, May 1995, and May 1993 episodes with values of 16, 18, 18, and 19 mm, respectively. The four cases with largest standard deviation of 6-h precipitation amounts were the July 1991, June 1992, July 1995, and August 1993 Table 4-1

a

RAMS Meteorological parameters used by URM-1ATM

Dimensions

Variable Name

2D 2D 2D 3D 3D 2D 3D 2D 2D 2D 3D 2D 3D

CVAR CVFR CVTP DENS KMIX MIXH MIXR PRCP SOLR STTP TEMP ULTV WIND

Description convective cloud cover area fraction (%) convective cloud precipitation fraction (%) convective cloud top height (m) air density (kg m-3) turbulent momentum diffusivity (m2 s -1) mixing depths (m) absolute humidity – water vapor mixing ratio (g kg -1) precipitation (mm) total incoming solar radiation (W m-2) stratiform cloud top height (m) air temperature (K) total incoming ultraviolet radiation (W m-2) u, v, and wa components of wind (m s -1)

w adjusted as described in Section 3.2.2.1.

4-5

ODMAN,

ET AL.

Table 4-3


Summary of National Weather Service surface observations, RAMS standard deviations, and performance statistics for temperatures and mixing ratios. Temperature (oC)

Episode February 94 March 93 April 95 May 93 May 95 June 92 July 91 July 95 August 93

Water Vapor Mixing Ratio (g kg -1 )

Mean

St. Dev.

Bias

RMSE

Mean

St. Dev.

Bias

RMSE

3.1 12.1 13.8 20.1 21.7 21.8 25.6 26.8 23.2

8.1 5.1 5.7 5.3 5.0 5.0 4.6 4.7 4.3

+0.30 -1.28 -0.92 -1.43 -0.98 -1.22 -0.71 +0.66 -0.44

2.82 3.01 2.37 2.69 2.34 2.36 2.25 2.45 2.12

4.3 7.0 6.5 10.0 11.5 11.6 15.5 15.7 13.5

2.8 2.0 2.2 2.4 2.8 3.3 3.3 3.0 2.7

-0.00 -0.16 -0.77 -0.30 -0.29 -0.32 -0.37 -1.45 -1.76

0.59 0.67 1.96 1.06 1.03 1.07 0.90 2.56 4.12

episodes with values of 11.5, 10.2, 9.2, and 8.9 mm, respectively. The main mode of precipitation in these summer episodes is convection. Simulation of clouds and precipitation remain one of the more difficult issues with numerical weather prediction or simulation and the SAMI episodes were no exception. One measure of model performance with regard to precipitation is the so-called equitable threat score (ET), which was originally described by Schaefer (1990) as the Gilbert Skill Score. Later it became known as the equitable threat score when applied to precipitation (Rogers et al., 1995; McBride and Ebert, 2000). It has a range from -1/3 to +1, where +1 would indicate a perfect simulation. It is defined by Equation 4-1, where the symbols are defined in Table 4-4. The variables W, X, Y, and Z can have either a value of 1 for TRUE or a value of 0 for FALSE. Equitable threat scores are usually calculated as a function of the forecast time period, precipitation amount, and the verification period.

ET =

XW - YZ (Y + Z )(X + Y + Z + W ) + (XW − YZ )

(4-1)

Equitable threat scores for several operational numerical weather prediction models at the National Centers for Environmental Prediction (NCEP) can be found at http://www.emc.ncep.noaa.gov/mmb/pscores/. If one examines the previous web page for the

Table 4-2

Summary of National Weather Service surface observations, RAMS standard deviations, and performance statistics for wind speeds and wind direction. Wind Speeds (m s -1 )

Episode February 94 March 93 April 95 May 93 May 95 June 92 July 91 July 95 August 93

Wind Direction (degrees)

Mean

St. Dev.

Bias

RMSE

Mean

St. Dev.

RMSE

3.7 3.2 3.6 2.9 3.4 2.7 2.5 2.5 2.5

2.2 1.9 2.1 1.9 2.0 1.8 1.7 1.7 1.8

+1.33 +0.88 +1.33 +0.38 +0.18 +0.66 +0.94 +0.49 +1.16

2.75 2.15 2.35 1.73 1.75 1.85 2.26 1.51 2.08

21 68 321 210 198 216 227 217 227

96 92 92 83 93 95 101 91 101

73.0 70.0 65.0 67.0 63.0 74.0 77.0 83.0 78.0

4-6

ODMAN,

ET AL.

Table 4-4


Four-cell contingency table used to calculated the equitable threat score (Schaefer, 1990). Forecasted Event Observed Event

YES

NO

YES

X

Y

NO

Z

W

performance of NCEP’s mesoscale prediction model, Eta, for a verification period of 24 h at the end of 24-, 36-, and 48-h forecasts for the Southern Midwest (consists of Arkansas, eastern Texas and Oklahoma, most of Tennessee, Louisiana, Mississippi, Alabama, and the panhandle of Florida), one observes maximum ET scores of about 0.6 for January 2001 and minimum ET scores of about 0.28 for July 2001. The same web page shows for a verification period of 3 h at the end of 48-h forecasts for the same model maximum ET scores of only about 0.15 regardless of the month. This admittedly small sample of results illustrates the behavior of numerical weather prediction models in general of having reduced skill in precipitation performance for the warm season (when precipitation is predominately convective), for longer forecast periods, and shorter verification periods. Calculation of RAMS simulated precipitation ET scores for 6 of the episodes every 6 hours (not shown) showed the best performance for the February 1994 and March 1993 episodes with maximum scores of 0.20-0.30 while all the other episodes which were in the warm season had maximum scores of 0.15 or less. These scores are consistent with the behavior of the operational models mentioned earlier. Model performance for the two winter episodes was reduced not only because of cloud and precipitation regimes, but also possibly because of a need for more vertical resolution with such a strong jet stream aloft. Additional issues for the winter cases involved a greater need for the ice microphysics to be activated (computer resources and schedule restraints prevented this) and demanding surface energy requirements with snow and ice cover which were not adequately treated by the version of RAMS used. Improved treatment of wind direction and speed in light wind situations (especially in complex terrain locations) would likely have resulted with a nested grid of 4-km or less but once again computer resources and schedule restraints prevented this option from being feasible. Even with these issues, RAMS was able to correctly simulate the observed synoptic scale flows. The 2o o m temperature bias and RMSE were generally below 1.0 C and 2.5 C, respectively, and both 2-m -1 mixing ratio bias and RMSE were generally below 1 g kg . A consistent positive bias of the wind speed -1 -1 at 10 m of 1 m s or less was observed and wind speed RMSE were typically 2.0 m s or less. Wind o direction RMSE values were typically 70 . The overall meteorological performance was considered to be adequate for the SAMI modeling requirements and consistent with the current capabilities of meteorological models in general.

4.4 Emission Inputs The EMS-95 (Wilkinson et al., 1994) is used to generate speciated day-specific, hour-by-hour gridded emission inputs to be used by URM-1ATM. EMS-95 separates the emissions into two categories: elevated point and ground-level source emissions. Ground-level sources include low-level point sources, mobile sources, anthropogenic area sources, non-road mobile sources, and biogenic sources. Point source and area source emissions estimates used in EMS-95 were based on data developed by the Pechan/Avanti Group (2001), as were on-road mobile source data (e.g. vehicle miles traveled by state, county, and roadway type; vehicle mix by state, county, and roadway type; speeds by vehicle type and roadway type) and were used to estimate on-road mobile source emissions using the EMS-95 Motor Vehicle Emissions Model (MoVEM). MoVEM uses MOBILE5b (USEPA, 1994) to compute vehicle-dependent emissions factors of CO, NOx, and total organic gases (TOG). Biogenic emissions were estimated using U.S. EPA’s Biogenic Emissions Inventory System, version 2 (BEIS2 – Pierce and Geron, 1996; Pierce, 1996; Pierce et al., 1990). The point source emissions estimates were 4-7

ODMAN,

ET AL.

Table 4-1


Emission species generated by EMS-95

General Species

Specific Species

Non-VOC Gas-Phase

CO, NO, NO2, NH3, SO2 ETHE, OLE1, OLE2, AAR1, AAR2, AAR3, TOLU, HCHO, CCHO, MEK, ACET, MEOH, ETOH, ISOP, APNE CARX, CRCX, CRKX, CRMX, NITX, ORGX, PMX, SODX, SULX

VOC Gas-Phase Particulate-Phase2 1

For an explanation of abbreviations refer to Table 3-1

2

X (1-4) represents the aerosol size bin

enriched with day specific emissions data obtained from major utility companies in the modeling region. Meteorological model results were used to estimate the temperature and radiation dependent biogenic emissions and the temperature dependent on-road mobile source emissions. Primary emissions species that are generated by EMS-95 for use in the SAPRC-based chemical mechanism utilized by URM-1ATM are listed in Table 4-1. Domain-wide average day emission estimates for the July 1995 episode are listed in Table 4-2. Emissions inventories were obtained from Pechan/Avanti Group via ftp for each season and scenario. After processing by EMS-95 (i.e., spatial and temporal allocation and development of emissions for mobile and biogenic sources), QA/QC procedures were applied to the emissions. One step included performing emissions totals for comparison with the unprocessed emissions and other emissions estimates. This often led to considerable regeneration and/or correction of emissions inventories. The output of this step were two emissions files, one for ground level emissions, the other for upper level emissions usually associated with larger point sources whose emissions are effectively emitted above the lowest layer in the model. 4.4.1

Emission Summary Tables

The total emissions were calculated both for the entire modeling domain and for the 8 SAMI states. Emissions in different categories were tabulated. Table 4-1 shows the 8 SAMI state total for an average day during the July-95 episode. Similar tables are available in Appendix A for other episodes both for the entire modeling domain and for the 8 SAMI states. 4.4.2

Day Specific Emissions After gridding of emissions, spatial plots (maps) were prepared for various categories over a

Table 4-2

Daily Average Emissions for July 1995 Episode

Species

Domain-wide Emissions (tons day -1 )

Biogenic Volatile Organic Compounds (VOCs) Anthropogenic Volatile Organic Compounds (VOCs) Sulfur Dioxide (SO2) Nitrogen Oxides (NOx) Ammonia (NH3) Carbon Monoxide (CO) Particulate Sulfate (SO42-) Particulate Nitrate (NO3-) Particulate Sodium (Na +) Particulate Carbon (EC & OC) Particulate Crustals (CRUS)

176,201 47,679 53,697 59,597 8,662 176,310 971 36 440 5,829 1,547

4-8

ODMAN,

ET AL.

Table 4-1


Daily average total emissions from the 8 SAMI states for the July-95 episode (in tons per day).

Area Biogenic Electric Generation Off-road mobile On-road mobile Point Daily total

VOCs

NOX

CO

SO2

PM10

PM2.5

NH3

4,176 54,069 119 1,660 2,734 2,190 64,947

660 520 6,210 1,646 3,222 1,733 13,991

3,875 0 181 10,324 22,602 2,836 39,818

972 0 11,533 285 153 3,837 16,780

10,849 0 607 438 258 814 12,966

2,406 0 147 204 113 365 3,235

1,178 0 0 1 132 65 1,376

portion of the domain covering the SAMI region. Figure 4-1 shows the elevated SO2 emissions for July 15, 1995. Figure 4-2 shows the ground level SO2 emissions for the same day. Figure 4-3 and Figure 4-4 show the elevated and ground-level NOx emissions, respectively. Finally, Figure 4-5 shows the daily total ground-level NH3 emissions for July 15, 1995. Maps for other modeled days can be found in Appendix A.

Figure 4-1

Gridded daily total emissions of elevated SO2 over the SAMI region for July 15, 1995. 4-9

ODMAN,

ET AL.


Figure 4-2

Gridded daily total emissions of ground level SO2 over the SAMI region for July 15, 1995.

Figure 4-3

Gridded daily total emissions of elevated NOx over the SAMI region for July 15, 1995.

4-10

ODMAN,

ET AL.


Figure 4-4

Gridded daily total emissions of ground-level NOx over the SAMI region for July 15, 1995.

Figure 4-5

Gridded daily total emissions of ground level NH 3 over the SAMI region for July 15, 1995.

4-11

ODMAN,

ET AL.


4.5 Initial and Boundary Conditions URM-1ATM requires initial (IC) and boundary conditions (BC) for both gaseous and particulate species. Initial conditions can have a major impact on the modeled concentrations at the beginning of the simulation, but the impact usually diminishes as the simulation proceeds. Many of the gaseous species of interest have relatively short lifetimes: they quickly undergo chemical transformation and deposition. The impact of ICs diminishes more quickly in comparison to particulate species. A two-day ramp-up period is used before each episode that is modeled and is generally sufficient to dampen the effect of initial conditions on gaseous species concentrations. However, fine PM has very low deposition velocities and longer lifetimes and can persist for longer periods. Therefore, the concentrations of particulate species have a stronger dependence on initial conditions and extra care must be used in setting those conditions. The same principles apply to the boundary conditions. The lifetimes of gaseous species are usually not long enough for transport from the domain boundaries into the region of interest. However, the particulate species can be transported considerable distances (e.g., thousands of kilometers) and they can impact the concentration of PM in the region of interest. The IC/BCs for the gaseous species were derived using data from the Aerometric Information Retrieval System (AIRS) (USEPA, 2001) and the North American Research Study on Tropospheric Ozone for the Northeast (NARSTO-NE) (Mueller, 1998) data archives as well as data from specialized studies and smaller networks. The gaseous species for which IC/BCs are developed include CO, SO2, NOx, VOCs and ozone. Initial conditions are based on observations that correspond to the time and date that the model simulation begins. Boundary conditions are based on observations that vary spatially and temporally over the duration of the modeling episode. Because the monitoring network does not correspond to the modeling grid, it is necessary to interpolate the observed values to the computational nodes on the modeling grid. For all the VOC species and CO, Dirichlet tessellation (Preparata and Shamos, 1985; Green and Sibson, 1978) is used to determine concentrations at each node on the modeling grid. The IC/BCs for SO2 and NOx are treated differently since the AIRS SO2 and NOx monitors are often located in areas with high concentrations (e.g. downwind from power plant plumes) and interpolation using Dirichlet tessellation overestimates the initial and boundary conditions over rural areas. Therefore, the following interpolation scheme (Equations 4-2 and 4-3) is used to derive SO2 and NOx IC/BCs such that the suspected high bias is minimized:

 n ln (c i )  ∑ d 2 c  i =1 i, j i  c j = exp  n for j = 1, g  1  ∑ d 2 c   i =1 i , j i  d i2, j = (x j − x i ) + ( y j − y i ) 2

2

(4-2)

(4-3)

where: c is concentration, i is the observation index, n is the number of observations, j is the node index, g is the number of nodes, x is the east-west coordinate of the observation or node location, and y is the north-south coordinate of the observation or node location. This method was tested and found to produce similar levels of SO2 and NOx in both the urban and rural areas as compared to levels that were produced by running the model for multiple days.

4-12

ODMAN,

ET AL.


The interpolation scheme that is used for ozone is an inverse distance squared weighting of the observations (Equation 7):

 n ci ∑ d 2 i =1 i,j cj =  n  1 ∑ 2  i =1 d i , j

   for j = 1, g   

(4-4)

Like the SO2 interpolation scheme, it too minimizes and localizes the impact of locally high ozone observations but not to the extent that the SO2 interpolation does. That is, locally high ozone concentrations are more likely to influence IC/BCs at points further away from the measurements. A summary of the procedures used for deriving the initial and boundary conditions for each of the gaseous species in the SAPRC mechanism (Carter, 1990) is shown in Table 4-1. The IC/BC values at the top of the domain are set to free troposphere values (Seinfeld and Pandis, 1998). Linear interpolation from the ground level values to the top of the domain is used to derive the IC/BC concentrations for layers in between. The IC/BCs for the other gas phase species in the SAPRC chemical mechanism (e.g., NH3, PAN, HNO3) are set equal to zero and are allowed to evolve during the simulation. The initial surface layer SO2 concentrations used for the July 11-19, 1995 episode are shown in Figure 4-1. Recall that there are two ramp-up days before each episode hence the actual simulation of this episode starts on July 9. These initial conditions were created using the inverse-concentrationsquared-distance weighting of the natural logarithm of observed SO2 concentrations (Equations 4-2 and 4-3).

Table 4-1

Number

Gas phase species in SAPRC mechanism and the methodology used for setting their surface layer initial and boundary conditions. Gas Phase Species

Method used for surface layer IC/BC derivation

(1)

Nitric Oxide (NO)

inverse-concentration-squared-distance weighting of the natural logarithm of observed concentrations (Equations 4-2 and 4-3)

(2)

Nitrogen Dioxide (NO2)

inverse-concentration-squared-distance weighting of the natural logarithm of observed concentrations (Equations 4-2 and 4-3)

(3)

Ozone (O3)

inverse- squared-distance weighting of the observed concentrations (Equations 4-4)

(4) (5) (6) (7)

Carbon Monoxide (CO) Formaldehyde (HCHO) Acetaldehyde (CCHO) Methyl Ethyl Ketone (MEK)

Dirichlet tessellation of observed concentrations Dirichlet tessellation of observed concentrations Dirichlet tessellation of observed concentrations Dirichlet tessellation of observed concentrations

(8) (9)

Ethene (ETHE) Sulfur Dioxide (SO2)

Dirichlet tessellation of observed concentrations inverse-concentration-squared-distance weighting of the natural logarithm of observed concentrations

(10) (11) (12-13) (14-16)

Isoprene (ISOP) Hydrochloric Acid (HCL) General Alkenes (OLE1, OLE2) General Alkanes and Aromatics (AAR1, AAR2, AAR3)

Dirichlet tessellation of observed concentrations Dirichlet tessellation of observed concentrations Dirichlet tessellation of observed concentrations Dirichlet tessellation of observed concentrations

17-60

Other 44 species

Set equal to zero

4-13

ODMAN,

ET AL.

Figure 4-1


Initial surface layer SO2 concentrations used for the July 9-19, 1995 episode.

The IC/BCs for the particulate species were derived from the Interagency Monitoring of Protected Visual Environments (IMPROVE) measurements. IMPROVE particulate matter measurements (NPS, 2000) are taken twice a week (Wednesday and Saturday) and are reported as a 24-hour average concentration for each species. Table 4-1 shows the URM-1ATM model species and the corresponding IMPROVE species along with any required conversion factors that were used to set the IC/BCs. Initial conditions are determined by examining the fine PM concentrations (PM2.5) for the day that most closely matches with the start of each episode. In the event that the episode starts on a day that is equally close to two IMPROVE measurement days, an average of the two closest days is used. Unlike the spatially varying conditions for gas-phase species, the initial conditions for each particulate species are set to a uniform concentration across the domain. The concentration is equal to the average of all valid IMPROVE observations in the modeling domain. Interpolation is not performed due to the limited amount of data. Similarly, boundary conditions (BC) were uniform along the boundaries, but not all boundaries shared the same particulate concentrations. A distinction was made between land (North, West, and Southwest) and marine (East and Southeast) boundaries. To determine the boundary conditions for the north, west, and southwest boundaries, IMPROVE data for stations west of the modeling domain were used. Observations were averaged over the duration of the episode and the BCs were set to the average value. The marine BCs were assumed to be 10% of the land BCs, except for the Na, Cl, and H, which were adjusted to reflect higher sea-salt concentrations. Based on data from the continental and marine vertical distribution of sulfate (Warneck, 1988), it was determined that the mixing ratios at 5 km were about 10% of those at the continental surface level. Therefore, linear interpolation was used to derive concentrations at layers in between. The species on the marine boundary were assumed to have a constant mixing ratio up to 5 km, except for Na, Cl, and H, which varied linearly with height. Above 5 km, it was assumed that the mixing ratio for all particulate species was constant. Next, the IC/BC concentrations are distributed to each of the four PM bins and a charge balance is performed to ensure that the electoneutrality equation (ENE) is satisfied. The size 4-14

ODMAN,

ET AL.

Table 4-1


Particulate Species Modeled by URM-1ATM and IMPROVE Observations

Species

URM-1ATM Model Species

IMPROVE Observations

Fine Sulfate Fine Nitrate Fine Ammonium Fine Sodium Fine Chloride Fine Hydrogen Ion Fine Elemental Carbon Fine Organic Carbon Fine Calcium Fine Magnesium Fine Potassium Fine PM “other” PM2.5

SUL1+SUL2+SUL3 NIT1+NIT2+NIT3 AMN1+ AMN2+ AMN3 SOD1+ SOD2+SOD3 CHL1+CHL2+CHL3 HYD1+HYD2+HYD3 CAR1+CAR2+CAR3 ORG1+ORG2+ORG3 CRC1+CRC2+CRC3 CRM1+CRM2+CRM3 CRK1+CRK2+CRK3 PM1+PM2+PM3 SUL1+SUL2+SUL3+NIT1+NIT2+NIT3+ORG1+ORG2+ ORG3+CRK1+CRK2+CRK3+CRC1+CRC2+CRC3+ CRM1+CRM2+CRM3+PM1+PM2+PM3+SOD1+SOD2+ SOD3+CAR1+CAR2+CAR3+CHL1+CHL2+CHL3+ AMN1+AMN2+AMN3+HYD1+HYD2+HYD3 SUL1+SUL2+SUL3+SUL4+NIT1+NIT2+NIT3+NIT4+ ORG1+ORG2+ORG3+ORG4+CRK1+CRK2+CRK3+ CRK4+CRC1+CRC2+CRC3+CRC4+CRM1+CRM2+ CRM3+CRM4+PM1+PM2+PM3+PM4+SOD1+SOD2+ SOD3+SOD4+CAR1+CAR2+CAR3+CAR4+CHL1+ CHL2+CHL3+CHL4+AMN1+AMN2+AMN3+AMN4+ HYD1+HYD2+HYD3+HYD4

BSO4 NO3 (BSO4*0.375)+(NO3*0.29) NA CL H EC1+EC2+EC3-OP (O1+O2+O3+O4+OP)*1.4 CA MG K Not Applicable MF

PM10

MT

distribution, which is shown in Table 4-2, was estimated from plots given in Seinfeld and Pandis (1998). IC/BC PM concentrations used for the July 9-19, 1995 episode are listed in Table 4-3.

4.6 Model Outputs The URM-1ATM model produces hourly averaged, gridded concentrations of gas-phase and particulate species for all seven layers of the model. Also, produced are hourly, gridded wet and dry deposition fluxes at the ground. Since each episode characterizes a certain fraction of a typical season or year, it is important to evaluate the performance of the model outputs across the range of conditions represented by these episodes. Comprehensive statistical calculations have been performed for each species contributing to the seasonal and annual metrics to determine the ability of the model to accurately estimate ambient ozone and PM concentrations and acid deposition mass fluxes.

Table 4-2

The estimated size distribution of particulate species used in BC/ICs

Species

Bin 1

Bin 2

Bin 3

Bin 4

Sulfate Ammonium Nitrate Sodium Chloride Organics Elemental Carbon Inert PM

5.0% 5.0% 5.0% 0.25% 5.0% 40.0% 40.0% 0.25%

45.0% 45.0% 20.0% 1.75% 5.0% 25.0% 35.0% 0.75%

40.0% 40.0% 35.0% 23.0% 10.0% 30.0% 20.0% 4.0%

10.0% 10.0% 40.0% 75.0% 80.0% 5.0% 5.0% 95.0%

4-15

ODMAN,

ET AL.

Table 4-3 Species


The IC/BC concentrations (in µg/m3) for particulate species in July 1995 episode. IC

N/W/SW BC

E/SE BC

5 km

Model Top

SUL1 SUL2

2.6111E-01 2.3500E+00

6.1111E-02 5.5000E-01

6.1111E-03 5.5000E-02

4.2778E-03 3.8500E-02

2.7500E-03 2.4750E-02

SUL3 SUL4

2.0889E+00 5.2222E-01

4.8889E-01 1.2222E-01

4.8889E-02 1.2222E-02

3.4222E-02 8.5556E-03

2.2000E-02 5.5000E-03

NIT1 NIT2 NIT3

2.5000E-02 1.0000E-01 1.7500E-01

1.3333E-02 5.3333E-02 9.3333E-02

1.3333E-03 5.3333E-03 9.3333E-03

9.3333E-04 3.7333E-03 6.5333E-03

6.0000E-04 2.4000E-03 4.2000E-03

NIT4

2.0000E-01

1.0667E-01

1.0667E-02

7.4667E-03

4.8000E-03

AMN1 AMN2 AMN3 AMN4

9.7917E-02 8.8125E-01 7.8333E-01 1.9583E-01

2.2917E-02 2.0625E-01 1.8333E-01 4.5833E-02

2.2917E-03 2.0625E-02 1.8333E-02 4.5833E-03

1.6042E-03 1.4438E-02 1.2833E-02 3.2083E-03

1.0313E-03 9.2813E-03 8.2500E-03 2.0625E-03

SOD1 SOD2

4.0000E-04 2.8000E-03

2.0000E-04 1.4000E-03

1.0000E-03 7.0000E-03

1.4000E-05 9.8000E-05

9.0000E-06 6.3000E-05

SOD3 SOD4

3.6800E-02 1.2000E-01

1.8400E-02 6.0000E-02

9.2000E-02 3.0000E-01

1.2880E-03 4.2000E-03

8.2800E-04 2.7000E-03

CHL1 CHL2 CHL3

6.2500E-03 6.2500E-03 1.2500E-02

2.5000E-03 2.5000E-03 5.0000E-03

7.7500E-02 7.7500E-02 1.5500E-01

1.7500E-04 1.7500E-04 3.5000E-04

1.1250E-04 1.1250E-04 2.2500E-04

CHL4

1.0000E-01

4.0000E-02

1.2400E+00

2.8000E-03

1.8000E-03

HYD1 HYD2 HYD3 HYD4

5.6189E-04 1.6672E-03 1.5747E-03 8.2532E-04

2.7678E-04 8.6977E-04 8.4622E-04 2.3850E-04

2.1611E-03 1.9648E-03 5.1673E-04 2.2058E-02

1.9375E-05 6.0884E-05 5.9235E-05 1.6695E-05

1.2455E-05 3.9140E-05 3.8080E-05 1.0732E-05

WAT1 WAT2 WAT3 WAT4

2.0000E-01 1.0000E+00 9.0000E-01 2.0000E-01

2.0000E-01 1.0000E+00 9.0000E-01 2.0000E-01

2.0000E-01 1.0000E+00 9.0000E-01 2.0000E-01

1.4000E-02 7.0000E-02 6.3000E-02 1.4000E-02

9.0000E-03 4.5000E-02 4.0500E-02 9.0000E-03

ORG1 ORG2

2.4589E+00 1.5368E+00

4.3579E-01 2.7237E-01

4.3579E-02 2.7237E-02

3.0505E-02 1.9066E-02

1.9611E-02 1.2257E-02

ORG3 ORG4

1.8442E+00 3.0737E-01

3.2684E-01 5.4474E-02

3.2684E-02 5.4474E-03

2.2879E-02 3.8132E-03

1.4708E-02 2.4513E-03

CAR1 CAR2 CAR3

2.6526E-01 2.3211E-01 1.3263E-01

6.3158E-02 5.5263E-02 3.1579E-02

6.3158E-03 5.5263E-03 3.1579E-03

4.4211E-03 3.8684E-03 2.2105E-03

2.8421E-03 2.4868E-03 1.4211E-03

CAR4

3.3158E-02

7.8947E-03

7.8947E-04

5.5263E-04

3.5526E-04

CRC1 CRC2 CRC3 CRC4

1.5000E-03 4.5000E-03 2.4000E-02 5.7000E-01

2.5000E-03 7.5000E-03 4.0000E-02 9.5000E-01

2.5000E-04 7.5000E-04 4.0000E-03 9.5000E-02

1.7500E-04 5.2500E-04 2.8000E-03 6.6500E-02

1.1250E-04 3.3750E-04 1.8000E-03 4.2750E-02

CRM1 CRM2 CRM3 CRM4

3.0000E-04 9.0000E-04 4.8000E-03 1.1400E-01

1.5000E-04 4.5000E-04 2.4000E-03 5.7000E-02

1.5000E-05 4.5000E-05 2.4000E-04 5.7000E-03

1.0500E-05 3.1500E-05 1.6800E-04 3.9900E-03

6.7500E-06 2.0250E-05 1.0800E-04 2.5650E-03

CRK1 CRK2

4.3000E-03 1.2900E-02

2.2500E-03 6.7500E-03

2.2500E-04 6.7500E-04

1.5750E-04 4.7250E-04

1.0125E-04 3.0375E-04

CRK3 CRK4

6.8800E-02 1.6340E+00

3.6000E-02 8.5500E-01

3.6000E-03 8.5500E-02

2.5200E-03 5.9850E-02

1.6200E-03 3.8475E-02

4-16

ODMAN,

ET AL.

Table 4-13 Species PM1 PM2 PM3 PM4 SO2 (ppb) HCl (ppb)


(Continued) The IC/BC concentrations (in µg/m3) for particulate species in July 1995 episode. IC

N/W/SW BC

E/SE BC

5 km

Model Top

1.6000E-02 4.8000E-02 2.5600E-01 6.0800E+00

1.0000E-02 3.0000E-02 1.6000E-01 3.8000E+00

1.0000E-03 3.0000E-03 1.6000E-02 3.8000E-01

7.0000E-04 2.1000E-03 1.1200E-02 2.6600E-01

4.5000E-04 1.3500E-03 7.2000E-03 1.7100E-01

obs. 1.00

obs. 1.00

0.05 0.10

0.05 0.10

0.05 0.10

Among the statistical measures examined are mean bias, normalized mean bias, mean error, and normalized mean error. As will be described later, the biases and errors were calculated in different ways for ozone than for PM and deposition. The monitoring networks used for comparison include the Aerometric Information Retrieval System (AIRS) for ozone, Interagency Monitoring of Protected Visual Environments (IMPROVE) for PM, and National Atmospheric Deposition Program (NADP) for wet deposition. Model performance for meteorology, ozone, wet deposition, and dry deposition was computed using observations within the 12-km grid. PM performance was determined from data collected at sites in the 12-, 24-, 48-, and 96-km grids because of the small quantity of data available on the finest grid alone. A detailed discussion of the model performance for each of these pollutants will be the topic of the next three chapters. A comprehensive set of results and performance statistics for all SAMI episodes can be found in Appendix A as well as on the Georgia Tech web site (GIT, 2001).

4-17

ODMAN,

ET AL.



4-18

ODMAN,

5

ET AL.


OZONE MODEL PERFORMANCE

5.1 AIRS Observation Database Statistical calculations were made to evaluate the ability of the model to accurately estimate ambient ozone concentrations for seven episodes representing the ozone season (April 1 – October 31). The February 8-13, 1994 and March 23-31, 1993 episodes were not examined. The ozone statistical calculations were done using the Modeling Analysis and Plotting System (MAPS) package (McNally et al., 1991). Observations used in these calculations were obtained from the Aerometric Information Retrieval System (AIRS) database (USEPA, 2001). Among the statistical measures examined are the normalized mean bias and normalized mean error. There are several hundred AIRS stations within the modeling domain reporting data during the seven ozone episodes. However, some of these stations fall into coarse resolution cells where the model predictions are subject to significant smoothing and the observations are representative of much smaller scales. Therefore, only the stations falling within the 12-km grid are used in the performance analysis.

5.2 Daily Maximum Ozone Spatial Plots Spatial plots or maps of simulated daily maximum ozone were prepared for each simulated day. Only the portion of the modeling domain over the SAMI region was focused in these maps. Figure 5-1 shows an example for July 12, 1995. This map and others for other days can be found in Appendix A.

Figure 5-1

Daily maximum ozone on July 12, 1995. 5-1

ODMAN,

ET AL.


5.3 Diurnal Station Plots 5.3.1

Comparison to AIRS Data for Selected Episodes and Sites

Time series plots were used to understand how well the model performed at each site and how performance varied by time of day. These plots present both simulated and observed hourly concentrations throughout the simulation period. With the time series plot, one can determine the model's ability to reproduce the peak, the presence or absence of significant bias and errors within the diurnal cycle, and whether the "timing" of the estimated peak agrees with the observation. Since SAMI’s main concern is for ozone in Class I areas, it is also important to discern between performance at urban sites and high-elevation rural sites. The simulated hourly ozone values at 33 sites were compared to observations using diurnal station plots. The names of these sites, their AIRS code as well as their location type are listed in Table 5-1.

Table 5-1 Number

List of sites for which diurnal plots were prepared. Site Address

AIRS Code

1

Clay Co, AL

01-027-0001

rural

2

Huntsville, AL

01-089-0021

urban

3 4

Sipsey Wilderness area, Lawrence Co, AL Shelby Co, AL

01-079-0002 01-117-0004

rural rural

5

Washington, DC

11-001-0017

urban

6

Dawsonville, Dawson Co, GA

13-085-0001

rural

7

South DeKalb, DeKalb Co, GA

13-089-0002

urban

8

Yorkville, Paulding Co, GA

13-223-0003

urban

9

Lexington, Fayette Co, KY

21-067-0012

urban

10

Taylorsville, Alexander Co, NC

37-003-0003

rural

11

Asheville , Buncombe Co, NC

37-021-0030

urban

12

Lenoir, Caldwell Co, NC

37-027-0003

rural

13

Winston-Salem, Forsyth Co, NC

37-067-0022

urban

14

Frying Pan, Haywood Co, NC

37-087-0035

high elevation

15

Purchase Knob, Haywood Co, NC

37-087-0036

high elevation

16

Crouse, Lincoln Co, NC

37-109-0004

rural

17

Co Line, Mecklenburg Co, NC

37-119-1009

urban

18

Swain Co, NC

37-173-0002

rural

19

Powdersville, Anderson Co, SC

45-007-0003

urban

20

Oconee Co, SC

45-073-0001

rural

21

Look Rock, Blount Co, TN

47-009-0101

high elevation

22

Bradley Co, TN

47-011-0004

rural

23

Nashville, Davidson Co, TN

47-037-0011

urban

24

Chattanooga, Hamilton Co, TN

47-065-0028

rural

25

Knoxville, Knox Co, TN

47-093-1020

urban

26

Cove Mountain, Sevier Co, TN

47-155-0101

rural

27

Clingmans Dome, Sevier Co, TN

47-155-0102

high elevation

28

Blountville, Sullivan Co, TN

47-163-2002

rural

29

Big Meadows, Madison Co, VA

51-113-0003

high elevation

30

Roanoke Co, VA

51-161-1004

urban

31

Whitetop Mt, Smyth Co, VA

51-173-0003

high elevation

32

Greenbrier Co, WV

54-025-0001

high elevation

33

Charleston, Kanawha Co, WV

54-039-0004

urban

5-2

Location Type

ODMAN,

ET AL.


Ozone (ppb)

Knoxville (TN): May 11-18, 1993 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 11

Figure 5-1

12

13

14

15

16

17

18

Observed (* ) and simulated (-) ozone levels at Knoxville, TN for May 11-18, 1993.

The examples given here show the hourly ozone estimates and measurements at some typical urban and rural sites. Figure 5-1 shows ozone model performance at Knoxville, TN (a high elevation urban site) for the May 1993 episode. The model underestimates the peaks on most days; however, the daytime variations and the timing of the peaks are in good agreement with the observations. On the other hand, modeled nighttime evolution of ozone is considerably different from observations. The modeled ozone rises after sunset until early morning hours while the observations continue dropping to near-zero levels. This is due to mixing from the upper layers of the model along with an underestimation of the nighttime NOx emissions. Figure 5-2 and Figure 5-3 show the ozone performance at typical high-elevation rural sites. The sites include the Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks for

Ozone (ppb)

Great Smoky Mountains: August 3-11, 1993 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 3

Figure 5-2

4

5

6

7

8

9

10

11

Observed (* ) and simulated (-) ozone levels at Great Smoky Mountains, TN for August 3-11, 1993. 5-3

12

ODMAN,

ET AL.


Ozone (ppb)

Shenandoah: August 3-11, 1993 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 3

Figure 5-3

4

5

6

7

8

9

10

11

12

Observed (* ) and simulated (-) ozone levels at Shenandoah, VA August 3-11, 1993.

the August 1993 episode. Under predictions of nighttime ozone at the high-elevation sites indicate a potential misrepresentation of nighttime mixing conditions at these locations. Increasing horizontal and vertical resolution might improve the results. Modeled altitudes were much lower than the actual altitudes of these sites because the 12-km resolution is larger than the peaks, hence the modeled area includes lower elevation regions as well. Ozone concentrations in third and fourth layers of the model were often in better agreement with the observations. 5.3.2

Comparison to SOS Surface Data for the July 11-19, 1995 Episode

The model estimated concentrations of ozone and nitrogen oxides were compared to intensive observations collected during the Southern Oxidant Study (SOS). Time series plots were prepared for Centerville, AL, Mammoth Cave, KY, Cove Mountain, TN, Giles County, TN, Oak Grove, MS and Yorkville, GA for the July-95 episode. As seen in Figure 5-1, the 12-km grid resolution is insufficient to capture early morning nitrogen oxide peaks on some days. The model-estimated nitrogen dioxide concentrations on the other hand are in good agreement with the observations especially from July 14 to 18 (Figure 5-2). The ozone concentrations are shown in Figure 5-3. The agreement between the model estimates and observations is encouraging.

5.4 Bias and Error Calculations The time series plots above are useful for looking at specific stations, but to get an idea of how well the model is performing over all stations, daily mean normalized bias and error were calculated using the seventy four stations falling into 12-km grid cells. In each case a daily mean normalized bias (MNB) and daily mean normalized error (MNE) were calculated as:

1 MNB = N

N

∑

(c

e i

i =1

5-4

)

− c io × 100% c io

(5-1)

ODMAN,

ET AL.


Figure 5-1

Nitrogen oxide (NO) concentrations at Giles County, TN during July 11-19, 1995.

Figure 5-2

Nitrogen dioxide (NO2) concentrations at Giles County, TN during July 11-19, 1995. 5-5

ODMAN,

ET AL.

Figure 5-3


Ozone (O 3) concentrations at Giles County, TN during July 11-19, 1995

e o 1 N ci − ci MNE = ∑ ×100% N i =1 cio

where,

(5-2)

cie is the model-estimated concentration at station i, c io is the observed concentration at station

i, and N equals the number of estimate-observation pairs drawn from all valid monitoring stations for the time period of interest. If the time period of interest is hourly, N will consist of the number of stations with valid monitoring data in that hour. If the time period is daily, N will consist of the sum of the number of stations with valid monitoring data for each of the 24 hours in the day of interest. Since the normalized quantities can become large when the observations are small, a cutoff (or threshold) value of 40 ppb is used in conjunction with Equations 4-1 and 4-2. Whenever the observation is smaller than the cutoff value, that estimate-observation pair was excluded from the calculation. Examples of hourly mean normalized bias and error for the period of July 9-19, 1995 are shown in Figure 5-4 and Figure 5-5, respectively. The normalized (i.e., percent) bias is negative for the nighttime indicating an underestimation. There are brief periods of positive bias before noon and later in the afternoon indicating overestimation of ozone concentrations but midday concentrations are also underestimated. The normalized (i.e., percent) error is usually largest during the night. Charts for other episodes can be found in Appendix A.

5-6

ODMAN,

ET AL.


Figure 5-4

Hourly mean normalized bias (12-km grid only) for the July 9-19, 1995 episode.

Figure 5-5

Hourly mean normalized error (12-km grid only) for the July 9-19, 1995 episode. 5-7

ODMAN,

ET AL.


30.00 20.00 10.00 0.00 -10.00 -20.00 -30.00 7/30/91

7/28/91

7/26/91

7/24/91

5/03/95

5/01/95

4/29/95

4/27/95

8/11/93

8/9/93

8/7/93

8/5/93

8/3/93

6/28/92

6/26/92

6/24/92

5/16/93

5/14/93

5/12/93

5/29/95

5/27/95

5/25/95

7/19/95

7/17/95

7/15/95

7/13/95

7/11/95

Normalized Bias (%)

Ozone (12km - 40 ppb cutoff)

Episode Day Figure 5-6

Ozone daily mean normalized bias for the seventy four AIRS sites in the 12-km grid during the seven ozone episodes.

Figure 5-6 and Figure 5-7 show the daily mean normalized bias and error for each day during the seven episodes where measurements were available along with corresponding bounds recommended by EPA for urban-scale modeling (USEPA, 1991). Note that the daily mean normalized biases are within ± 15% with the exception of July 19, 1995; May 13, 1993; August 4, 1993; and the entire April 26 – May 3, 1995 episode. The daily mean normalized errors are less than 35% on all the modeled days. Mean normalized biases and errors show no evidence of growth on sequential days of any episode and there are no clear signs of a systematic bias, except for the April 26 – May 3, 1995 episode. This episode is biased low for every day that was modeled. Possible explanations include an underestimation in the biogenic VOC emissions and/or an underestimation of the NOx boundary conditions. Typically, urban applications involve much smaller domains and shorter simulation periods; therefore, urban-scale models can afford finer grid resolution (typically 4 or 5 km) and capture the ozone formation and deposition processes with more details. Meeting guidelines set for urban-scale

40.00 35.00 30.00 25.00 20.00 15.00 10.00 5.00 0.00 7/30/91

7/28/91

7/26/91

7/24/91

5/03/95

5/01/95

4/29/95

4/27/95

8/11/93

8/9/93

8/7/93

8/5/93

8/3/93

6/28/92

6/26/92

6/24/92

5/16/93

5/14/93

5/12/93

5/29/95

5/27/95

5/25/95

7/19/95

7/17/95

7/15/95

7/13/95

7/11/95

Normalized Error (%)

Ozone (12km - 40 ppb cutoff)

Episode Day Figure 5-7

Ozone daily mean normalized error for the seventy four AIRS sites in the 12-km grid during the seven ozone episodes. 5-8

ODMAN,

ET AL.


75

Model Ozone Error (ppb)

50 25 0 -25 -50 -75 0

50

100

150

Observed Daily Maximum 8-hr Ave. Ozone (ppb) Figure 5-8

Model estimated error versus observed daily maximum 8-hour average ozone.

models with a regional scale model and for a wide variety of meteorological conditions is encouraging. Ozone modeling performance was also analyzed based on the maximum daily 8-hour average ozone concentration, [O3]mx8, near the surface. Data from both urban and rural sites were used to compute error statistics for modeled ozone interpolated spatially to monitoring locations. Model error, eiO3, for estimate-observation pair i is calculated as:

eiO 3 = cie − cio e

o

(5-3) O3

where, c i is the model estimated value and c i is the observed value of [O3]mx8. Figure 5-8 plots e versus [O3]mx8 for all stations within the 12-km grid and all 7 episodes for which ozone results were analyzed. The model consistently overestimated [O3]mx8 below 40 ppb and underestimated ozone above 80 ppb for all episodes. This behavior was observed in other model applications (Irving, 1990), so it is not unique to URM-1ATM. This is believed to be the result of an inability of the modeling system (emissions, meteorological and air quality) to reproduce the full range of conditions that control ambient O3 pollutant concentrations. Finally, episodic averages of grid-wide e were compared to episodic biases in meteorology (air temperature, wind speed and water vapor mixing ratio), but no relationship was found.

5-9

ODMAN,

ET AL.



5-10

ODMAN,

6

ET AL.


AEROSOL MODEL PERFORMANCE

6.1 IMPROVE Observation Database PM model performance was evaluated by comparing modeling results to observations taken from the Interagency Monitoring of Protected Visual Environments (IMPROVE) monitoring network (NPS, 2000). IMPROVE measurements are taken on Wednesday and Saturday each week and are reported as twenty-four hour average concentrations. (Since 2000, IMPROVE measurements moved to a one-in-three-day schedule.) Therefore, measurements are only available for 23 of the 69 days modeled during the nine episodes. The species that were compared include fine (particles with an aerodynamic diameter < 2.5 µm) sulfate, nitrate, ammonium, elemental carbon, organic carbon, and soils (crustal minerals); total PM2.5 and PM10; and PM coarse (particles with an aerodynamic diameter between 2.5 and 10.0 µm). There are eighteen IMPROVE monitoring sites in the modeling domain. However, five of those sites are located near or on the boundary of the modeling domain and are heavily influenced by the boundary conditions. Therefore, only observations from the remaining thirteen stations are used to determine the PM model performance. Table 6-1 lists the stations and the model grid size at the corresponding location.

6.2 Spatial Plots of Daily Average Concentrations Spatial plots or maps of simulated daily average PM2.5 concentrations were prepared for each IMPROVE day. Only the portion of the modeling domain over the SAMI region was focused in these maps. Figure 6-1 shows an example for July 15, 1995. Maps for other days as well as sulfate, nitrate, ammonium, organic and elemental carbon, soils and PM10 can be found in Appendix A.

6.3 Daily Average Concentrations at IMPROVE Stations The four nearest grid nodes (i.e., corners of a grid cell) to each IMPROVE station are distance weighted (using bilinear interpolation) to determine the speciated PM concentration at each monitoring site. Simulated PM2.5 concentrations are compared to all available observations at each station for each episode. Figure 6-2 shows a comparison of daily average concentrations at Great Smoky Mountains Table 6-1

IMPROVE monitoring stations and URM-1ATM grid resolution at station location.

IMPROVE Station

Station ID

Grid cell size (km)

Brigantine, NJ

BRIG

48

Dolly Sods/Otter Creek, WV

DOSO

12

Great Smoky Mountains, TN

GRSM

12

Jefferson/James River Face, VA

JEFF

12

Lye Brook, VT

LYBR

96

Mammoth Cave, KY

MACA

24

Okefenokee, GA

OKEF

96

Cape Romain, SC

ROMA

48

Shenandoah, VA

SHEN

12

Shining Rock, NC

SHRO

12

SIPS

12

Upper Buffalo, AR

UPBU

96

Washington, DC

WASH

24

Sipsey, AL

6-1

ODMAN,

ET AL.

Figure 6-1


Daily average PM2.5 concentrations over the SAMI region on July 15, 1995.

for the July-1995 episode. Observations were available on three days of this episode: July 12, 15 and 19. These comparison charts are available in Appendix A for all the episodes and all IMPROVE sites and contain the sulfate, nitrate, ammonium, organic and elemental carbon and soil components of PM2.5 as well as PM10. Figure 6-3 shows a comparison of the simulated PM2.5 concentrations to the observations at every IMPROVE station on July 15, 1995. Note that the model underestimates PM2.5 at several stations and the underestimation is more pronounced for stations falling into coarse grid cells, for example

Figure 6-2

Simulated and observed PM2.5 concentrations at Great Smoky Mountains: July-1995 episode. 6-2

ODMAN,

ET AL.

Figure 6-3


Simulated and observed PM2.5 concentrations at all IMPROVE stations on July 15, 1995.

Brigantine, NJ (48 km) and Washington, DC (24 km). Similar charts are available in Appendix A for each simulated IMPROVE day. Individual comparisons of PM2.5 components (i.e., sulfate, nitrate, ammonium, organic and elemental carbon and soil) as well as PM10 are also available. Figure 6-4, shows a stacked bar chart comparing the modeled and measured PM 2+ concentrations on July 15, 1995 for sulfate (SO4 ), organics (ORG), ammonium (NH4 ), nitrate (NO3 ), elemental carbon (EC), soils, and other unidentified PM mass. Of all the IMPROVE days modeled, this day produced some of the highest PM2.5 concentrations in the Appalachian Mountains. The modeled concentrations are shown by the left stack while IMPROVE observations are displayed by the right stack for each station. The unidentified mass is expected to be mostly related to water associated with the particles and/or the assumption that the total organic mass is 1.4 times the organic carbon (OC) mass (Andrews et al., 2000). The model results do not contain any unidentified mass. The most abundant species in the fine PM (not including the unidentified mass) is sulfate, followed by organics and ammonium. The largest difference between modeled and observed PM2.5 on this day occurred at the Brigantine, NJ site. However, note that this monitoring station is located in a 48-km grid cell where

Simulated (L) and Observed (R) PM2.5 for July 15, 1995

PM 2.5 ( µ g/m3)

70.00 60.00

Other

50.00

Soils EC

40.00

NO3 NH4 ORG

30.00 20.00

SO4

10.00 0.00 BRIG DOSO GRSM JEFF LYBR MACA OKEF ROMA SHEN SHRO SIPS

Station Figure 6-4

Simulated (left) and Observed (right) speciated PM2.5 on July 15, 1995. 6-3

UPBU WASH

ODMAN,

ET AL.


the model estimates would be less reliable.

6.4 Scatter Plots Scatter plots were created by plotting the daily average simulated concentrations versus IMPROVE observations. All available observation-simulation pairs are shown as points on these plots. A point on the diagonal corresponds to a simulated value that perfectly matches the observed value. If the points are not too scattered around the diagonal than the correlation is strong between the simulated and observed values. The points above the diagonal correspond to overestimated values (positive bias, or biased high) and those below it correspond to underestimated values (negative bias, or biased low). The ±50% bias lines are also shown in these plots: the +50% bias line has a slope of 1.5 while the -50% bias line has a slope of 0.5. Finally, points were color coded by episode so that the behavior of the bias for different episodes can be seen. Figure 6-1 compares the simulated daily average fine sulfate concentrations with observations at every station on every IMPROVE day during the nine episodes. Overall, the model underestimates 3 sulfate concentrations. For observations below 3 µg/m the model generally overestimates sulfate, 3 often by more than 50%. Above 3 µg/m , there are no positive biases larger than +50%, underestimations are dominant and negative bias often exceeds -50%. Sulfate is underestimated during the May 1993, August 1993, and April 1995 episodes. For the other episodes, there is no clear indication of a systematic bias. The scatter in daily average concentrations of fine nitrate particles is 3 shown in Figure 6-2. There are large positive biases for concentrations below 1 µg/m where most of 3 the observed values fall. Nitrate concentrations above 1 µg/m are generally underestimated. The bias seems to be more dependent on the value of the concentration than episodic factors. The simulatedobserved pairs of daily average fine ammonium concentrations are shown in Figure 6-3. The pattern of the scatter is very similar to that of particulate sulfate (Figure 6-1) with a more prominent low bias. Most episodes show a clear underestimation. There is no sign of any systematic bias in organic PM (total organic matter) concentrations shown in Figure 6-4. During the May 1993 and April 1995 episodes, the organic PM concentrations are 3 generally less than 4 µg/m and the model estimates are typically low biased. The scatter plot for the elemental carbon concentrations in Figure 6-5 does not show any systematic bias. Most of the points fall between the ±50% bias lines. The soil concentrations on the other hand are severely overestimated as shown in Figure 6-6. This is most likely due to the large uncertainty in soil emission inventories. Finally, Figure 6-1 compares the measured and modeled PM2.5 for every station during the nine episodes. The majority of the modeled values fall within the ±50% bias lines. In general, the lower concentrations are biased high and the higher concentrations are biased low. Recall that IMPROVE observations may contain water which is not included in the model estimates. This may be one reason for underestimations but the overestimation of soils plays a compensating role. The underestimation of sulfate and ammonium, which are the largest components of PM 2.5 along with organics, is likely the most important factor leading to the underestimation of PM2.5.

6-4

ODMAN,

ET AL.


Sulfate Aerosol Concentration

URM Model Conc. ( g/m3)

25.0

July 1995 May 1995 May 1993 March 1993 February 1994 July 1991 June 1992 August 1993 April 1995

20.0

15.0

10.0

5.0

0.0 0.0

5.0

10.0

15.0

20.0

25.0

3

IMPROVE Measurements (µ g/m ) Figure 6-1

IMPROVE measurements vs. simulated concentrations for sulfate. Also shown are the 1:1 and ±50% bias lines.

Nitrate Aerosol Concentration


5.0


4.0

3.0

2.0

1.0

0.0 0.0

1.0

2.0

3.0

4.0

5.0

IMPROVE Measurements (µ g/m3) Figure 6-2

IMPROVE measurements vs. simulated concentrations for nitrate. Also shown are the 1:1 and ±50% bias lines.

6-5

ODMAN,

ET AL.


Ammonium Aerosol Concentration

URM Model Conc. ( g/m3 )

10.0


8.0

6.0

4.0

2.0

0.0 0.0

2.0

4.0

6.0

8.0

10.0 3


IMPROVE measurements vs. simulated concentrations for ammonium. Also shown are the 1:1 and ±50% bias lines.

Organic Aerosol Concentration


12.0


10.0 8.0 6.0 4.0 2.0 0.0 0.0

2.0

4.0

6.0

8.0

10.0

12.0

3


IMPROVE measurements vs. simulated concentrations for organics. Also shown are the 1:1 and ±50% bias lines.

6-6

ODMAN,

ET AL.


EC Aerosol Concentration


4.0


3.0

2.0

1.0

0.0 0.0 Figure 6-5

2.0

3.0

4.0

IMPROVE measurements vs. simulated concentrations for elemental carbon. Also shown are the 1:1 and ±50% bias lines.

Soil Aerosol Concentration

10.0


1.0

IMPROVE Measurements (µ g/m3)


8.0

6.0

4.0

2.0

0.0 0.0

2.0

4.0

6.0

8.0

10.0 3


IMPROVE measurements vs. simulated concentrations for soils. Also shown are the 1:1 and ±50% bias lines.

6-7

ODMAN,

ET AL.


PM2.5 Aerosol Concentration URM Model Conc. ( g/m3)

70.0 60.0


50.0 40.0 30.0 20.0 10.0 0.0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

IMPROVE Measurements (µ g/m3 ) Figure 6-1

IMPROVE measurements vs. simulated concentrations for PM2.5. Also shown are the 1:1 and ±50% bias lines.

6.5 Bias and Error Calculations The mean observed concentration (MOC), mean bias (MB) and mean error (ME) were calculated as follows:

MOC =

MB =

ME = where

1 N

1 N

N

∑c

(6-1)

i =1

∑ (c N

e i

i =1

o i

− cio )

1 N e ci − c io ∑ N i =1

(6-2)

(6-3)

cie is the model-estimated daily average concentration, c io is the observed daily average

concentration and i denotes a daily estimate-observation pair at any given station. All reporting IMPROVE stations in the 12, 24, 48, and 96 km grid cells were used. N is the number of estimateobservation pairs drawn from all valid monitoring station data for the comparison time period of interest. Next, the normalized mean bias (NMB) and normalized mean error (NME) were calculated as 6-8

ODMAN,

ET AL.


NMB =

MB × 100% MOC

(6-4)

NME =

ME ×100% MOC

(6-5)

Note that these definitions are slightly different then the definitions of mean normalized bias (MNB) and mean normalized error (MNE) used for ozone performance evaluation. Since N was typically small, the normalized bias and error were very sensitive to a cutoff value. In the absence of a cutoff value, the expressions in Equations 5-1 and 5-2 can become very large when the observed value is small. Therefore, the definition of normalized bias and error were altered as in Equations 6-4 and 6-5. Table 6-1 shows mean observed concentration, mean bias and mean error for PM2.5 along with normalized mean bias and error. Similar tables are available in Appendix A for sulfate, nitrate, ammonium, organic and elemental carbon, and soil components of PM2.5 as well as for PM10.

6.6 Discussion of Performance Results Figure 6-1 and Figure 6-2 contain modeled and observed gravimetric PM2.5 at Great Smoky Mountains and Shenandoah National Parks for all 23 days that IMPROVE measurements were taken. The model typically under predicts the PM2.5 concentrations at both sites. This is partially due to the model not accounting for the unidentified mass in the observations for which there is no corresponding Table 6-1

Date 7/12/1995 7/15/1995 7/19/1999 5/24/1995 5/27/1995 5/12/1993 5/15/1993 3/24/1993 3/27/1993 3/31/1993 2/9/1994 2/12/1994 4/26/1995 4/29/1995 5/3/1995 6/24/1992 6/27/1992 8/4/1993 8/7/1993 8/11/1993 7/24/1991 7/27/1991 7/31/1991

Daily mean concentration, bias, normalized bias, error and normalized error of PM2.5 by IMPROVE day. Number of Reporting Stations

Mean Conc. (µg/m3)

Mean Bias (µg/m3)

Mean Norm. Bias (%)

Mean Error (µg/m3)

Mean Norm. Error (%)

9 10 9 9 10 7 7 7 7 6 7 5 10 10 9 7 7 7 7 7 3 3 3

26.91 33.42 14.15 17.26 15.90 19.94 17.06 9.50 12.50 8.58 9.26 6.79 10.35 11.18 10.02 21.59 18.05 17.50 24.87 21.97 19.97 25.10 34.50

-6.27 -7.13 -0.54 -6.16 1.60 -4.06 -4.42 1.42 -1.81 3.24 -3.85 11.58 -4.97 -5.51 -2.80 -11.43 -4.38 -5.16 -11.79 -4.50 -0.23 1.53 -8.90

-23.31 -21.33 -3.83 -35.71 10.08 -20.37 -25.93 14.96 -14.50 37.81 -41.58 170.55 -48.01 -49.28 -27.96 -52.92 -24.26 -29.47 -47.41 -20.48 -1.14 6.08 -25.79

6.41 7.18 4.08 6.84 2.65 6.92 4.70 1.88 5.66 6.74 4.25 12.34 5.06 5.51 3.49 11.43 6.07 5.47 12.32 6.00 3.17 4.97 12.90

23.81 21.48 28.84 39.61 16.65 34.72 27.57 19.74 45.24 78.53 45.90 181.74 48.89 49.28 34.83 52.92 33.64 31.27 49.53 27.32 15.86 19.81 37.38

6-9

ODMAN,

ET AL.


3

PM 2.5 Conc. ( µg/m )

Great Smoky Mt. National Park 50.0 40.0 30.0 20.0 10.0 0.0 07/19/95

07/15/95

07/12/95

05/27/95

05/24/95

050/3/95

04/29/95

04/26/95

02/12/94

020/9/94

08/11/93

Figure 6-1

08/07/93

08/04/93

05/15/93

05/12/93

03/31/93

03/27/93

03/24/93

06/27/92

06/24/92

07/31/91

07/27/91

07/24/91

Model

Observations

Simulated and Observed (gravimetric) PM2.5 at Great Smoky Mountains National Park.

species in the model. Table 6-1 contains a summary of the performance statistics for each particulate species. In addition to using Equations 6-4 and 6-5 to calculate the normalized mean bias (NMB) and normalized mean error (NME) for each species, the fractional bias (FB) and fractional error (FE) are also presented for comparison:

1 N 2(cie − cio ) FB = ∑ e N i =1 (ci + cio )

(6-6)

3

PM 2.5 Conc. ( µ g/m )

Shenandoah National Park 50.0 40.0 30.0 20.0 10.0 0.0 07/19/95

07/15/95

07/12/95

05/27/95

6-10

05/24/95

Simulated and Observed (gravimetric) PM2.5 at Shenandoah National Park.

050/3/95

Observations

04/29/95

04/26/95

02/12/94

020/9/94

08/11/93

Figure 6-2

08/07/93

08/04/93

05/15/93

05/12/93

03/31/93

03/27/93

03/24/93

06/27/92

06/24/92

07/31/91

07/27/91

07/24/91

Model

ODMAN,

ET AL.


Table 6-1

Performance statistics for speciated PM2.5 (all episodes).

Species

Number Mean Conc. (µg m -3 )

MB (µg m -3 )

NMBb (%)

FBc (%)

ME (µg m -3 )

NMEd (%)

FEe (%)

Sulfate

219

6.63

-1.51

-22.8

-29.4

2.48

37.5

49.9

Nitrate

219

0.57

0.10

17.2

-0.32

0.56

98.3

86.9

Ammonium

219

2.65

-0.84

-31.7

-35.8

1.12

42.1

52.2

Organics

218

3.07

-0.33

-10.8

-12.2

1.17

38.2

40.6

Elem. Carbon

217

0.52

-0.09

-16.9

-20.7

0.23

44.2

50.8

Soils

221

0.55

0.92

167.1

68.5

1.03

187.0

86.0

PM2.5

216

16.31

-3.69

-22.6

-23.9

6.01

36.8

43.3

PM10

206

22.34

-1.44

-6.4

-5.9

8.25

36.9

37.7

Coarse PM

202

6.76

1.91

28.2

33.3

5.07

75.1

62.1

a

a

Ammonium from IMPROVE is assumed to be in form of ammonium sulfate, (NH4) 2SO4, and ammonium nitrate, NH4NO3. Using Equation 6-4 c Using Equation 6-6 b

d

Using Equation 6-5

e

Using Equation 6-7

1 FE = N where

N

2 cie − cio

∑ (c e + c o ) i =1

i

(6-7)

i

cie is the model-estimated daily average concentration, c io is the observed daily average

concentration, i denotes a daily estimate-observation pair at any given station, and N is the number of estimate-observation pairs drawn from all valid monitoring station data for the comparison time period of interest. Note that these definitions bound the fractional bias (FB) between -200% and +200% and bound the fractional error (FE) between 0% and +200%. The biases calculated using Equations 6-4 and 6-6 are directionally consistent (except for nitrate), and are similar in magnitude (except for nitrate and soils). The errors calculated using Equations 6-5 and 6-7 are all within 13% of each other (except for soils). The normalized biases and errors mentioned in the rest of this discussion are those calculated using Equations 6-4 and 6-5. Typically, the largest portion of PM2.5 consists of sulfate (approximately 40%). Sulfate is produced by the gas-phase oxidation of SO2 by OH or in the aqueous phase by peroxides (H2O2 and organic peroxides) and/or ozone when a rain, cloud, or fog droplet is present. The normalized mean error for sulfate is 37.5%. Particulate nitrate, formed by gas-to-particle conversion of -3 nitric acid, is usually found in low levels (less than 1.0 µg m ); therefore, the normalized error can be very high even though the absolute error is quite low. Particulate ammonium is usually found as ammonium sulfate ((NH4)2SO4), ammonium bisulfate (NH4HSO4), ammonium nitrate (NH4NO3), and/or mixed salts. The ammonium concentration primarily depends on the amount of sulfate and gas-phase ammonia that is available. Direct ammonium measurements began at GRSM, SHEN, and DOSO in 1997; however, there are no direct measurements of ammonium at any of the IMPROVE stations during the modeling periods (19911995). Therefore, for the purpose of ammonium performance evaluation, it has been assumed that the + sulfate and nitrate are completely neutralized with NH4 . The particulate ammonium predictions are biased low at most stations and the normalized mean error is approximately 42%. The under prediction 6-11

ODMAN,

ET AL.


in ammonium might be attributed to the under prediction in sulfate and/or the assumption that the observed sulfate is fully neutralized as ammonium sulfate. Analysis of direct ammonium measurements at GRSM, SHEN, and DOSO from 1997-1999 indicates that the sulfate typically is not fully neutralized (ARS, 2001). Organic PM2.5 is biased low and the normalized mean error is approximately 38%. The error can possibly be attributed to inaccuracies in the organic aerosol yields that were used or a deficiency in the emission inventory. Also, IMPROVE measures organic carbon (OC) and use a multiplier of 1.4 (to account for hydrogen and oxygen) to convert OC to total organic matter. It should be noted that this value may be low. It has been suggested that the conversion factor should be 1.6 for urban areas and 2.1 for rural areas (Turpin and Lim, 2001). When these conversion factors are applied to the -3 observations, the mean concentration increases to 4.44 µg m and the resulting performance statistics -3 -3 are: MB = -1.29 µg m , NMB = -38.40%, ME = 2.03 µg m , and NME = 45.72%. -3

Like nitrate, the observations for elemental carbon (EC) are low ( 9.93

Nitrate Ammonium (as Sulfate) Ammonium (as Bisulfate) Organics Elemental Carbon Soils PM2.5 PM10 Coarse PM

0.57 2.65 1.41 3.07 0.52 0.55 16.31 22.34 6.76

< 0.20 < 1.30 < 0.70 < 1.65 < 0.25 < 0.25 < 8.40 < 12.69 < 2.60

0.20 - 0.71 1.30 - 3.83 0.70 - 2.05 1.65 - 4.16 0.25 - 0.65 0.25 - 0.77 8.40 - 23.40 12.69 - 29.27 2.60 - 9.68

> 0.71 > 3.83 > 2.05 > 4.16 > 0.65 > 0.77 > 23.40 > 29.27 > 9.68

6-12

ODMAN,

ET AL.

Table 6-3


Mean and normalized bias as a function of classification for each particulate species. Low

Species

MB

Sulfate Nitrate Ammonium (as Sulfate) Ammonium (as Bisulfate) Organics Elemental Carbon Soils PM2.5 PM10 Coarse PM

Middle

High

NMB

MB

NMB

MB

NMBb

0.66

31.9

-1.45

-25.4

-3.82

-27.6

0.47 0.30 0.70 0.48 0.11 0.98 2.06 5.02 5.99

321.2 33.5 141.0 40.6 62.8 591.2 33.7 55.1 301.3

0.21 -0.75 0.37 -0.24 -0.08 0.91 -3.26 -0.72 3.24

55.7 -32.1 29.6 -8.8 -19.4 187.1 -22.5 -3.5 57.2

-0.62 -2.23 0.19 -1.42 -0.30 0.90 -10.77 -10.06 -6.23

-39.0 -41.8 6.9 -23.6 -25.8 76.6 -33.5 -24.5 -41.8

a

a

Mean bias (µg m-3) calculated using Equation 6-2

b

Normalized mean bias (%) calculated using Equation 6-4

b

a

b

a

observed concentrations increase. Particulate ammonium predictions are typically biased high at low concentrations, especially when bisulfate is assumed. However, the bias for the middle and high values depends on the ammonium sulfate/bisulfate assumption. Assuming that the sulfate is present as ammonium sulfate leads to results being biased low. If instead, the measured sulfate was assumed to be ammonium bisulfate, then the model results were biased high. Results from the Southeastern Aerosol and Visibility Study (SEAVS) suggests that at low sulfate levels, there is ample ammonia to form ammonium sulfate, though as sulfate increases, more ammonium bisulfate is formed (EPRI, 1998). This is consistent with the model results. Biases for organics and EC both tend to become more negative as the observed concentrations increase. However, soils show a consistent overestimation with little variation in mean bias across classes. Modeled soils consist of more species than those measured by IMPROVE, and there is a suspected bias in the emission inventory. Model performance for PM2.5 and PM 10 tend to be fairly similar. Both are overestimated at low concentrations and bias tends to become negative as the observed concentrations increase. Coarse -3 PM typically was positively biased for observations less than 10.0 µg m and negatively biased when -3 the concentrations were greater than 10.0 µg m . EPA guidance criteria does not yet exist for determining whether PM model performance is acceptable; however, the errors in PM 2.5 and PM 10 are consistent with, if not better than, those obtained by other models (Seigneur, 2001). This is especially encouraging since the duration of the SAMI episodes were typically longer than those of previous studies and cover a wide range of meteorological conditions including winter, spring, and summer episodes. Further, the regional domain modeled is larger than those used in previous studies (e.g., the Los Angeles basin).

6-13

ODMAN,

ET AL.



6-14

ODMAN,

7

ET AL.


WET DEPOSITION MODEL PERFORMANCE

7.1 NADP Observation Database Wet acid deposition performance was evaluated by comparing modeling results to observations taken from the National Atmospheric Deposition Program (NADP) monitoring network. The species that were compared include sulfate, nitrate, ammonium, hydrogen ion, and crustal cations 2+ 2+ (Mg and Ca ). Weekly NADP samples are collected on Tuesdays and the concentrations and precipitation are reported as seven-day totals. Multiplying the measured concentration by the precipitation results in deposition mass flux. There are eighty-three NADP monitoring sites in the modeling domain. However, since wet deposition is very localized and coarse grids are not adequate to resolve localized phenomenon, only data from the stations in the 12-km grids are used for performance evaluation. These stations are listed in Table 7-1.

7.2 Spatial Plots of 7-Day Cumulative Deposition Spatial plots or maps of simulated weekly cumulative wet deposition fluxes were prepared for each episode. Only the portion of the modeling domain over the SAMI region was focused in these maps. Figure 7-1 shows the map of sulfate wet deposition for the week of July 11-18, 1995. Maps for other episodes as well as wet deposition fluxes of nitrate, ammonium, calcium, magnesium and hydronium ion can be found in Appendix A.

7.3 7-Day Cumulative Deposition at NADP Stations The large spatial variability in precipitation and difficulty in modeling convective precipitation sometimes resulted in a poor match between the modeled and observed deposition fluxes. Therefore, NADP observations were compared to the best (i.e., closest in magnitude to the observation) model result within a 30 km radius of the monitoring site. Although the cell values sometimes have a tendency to miss the observed sulfate mass flux, the best values usually match well and were used in all the model performance calculations. Figure 7-2 shows a comparison of NADP observations for sulfate to the cell and best model predictions for the seven-day cumulative period of March 23-30, 1993. Note that there may be large differences between the observation and the cell value, but the best value Table 7-1

NADP stations falling into the 12-km grid cells.

NADP Station Sand Mountain Exp. Station Georgia Station Lilley Cornett Woods Coweeta Piedmont Research Station Mt. Mitchell Walker Branch Watershed Great Smoky Mountains National Park – Elkmont Wilburn Chapel Charlottesville Horton’s Station Shenandoah National Park – Big Meadows Babcock State Park Parsons

7-1

Station ID

County

State

AL99 GA41 KY22 NC25 NC34 NC45 TN00 TN11 TN98 VA00 VA13 VA28 WV04 WV18

De Kalb Pike Letcher Macon Rowan Yancey Anderson Sevier Giles Albemarle Giles Madison Fayette Tucker

AL GA KY NC NC NC TN TN TN VA VA VA WV WV

ODMAN,

ET AL.

Figure 7-1


Weekly cumulative sulfate wet deposition for July 11-18, 1995.

generally agrees well with the observation. Station WV04 (Babcock State Park, WV) demonstrates that an error in the location of the precipitation by a couple of grid cells leads to large changes in precipitation rates and wet depositional mass fluxes.

Mass Flux (mg/m 2)

Plots were also prepared for wet deposited concentrations, which were obtained by dividing the deposition flux by the precipitation (in mm). Figure 7-4 gives an example for sulfate for the week of March 23-30, 1993. Similar plots for other episodes as well as other species can be found in Appendix A.

350 300 250 200 150 100 50 0

Sulfate Wet Deposition

Cell Best NADP

AL99 GA41 KY22 NC25 NC34 NC45 TN00 TN11 TN98 VA00 VA13 VA28 WV04 WV28

Station Figure 7-2

Simulated and observed (NADP) sulfate wet deposition fluxes at 14 monitoring sites in the 12-km grid for the week of March 23-30, 1993. 7-2

ODMAN,

ET AL.

Figure 7-4


Simulated and observed wet-deposited sulfate concentrations for the week of March 23-30, 1993.

Figure 7-3 and Figure 7-1 contain modeled (cell and best) and observed sulfate and nitrate wet deposition mass fluxes at Great Smoky Mountains. It includes results for all the episodes that NADP measurements were available. No measurements were available at Great Smoky Mountains for the July 1995 episode. Biases in the depositional flux are directly correlated to the biases in the cumulative precipitation (Figure 7-2) for each episode.

7.4 Scatter Plots of Seven-day Cumulative Wet Deposition Figure 7-3 shows a scatter plot of observed versus modeled weekly cumulative wet deposition fluxes of sulfate at all the NADP stations in the 12 km grid for each episode. There is a strong correlation between modeled and observed mass fluxes with very few data points falling outside the 2 ±50% bias lines. Fluxes smaller than about 40 mg/m are generally overestimated and larger fluxes are underestimated. The underestimation is most severe for the February 1994 episode. The correlation is also strong between the simulated and observed fluxes of nitrate wet deposition especially for values 2 below 50 mg/m (Figure 7-4). However there are some severe underestimations for larger fluxes. The

Deposition Flux (mg/m2)

Great Smoky Mnts - Elkmont (TN11) - Sulfate Deposition 300.0 250.0 200.0 150.0 100.0 50.0 0.0 Jul 1991

Jun1992

Mar 1993

May 1993

Model Cell Value Figure 7-3

Aug 1993

Feb 1994

Model "Best" Value

Sulfate deposition mass flux at Great Smoky Mountains National Park 7-3

Apr 1995

Observation

May 1995

Jul 1995

ODMAN,

ET AL.



Great Smoky Mnts - Elkmont (TN11) - Nitrate Deposition 150.0 125.0 100.0 75.0 50.0 25.0 0.0 Jul 1991

Jun1992

Mar 1993

May 1993


Aug 1993

Feb 1994

Model "Best" Value

Apr 1995

May 1995

Jul 1995

Observation

Nitrate deposition mass flux at Great Smoky Mountains National Park

July 1995 episode is low biased at almost all NADP stations. There is a systematic high bias in ammonium wet deposition fluxes across all episodes (Figure 7-5). To date, the reason for this overprediction is unknown. Currently, the model is being equipped with a tool to calculate the reduced-nitrogen balance ýn order to investigate this overproduction. Very 2 few fluxes exceed 20 mg/m and these are underestimated. Calcium wet deposition is high biased for the March 1993 and April 1995 episodes and low biased for the July 1995 episode (Figure 7-6). On the other hand the correlation for the May 1993 episode is very strong. For the other episodes the scatter is 2 large generally with overestimations for fluxes below 3 mg/m and underestimates for larger fluxes. Similar to the ammonium wet deposition there is a strong positive bias in the magnesium wet 2 deposition fluxes (Figure 7-7). The bias diminishes for the very few values exceeding 1 mg/m . In interpreting these results, note that the precipitation values estimated by RAMS, which are input to the URM-1ATM model, were high biased for observed values below 30 mm and low bias for larger values (Figure 7-8). The precipitation for the March 1993 and February 1994 episodes were significantly underestimated.

Great Smoky Mnts - Elkmont (TN11) - Precipitation Precipitation (mm)

150.0 125.0 100.0 75.0 50.0 25.0 0.0 Jul 1991

Jun1992

Mar 1993

May 1993


Aug 1993

Feb 1994

Model "Best" Value

Precipitation at Great Smoky Mountains National Park 7-4

Apr 1995

Observation

May 1995

Jul 1995

ODMAN,

ET AL.


Sulfate Wet Deposition Flux

URM Model Flux (mg/m2)

400.0


300.0

200.0

100.0

0.0 0.0

100.0

200.0

300.0

400.0

NADP Measurements (mg/m2) Figure 7-3

National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated sulfate flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines.

Nitrate Wet Deposition Flux


160.0


120.0

80.0

40.0

0.0 0.0

40.0

80.0

120.0

160.0 2

NADP Measurements (mg/m ) Figure 7-4

National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated nitrate flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines.

7-5

ODMAN,

ET AL.


Ammonium Wet Deposition Flux


60.0


50.0 40.0 30.0 20.0 10.0 0.0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

NADP Measurements (mg/m2) Figure 7-5

National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated ammonium flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines.

Calcium Wet Deposition Flux


14.0 12.0


10.0 8.0 6.0 4.0 2.0 0.0 0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

2


National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated calcium flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines.

7-6

ODMAN,

ET AL.


Magnesium Wet Deposition Flux


7.0


6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

2


National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM simulated magnesium flux using the "best" cell value. Also shown are the 1:1 and ±50% bias lines.

Precipitation

URM Model Prec. (mm)

160.0


120.0

80.0

40.0

0.0 0.0

40.0

80.0

120.0

160.0

NADP Measurements (mm) Figure 7-8

National Atmospheric Deposition Program (NADP) measurements (stations in the 12-km grid) vs. URM-1ATM model input precipitation using the "best" cell value. Also shown are the 1:1 and ±50% bias lines.

7-7

ODMAN,

ET AL.


7.5 Bias and Error Calculations Mean observed flux (MOF), mean bias (MB) and mean error (ME) were calculated as follows:

MOF =

MB =

ME = where

1 N

1 N

N

∑f

o i

(7-1)

i =1

∑(f N

e i

i =1

− f io )

1 N e fi − f i o ∑ N i =1

(7-2)

(7-3)

f i e is the model-estimated weekly cumulative deposition flux, f i o is the observed weekly

cumulative deposition flux and i denotes a weekly estimate-observation pair at any given station. Only reporting NADP stations in the 12-km grid cells were used. N is the number of estimate-observation pairs drawn from all valid monitoring station data for the comparison time period of interest. Normalized mean bias (NMB) and normalized mean error (NME) were calculated as:

NMB =

MB × 100% MOF

(7-4)

NME =

ME × 100% MOF

(7-5)

Note that these definitions are identical to those used for PM in Section 6.5. Table 7-1 shows mean observed flux, mean bias and mean error for sulfate wet deposition along with normalized mean bias and error. Similar tables are available in Appendix A for nitrate, ammonium, calcium, magnesium and hydronium ion. Performance statistics for all these tables were calculated using the best model prediction.

7.6 Discussion of Performance Results Table 7-2 contains a summary of the performance statistics for all the species undergoing wet deposition. In addition to using Equations 7-4 and 7-5 to calculate the normalized mean bias (NMB) and normalized mean error (NME) for each species, the fractional bias (FB) and fractional error (FE) are also presented for comparison:

1 N 2( f i e − f i o ) FB = ∑ e N i=1 ( f i + f i o ) 7-8

(7-6)

ODMAN,

ET AL.

Table 7-1


Weekly mean deposition flux and model bias, normalized bias, error and normalized error of sulfate by episode. Mean Flux Number of Reporting Stations (mg/m2 )

Date July 11-18, 1995 May 23-30, 1995 May 11-18, 1993 March 23-30, 1993 Feb 8-15, 1994 April 26 - May 3, 1995 August 3-11, 1993 June 24-29, 1992 July 23-30, 1991

9 12 10 14 13 14 14 14 11

Norm. Bias (%)

-9.67 -3.74 -11.50 1.50 -32.06 1.28 4.51 -6.02 -9.31

-14.91 -8.79 -10.10 1.80 -26.89 1.94 5.73 -9.99 -7.88

64.89 42.56 113.81 83.40 119.22 66.00 78.62 60.26 118.09

1 FE = N where

Mean Bias (mg/m2 )

N

Mean Error Norm. Error (mg/m2 ) (%) 14.29 24.09 18.73 13.72 32.19 5.50 19.82 10.14 21.06

22.02 56.61 16.46 16.45 27.00 8.34 25.21 16.83 17.83

2 fie − fio

∑ (f e + f o) i =1

i

(7-7)

i

f i e is the model-estimated weekly cumulative deposition flux, f i o is the observed weekly

cumulative deposition flux, i denotes a weekly estimate-observation pair at any given station, and N is the number of estimate-observation pairs drawn from all valid monitoring station data for the comparison time period of interest. Note that these definitions bound the fractional bias (FB) between -200% and +200% and bound the fractional error (FE) between 0% and +200%. A total of 105 estimate-observation pairs were used from all the episodes to calculate these values. All performance statistics were calculated using the best model prediction. The biases calculated using Equations 7-4 and 7-6 are directionally consistent (except for sulfate), but may show a large difference in magnitude. The errors calculated using Equations 7-5 and 7-7 are all within 3% of each other (except for ammonium, hydrogen ion, and magnesium). The normalized biases and errors mentioned in the rest of this discussion are those calculated using Equations 7-4 and 7-5. Sulfate and nitrate both have a small negative bias and the normalized mean errors are less than 25%. The wet deposition of ammonium was biased high for all episodes.

Table 7-2

Performance statistics for the wet deposition species (all episodes).

Species

Mean Flux (mg m-2)

MB (mg m-2)

NMBa (%)

NMBb (%)

ME (mg m-2)

NMEc (%)

NMEd (%)

Sulfate

83.48

-6.55

-7.8

5.9

20.71

24.8

28.1

Nitrate

45.99

-5.43

-11.8

-2.9

10.35

22.5

24.6

Ammonium

9.32

10.42

111.8

83.0

10.72

115.0

84.9

Hydrogen Ion

1.76

-1.01

-57.2

-69.6

1.03

58.6

75.8

Calcium

3.14

0.72

22.9

30.8

1.39

44.4

46.4

Magnesium

0.50

0.70

140.4

52.9

0.80

160.7

78.4

Precipitation (mm)

40.14

-6.53

-16.27

-4.8

10.08

25.12

27.8

a

Using Equation 7-4

b

Using Equation 7-6 Using Equation 7-5

c d

Using Equation 7-7

7-9

ODMAN,

ET AL.


Both the calcium and magnesium deposition fluxes were biased high. While some of the discrepancies between modeled and measured fluxes can be attributed to differences in the modeled and observed precipitation rates, the high bias for calcium and magnesium is more likely due to a high bias in the crustal emission rates. Again, EPA guidance criteria does not exist for determining whether wet deposition model performance is acceptable. Deposition bias was evaluated as a function of observed wet deposition flux in a manner similar to that used for PM. Each of the modeled wet deposition species has been assigned to a flux bin (low, middle, and high) based on its observed value. The “low” category contains the lowest 20% by deposition flux, the “high” category contains the highest 20%, and the “middle” category contains the remaining 60% between the low and high categories. Table 7-3 shows the average mass flux and the ranges associated with each species. Table 7-1 shows the mean and normalized bias as a function of classification bin for each wet deposition species. Bias for the simulated wet deposition of sulfate and nitrate were both positive for the low observations and tended to become more negative with increasing observed mass flux. This is similar to ozone, particulate sulfate and particulate nitrate trends. Simulated ammonium deposition was almost always greater than observed, while the hydrogen ion predictions were almost always underestimated. The ammonium normalized bias decreases with increasing ammonium flux; the higher modeled fluxes -2 (greater than 12.0 mg m ) tend to match better with the observations. The hydrogen bias tends to increase as the hydrogen deposition flux increases. Both the simulated calcium and magnesium deposition fluxes were usually biased high. However, calcium bias tended to be negative as the observed mass flux increased. There is no clear bias trend for the magnesium flux as a function of deposition flux bin. Precipitation amount was often overestimated when the observed rainfall was less than 14 mm and underestimated as the observed precipitation increased. It should be noted that the trends in sulfate and nitrate biases were very similar to the trend in precipitation. If the bias in precipitation could be reduced, the performance of many deposition species would likely improve.

7.7 Comparison to Other Studies Probably the closest modeling study to the SAMI atmospheric modeling conducted to date was that performed for the National Acid Precipitation Assessment Program (NAPAP). As part of NAPAP, the Regional Acid Deposition Model (RADM) was developed and applied to better quantify how controls will affect future levels of acid deposition, very similar to part of the objective here (they were not considering other endpoints such as ozone and visibility). In the NAPAP study (NAPAP, 1990), they evaluated RADM using data from a 33-day period in the summer of 1988 as part of the Eulerian Model Evaluation Field Study (EMEFS), and also used an aggregation of 30, three-day simulations from the early 1980’s. RADM was evaluated for its ability to simulate, among other

Table 7-3

Average observed (NADP) mass flux and classification bin for each wet deposition species at the 14 monitoring stations in the 12-km grid. Range of Observations (mg m -2 )

Species

Average

Low

Middle

High

Sulfate

83.48

< 31.10

31.10 - 120.80

> 120.80

Nitrate Ammonium Hydrogen Ion Calcium Magnesium Precipitation (mm)

45.99 9.32 1.76 3.14 0.50 40.14

< 18.75 < 3.00 < 0.68 < 1.22 < 0.20 < 13.75

18.75 - 62.60 3.00 - 12.75 0.68 - 2.60 1.22 - 5.00 0.20 - 0.75 13.75 - 60.00

> 62.60 > 12.75 > 2.60 > 5.00 > 0.75 > 60.00

7-10

ODMAN,

ET AL.

Table 7-1


Mean and normalized bias as a function of classification bin for each wet deposition species at the 14 monitoring stations in the 12-km grid. Low

Middle

Species

MBaa

NMBb

Sulfate

14.00

72.0

Nitrate Ammonium Hydrogen Ion Calcium Magnesium Precipitation (mm)

4.57 10.36 -0.05 1.58 0.55 2.71

35.8 632.2 -13.4 232.4 431.5 40.6

a

Mean bias (µg m-3) calculated using Equation 7-2

b

Normalized mean bias (%) calculated using Equations 7-1 and 7-4

High NMB Bb

MBaa

NMBb

0.79

1.1

-49.13

-27.6

-1.70 12.56 -0.76 1.07 0.87 -3.28

-4.1 169.7 -50.2 39.4 209.0 -10.0

-26.61 4.03 -2.70 -1.19 0.36 -25.52

-28.7 17.7 -69.9 -17.3 32.0 -26.7

MBaa

variables, ambient levels of ozone, sulfate and nitrate, concentrations of sulfate and nitrate in precipitation, and the deposition of sulfate and nitrate. It is difficult to conduct a comparison of the two model evaluations because different metrics were chosen to conduct the evaluations due, in part, to the data sets being used and the study objectives. Further, they used a continuous period during the summer, not a variety of episodes through a number of seasons and years. To the degree that they can be compared, a description of the results follows. Like here, when compared to the EMEFS data, they found a tendency to under-predict the highest, peak ozone concentrations and over predict the lowest peak ozone concentrations. They also found a consistent bias in over-predicting the low ozone concentrations. Their interest in correctly predicting ozone levels, however, was driven by using it as an indicator that the atmospheric chemistry was being reliably simulated. As part of EMEFS, measurements of SO2 and sulfate were made. In the NAPAP evaluation, they compared the 33-day averages as well as daily concentrations. For SO2, which was examined to a limited amount in SAMI, they found a significant high bias across all percentiles of observed concentrations. For sulfate, they were biased low across all percentiles of the observations, with an average deviation of about 40%. They concluded that RADM was unable of accurately estimating the annual average sulfate concentrations, though, again, their interest was in simulating acid deposition. SAMI modeling had a smaller bias and error, with a very good correlation. Sulfate deposition was also evaluated as part of NAPAP, though in a less direct way. Here, they used 30, three-day, meteorological simulations for 1982 to 1985 to drive the model, but they used observed deposition data from 1987 and 1988 for comparison. Errors, which were broken down into three groups, by season, depending on concentration, ranged from 26 to 173% when averaged across sites. They do not provide evaluation data for individual sites, and cannot do an episode-by-episode comparison. More often then not, the simulations appear to overestimate sulfate deposition, except for the highest third of the measurements in the spring and summer. The SAMI modeling was not as consistently biased. Finally, they also looked at nitrate wet deposition in a similar fashion as they did sulfate wet deposition, that is aggregating over 30, three-day samples. For spring and fall, they found good agreement between the simulated and observed rates. They under-estimated the rates in the summer, and the winter found significant scatter between observations and simulated values. A quantitative tabulation was not given. SAMI found a slightly high bias in the simulations, in part likely due to the boundary conditions used. The comparison of the simulated nitric acid to the observed, by percentile, showed a consistent over prediction of about 20-50%. 7-11

ODMAN,

ET AL.


Because of the differences in the intent of the studies, and the availability of data sets, NAPAP and SAMI did not perform the same level of model evaluation. However, the SAMI modeling has better aerosol model performance for sulfate (other species were not tested as part of NAPAP), and the deposition results appear to be as good or better. The ozone predictions from the SAMI study appear to be more accurate, though the model set-up for NAPAP was not intended to capture the smaller scale features that can significantly impact ozone concentrations.

7-12

ODMAN,

8

ET AL.


SEASONAL AND ANNUAL AIR QUALITY

8.1 Calculation of Seasonal and Annual Metrics In contrast to many air quality modeling efforts, SAMI’s primary interest is in assessing the impact of emission control scenarios on visibility and the ecosystems (aquatic and terrestrial) in the Southern Appalachian Mountains. Specific focus is placed on the air quality and air quality values in the national parks and wilderness areas designated as Class I. This requires consideration of longer term air quality metrics, such as the response of seasonal cumulative ozone, annual average PM2.5 (by species), and annual average acid deposition to emission controls. Given the choice of modeling several continuous years (one year likely will not be representative), or selecting a series of episodes specifically chosen to characterize longer term meteorology and air quality, SAMI opted for the latter as being more tractable. Data classification and optimization techniques (Deuel and Douglas, 1998) were used to select nine episodes, each 6 to 9 days long (plus 2 ramp-up days), between the years 1991 and 1995. As part of the episode selection process, each day (or week in the case of acid deposition) is assigned a statistically-defined category or class depending on the observed pollutant levels. For SAMI, the classes were defined at two national parks, Great Smoky Mountains (GRSM) in Tennessee/North Carolina and Shenandoah (SHEN) in Virginia. SAMI categorized individual days into one of four ozone classes. The primary classifying variable was the observed daily cumulative ozone exposure using the W126 metric (Lefohn and Runeckles, 1987) to calculate a weighted sum of hourly ozone concentrations. Ozone W126 was selected because it can be used to evaluate the ozone effects on forests and vegetation. Only the days during the ozone season (April-October) were considered. Acid deposition class (1 through 4) was defined as one of four levels representing the observed sum of 2+ 2+ selected cations (Ca and Mg ) and anions (sulfate and nitrate) in weekly precipitation. Visibility class (1 through 5) was defined as one of five levels representing the measured daily total fine particulate mass (sulfate, nitrate, organics, elemental carbon and soils) and not in terms of visibility metrics such as extinction coefficients or deciviews. Note that wet deposition class (1 through 4) was defined as one of four levels representing the observed sum of selected cations (calcium and magnesium) and anions (sulfate and nitrate) in weekly precipitation. Dry deposition classes were assigned after the episodes were selected. Wet and dry deposition classes were based on weekly monitoring data because daily measurements were not available. In each case, class number increased with the severity of pollutant levels with Class 1 days being the least polluted and class 4 or 5 days being the most polluted. Table 8-1 shows the percentage of days that are represented by each class over a season or year. These classes were then used to select a set of multi-day episodes to represent the full spectrum of ozone, deposition and visibility conditions for modeling. The nine episodes listed in Table 4-1 were chosen for detailed modeling. A general description of the severity of the ozone, PM, and acid deposition levels is also included for each episode that will be used to calculate the corresponding seasonal and annual air quality metrics. This modeling approach was designed to provide insight into the atmospheric response to different emission strategies under a variety of conditions. It was also meant to provide a framework for scaling episodic model results up to seasonal and annual impacts.

Table 8-1 Species Ozone Wet Deposition Dry Deposition PM

Percent of days falling into each class based on severity of pollutant levels (Deuel and Douglas, 1998) Class 1

Class 2

Class 3

Class 4

Class 5

70% 70% 70% 20%

20% 20% 20% 30%

7% 7% 7% 30%

3% 3% 3% 17%

N/A N/A N/A 3%

8-1

ODMAN,

ET AL.


Table 8-2 lists the ozone and visibility classes for modeled days and gives the percentage contribution (weight) to the seasonal ozone and annual visibility metrics at the Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks. Table 8-3 lists the weeklong periods for which classes were assigned and the percentage contribution to the annual wet and dry deposition metrics at GRSM and SHEN. Dry deposition was not optimized during the episode selection process. Instead it was classified after the fact leading to episodes that under represent the amount of dry deposition contributing to the annual metric. To calculate a seasonal or annual average metrics at each site, the following relation was used:

N

maverage = ∑ i =1

wi m i 100

(8-1)

where, maverage is the simulated seasonal or annual average pollutant value, N is the number of weighted periods (days or weeks) contributing to the metric, wi is the percent contribution of the period i to the metric, and mi is the simulated daily or weekly pollutant value. It should be noted that these averages are composites based on weighting of N events taken between 1991 and 1995. They are surrogates for a seasonal or annual average, but are not real in the sense of a true averaging of concentrations simulated throughout an entire season or a year. Metrics were also calculated for certain periods of the season or year such as a single class or a combination of classes. In that case the following formula was used.

N class

m

class average

=

∑wm i =1

i

i

(8-2)

N class

∑w

i

i =1

class

where N is the number of periods (days or weeks) in a certain class or group of classes. Note that the sum of the percent contributions of days or periods in a single class or a group of classes does not add up to 100% hence the difference from Equation 8-1. Annual weights are only defined at two sites (GRSM and SHEN). However, these weights are applied to calculate seasonal ozone, annual average PM, and deposition at other Class I areas by assuming that sites south of the Virginia-North Carolina and Kentucky-Tennessee borders are represented by the weights at GRSM and sites north of these borders are represented by the SHEN weights.

8-2

ODMAN,

ET AL.

Table 8-2


Ozone and PM classes and their contribution (weight) to the seasonal cumulative ozone (W126) and annual average visibility metrics at Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks. Ozone

Visibility

GRSM

SHEN

GRSM

SHEN

Date (MM/DD/YY)

Class

Weight (%)

Class

Weight (%)

Class

Weight (%)

Class

Weight (%)

07/23/91

–

–

3

0.17

–

–

–

–

07/24/91 07/25/91

– –

– –

– –

– –

– –

– –

4 –

2.10 –

07/26/91 07/27/91

3 2

1.46 0.77

2 –

1.19 –

– 5

– 1.39

– –

– –

07/28/91 07/29/91

1 –

4.00 –

2 –

1.19 –

– –

– –

– –

– –

07/30/91 07/31/91

– 2

– 0.77

1 2

5.85 1.19

– 5

– 0.38

– 4

– 2.19

06/24/92 06/25/92

4 –

1.13 –

2 1

1.19 6.10

4 –

4.03 –

4 –

0.79 –

06/26/92 06/27/92

1 –

2.87 –

2 –

1.19 –

– –

– –

– 4

– 0.79

06/28/92 06/29/92 03/23/93

3 – –

1.27 – –

2 3 –

1.19 2.30 –

– – –

– – –

– – –

– – –

03/24/93 03/25/93

– –

– –

– –

– –

2 –

26.41 –

2 –

1.37 –

03/26/93 03/27/93

– –

– –

– –

– –

– 1

– 10.39

– –

– –

03/28/93 03/29/93

– –

– –

– –

– –

– –

– –

– –

– –

03/30/93 03/31/93

– –

– –

– –

– –

– –

– –

– 1

– 17.90

05/11/93 05/12/93

2 –

0.84 –

3 3

0.17 2.02

– –

– –

– 4

– 2.55

05/13/93 05/14/93

1 –

10.66 –

1 –

5.85 –

– –

– –

– –

– –

05/15/93 05/16/93 05/17/93

2 2 2

3.07 0.84 0.52

– – –

– – –

3 – –

6.99 – –

3 – –

4.50 – –

08/03/93 08/04/93

– –

– –

– –

– –

– 3

– 4.60

– 3

– 2.95

08/05/93 08/06/93

2 1

3.07 10.66

1 1

6.10 5.85

– –

– –

– –

– –

08/07/93 08/08/93

1 2

4.00 1.58

– –

– –

3 –

4.60 –

4 –

2.55 –

08/09/93 08/10/93

2 –

0.77 –

2 2

1.19 1.19

– –

– –

– –

– –

08/11/93 02/08/94

2 –

3.07 –

2 –

1.19 –

4 –

4.03 –

4 –

1.58 –

02/09/94 02/10/94

– –

– –

– –

– –

1 –

14.98 –

1 –

17.90 –

02/11/94 02/12/94

– –

– –

– –

– –

– –

– –

– –

– –

8-3

ODMAN,

ET AL.

Table 8-2


(Continued) Ozone and PM classes and their contribution (weight) to the seasonal cumulative ozone (W126) and annual average visibility metrics at Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks. Ozone

Visibility

GRSM

SHEN

GRSM

SHEN

2

Date (MM/DD/YY) 04/28/95

Class

Class 1

Weight (%) 5.45

Class

–

Weight (%) –

–

Weight (%) –

–

Weight (%) –

04/29/95 04/30/95 05/01/95

3 – 1

0.64 – 10.66

2 – 1

3.77 – 5.85

3 – –

9.16 – –

2 – –

10.14 – –

05/02/95 05/03/95

1 2

10.66 1.58

1 1

5.85 5.85

– –

– –

– 2

– 10.14

05/24/95 05/25/95

3 3

1.27 0.64

3 2

2.02 1.19

– –

– –

4 –

2.10 –

05/26/95 05/27/95

3 2

0.37 1.58

– –

– –

– 4

– 4.03

– 3

– 3.26

05/28/95 05/29/95

– 1

– 10.66

1 1

5.85 5.85

– –

– –

– –

– –

07/11/95 07/12/95

3 3

0.52 1.27

3 3

2.02 0.67

– 4

– 4.03

– 5

– 2.05

07/13/95 07/14/95

3 4

1.27 1.57

4 4

0.62 1.76

– –

– –

– –

– –

07/15/95 07/16/95

3 –

0.37 –

4 –

1.01 –

5 –

1.39 –

5 –

2.05 –

07/17/95 07/18/95 07/19/95

3 – 4

0.92 – 1.57

2 2 2

2.00 1.19 3.62

– – –

– – –

– – 3

– – 2.95

Table 8-3

3.59

2

10.14

Class

Wet and dry deposition classes and their contributions (weight) to the annual metrics at Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks. Wet Deposition GRSM

Dry Deposition SHEN

GRSM

SHEN

Period (MM/DD/YY)

Class

Weight (%)

Class

Weight (%)

Class

Weight (%)

Class

Weight (%)

07/23/91 – 07/30/91

4

3.81

4

2.00

1

14.70

1

13.97

06/23/92 – 06/30/92 03/23/93 – 03/30/93 05/11/93 – 05/18/93

2 2 3

18.36 9.13 4.46

1 3 4

30.79 13.88 2.00

– 1 –

– 14.67 –

1 1 2

13.97 8.89 6.00

08/03/93 – 08/10/93 02/08/94 – 02/15/94

3 2

4.46 9.13

– 2

– 4.75

1 1

14.67 11.86

1 1

13.90 8.94

04/25/95 – 05/01/95 05/23/95 – 05/30/95

1 1

14.61 36.04

2 2

11.05 4.75

1 1

14.67 14.72

– 1

– 13.95

07/11/95 – 07/18/95

–

–

1

30.79

1

14.70

2

20.37

8-4

ODMAN,

ET AL.


8.2 Model Performance in Terms of Seasonal and Annual Metrics Ozone W126 is used to represent the cumulative seasonal exposure to ozone. The W126 exposure index was selected to characterize ozone trends and relate vegetation yield reduction to ozone exposure. The daily cumulative ozone W126 is calculated using Equation 8-3 and the sigmoidally weighted function, fi, given in Equation 8-4 (Lefohn and Runeckles, 1987):

24

W 126 = ∑ c i f i

(8-3)

1 1 + Me − Aci

(8-4)

i =1

fi =

-1

where, ci is the hourly ozone concentration in ppm, M is 4403, and A is 126 ppm . The W126 index focuses on the higher hourly average concentrations, while retaining the mid- and lower-level values. Equation 8-1 was used to reconstruct the cumulative seasonal ozone W126 using model estimated ozone concentrations for each weighted day at GRSM and SHEN. Multiplying the daily average cumulative reconstructed ozone W126 values by the number of days in the ozone season (214) results in the seasonal cumulative reconstructed ozone W126. The modeled and observed seasonal cumulative reconstructed ozone W126 are compared at GRSM (Look Rock—elevation of 793 m) and SHEN (Big Meadows—elevation of 1073 m) in Figure 8-1. The ozone observations were taken from the Aerometric Information Retrieval System (AIRS) database (USEPA, 2001a). The model underestimates the seasonal W126 at GRSM and SHEN by 39% and 50%, respectively. This is due to

Ozone W126 (ppm-hrs)

Seasonal Ozone W126 80.0

60.0

40.0

20.0

0.0

Great Smoky Mountains Observed Figure 8-1

Shenandoah Modeled

Comparison of modeled and observed seasonal ozone W126 at Great Smoky Mountains (Look Rock) and Shenandoah (Big Meadows) National Parks. 8-5

ODMAN,

ET AL.


the underestimation of ozone peaks that carry the greatest weight in calculating the seasonal ozone W126. Figure 8-2 shows a comparison of the annual averaged observations and model simulated 2+ concentrations of fine sulfate (SO4 ), nitrate (NO3 ), ammonium (NH4 ), organics (OC), elemental carbon (EC), and soils at GRSM (Look Rock) and SHEN (Big Meadows) by using Equation 8-1 to weight each episode day. Speciated daily average observations were extracted from the Interagency Monitoring of Protected Visual Environments (IMPROVE) network (NPS, 2000). Annual average sulfate and ammonium are underestimated by approximately 17% at GRSM, but show little bias at SHEN. Nitrates and soils are overestimated at both sites. Model estimates for organic and elemental carbon are within 7% of the observed values at GRSM and SHEN. A comparison of the annual averaged wet deposition using weekly observations from the 2National Atmospheric Deposition Program (NADP) and model estimated mass fluxes for sulfate (SO4 + 2+ 2+ ), nitrate (NO3 ), ammonium (NH4 ), calcium (Ca ), and magnesium (Mg ) at GRSM (Elkmont— elevation of 640 m) and SHEN (Big Meadows—elevation of 1074 m) are shown in Figure 8-1. Sulfate and nitrate wet deposition are overestimated at GRSM by 86% and 95%, respectively. However, sulfate and nitrate wet depositions are underestimated at SHEN by 5% and 35%, respectively. Ammonium is overestimated by approximately 600% at GRSM and 140% at SHEN. Calcium and magnesium are overestimated at both sites, but the deposition mass fluxes are small compared to the other species.

8.3 Sources of Modeling Uncertainty There are many known sources of error in running the RAMS/EMS-95/URM-1ATM atmospheric modeling system. First, there are a number of errors associated with the meteorological inputs in general. Some of the more important meteorological parameters, such as temperatures, wind speed, and wind direction can bias the simulated pollutant concentrations and skew the source/receptor relationships. However, the most difficult meteorological parameters to simulate were

Concentration (µg/m 3)

Annual Average Aerosol Concentrations 14.0 12.0

GRSM

SHEN

10.0

Soils EC OC NH4 NO3 SO4

8.0 6.0 4.0 2.0 0.0 MODEL

Figure 8-2

IMPROVE

MODEL

IMPROVE

Comparison of modeled and observed annual averaged fine PM concentrations at Great Smoky Mountains (GRSM—Look Rock) and Shenandoah (SHEN—Big Meadows) National Parks. 8-6

ODMAN,

ET AL.



Annual Average Wet Deposition Mass Fluxes 200.0

GRSM 150.0

Mg Ca

100.0

NH4 NO3

50.0

SO4

0.0

MODEL Figure 8-1

SHEN

NADP

MODEL

NADP

Comparison of modeled and observed annual averaged wet deposition mass fluxes at Great Smoky Mountains (GRSM—Elkmont) and Shenandoah (SHEN—Big Meadows) National Parks.

precipitation and the extent and physical characteristics of the clouds. Errors in the magnitude and location of precipitation directly affect the model’s ability to estimate wet deposition and the bias in scavenging can impact deposition and pollutant concentrations downwind. Also, the meteorological model was run ignoring ice and snow microphysics, which could potentially affect the pollutant concentrations in two winter episodes. The uncertainties in the emission inventories of some gas and particulate species are well acknowledged. No attempts were made here to quantify these uncertainties, but related studies show that the uncertainties in emissions may be significant for some species (Mendoza-Dominguez and Russell, 2001). Some of the more uncertain emissions include ammonia and PM emissions of organic carbon, elemental carbon, and crustal minerals. Biogenic emissions, especially for transition seasons when vegetation is changing, are highly uncertain. There are also uncertainties in the amount of secondary organic PM formed by using fixed organic aerosol yields. This method condenses speciated oxidized gas phase organic compounds to the particle phase at a fixed rate without considering the amount of existing organic PM and does not allow volatilization from the particle phase to the gas phase. The work of Odum et al. (1996) and Saxena and Hildemann (1995) suggests that organic aerosols partition between the gas and particle phase based on the existing organic PM mass concentrations, though to accurately do such partitioning requires detailed knowledge as to the composition of the organic species present and their specific water solubility. Here, a lumped condensible organic specie is used. In evaluating the performance of the air quality model, certain assumptions had to be made in order to compare the model result to observations. Since there were no IMPROVE measurements of ammonium for this period, an assumption was made about the form of the particulate ammonium in order to estimate the ammonium concentrations. Whether the ammonium is in the form of ammonium sulfate, ammonium bisulfate, or other ammonium salts will bias the comparisons to the model. Also, the assumption that the measured total organic mass is 1.4 times the measured organic carbon mass could bias the results. It has been suggested by Turpin and Lim (2001) that this factor could be as high as 1.6 in the urban areas and 2.1 in the rural areas. Finally, there are uncertainties in evaluating nitrate PM performance due to the difficulties discussed earlier associated with accurately capturing nitrate on 8-7

ODMAN,

ET AL.


a filter pack. The measurements of PM and deposition fluxes were daily and weekly, respectively. Continuous measurements are needed (hourly or more frequent) to evaluate the diurnal variation in model performance and to improve process representations. The scant monitoring network for PM and some gas-phase species can be a source of model uncertainty. Observations are used to set the initial and boundary conditions (IC/BCs) for many gas and particulate species. However, because of the minimal amount of data, defining the IC/BCs for every grid cell in the domain becomes difficult, especially since monitors are usually placed in areas where higher concentrations are expected. Furthermore, there are even fewer observations to help determine the vertical profiles of the IC/BCs. Finally, there are model uncertainties caused by a lack of grid resolution. Coarse horizontal grid resolution over complex terrain is generally not adequate to capture some very important processes. Also, there is uncertainty in evaluating a grid-averaged modeled concentration against a pollutant concentration that is varying on a sub-grid scale. There are additional errors caused by not modeling large point-source plumes at sub-grid scale. This can result in plumes being diluted too fast and the chemistry pushed into a different chemical regime. As a consequence, the spatial relationship between pollutant sources and downwind pollutant concentrations can be affected. Similarly, the vertical resolution is probably too coarse to accurately capture the depth of the mixed layer during the day, nocturnal jets, convective mixing, and cloud scavenging. This list is certainly not exhaustive, but addresses some of the major issues SAMI has had to face regarding sources of uncertainty in its modeling.

8-8

ODMAN,

9

ET AL.


MODEL INPUTS FOR FUTURE YEAR SIMULATIONS

9.1 Meteorological Inputs The meteorological inputs for future year simulations are identical to those used for the basecase simulations. The basic assumption is that the nine simulated episodes characterize a typical or mean year between 1991 and 1995, and can be used as typifying future mean years. Clearly, the assumption relies on the climatological representativeness of the simulated episodes. Also, it ignores any potential feedback of future landuse and emissions on the meteorology, which would be inevitable at least on local scales. A detailed description of how the meteorological inputs were prepared can be found in Doty et al, 2001. A brief summary of how these inputs compare with the observations is provided in Section 4.3 above.

9.2 Emission Inputs 9.2.1

Overview of Future Year Emission Strategies

Three emission strategies were developed by SAMI’s Policy Committee to reflect their assumptions about future growth in the demand for goods and services and the implementation of regulations and incentives (SAMI, 2001). Each strategy proposes progressively more stringent emission controls in each of five major source categories: utility, industrial, highway vehicle, non-road engines, and area sources for the years 2010 and 2040. From least stringent to most stringent, the three strategies are “on the way” (OTW), “bold with constraints” (BWC), and “beyond bold” (BB). These names are not meant to provide a full strategy description, but to roughly indicate the relative stringency of a strategy. Strategy results for 2040 are highly dependent on the assumptions about the future decisions that businesses, government, and individuals may make. The purpose of the strategies is to approximate both the upper and lower boundaries of likely future emissions under SAMI’s assumptions. Each strategy used in this modeling exercise is not necessarily being considered as an option for policy recommendation. A brief description of the emission control assumptions for each strategy follows. For a more complete description, refer to Pechan (2001). The OTW strategy assumes reductions of volatile organic compounds (VOCs) and oxides of nitrogen (NOx ) emissions from the laws and regulations mandated by the Clean Air Act (CAA) as amended in 1977 and 1990 to comply with the 1-hour ozone standard; reductions of SO2 and NOx from utility sources under Title IV of the CAA amendments; and reductions of NOx and VOCs from mobile sources under Tier I tailpipe standards and fuel rules. In addition, this strategy assumes emission reductions from several recently promulgated regulations: regional NOx reductions which will be included in “State Implementation Plans” to reduce ozone (USEPA, 1998); NOx and VOC reductions resulting from implementation of Tier II and low sulfur rules (USEPA, 2001b); and VOC reductions resulting from Maximum Achievable Control Technology (MACT) standards (USEPA, 1990). OTW does not include the emissions reductions that might be required for the 8-hour ozone National Ambient Air Quality Standards (NAAQS), the new PM2.5 NAAQS, or the regional haze rule. OTW is applied for all the eastern United States and is the reference strategy against which the two other strategies will be evaluated. The BWC strategy simulates emission reductions from the OTW reference strategy’s inventory, plus state-of-the-art emission controls applied to all sources as soon as technologically feasible, and “off the shelf” controls for 2010 and existing prototypes for 2040. It also includes logistical constraints to the implementation of emissions controls. BB assumes emission reductions from the OTW reference strategy’s inventory, plus the most advanced existing and evolving technologies applied to all sources for 2010 and 2040. This strategy is intended to approximate an upper bound of 9-1

ODMAN,

ET AL.


emission reductions (given the current assumptions) without consideration of economic or technical feasibility. BWC and BB emissions reductions are applied only within the eight SAMI states; emissions in the rest of the eastern U.S. are assumed to remain the same as the OTW strategy. The SAMI emission inventories include ammonia (NH3), carbon monoxide (CO), SO2, NOx , VOCs, and speciated particles less than 10 and less than 2.5 micrometers in diameter (PM10 and PM2.5). Annual inventories were developed for each of the SAMI emission reduction strategies for 1990, 2010 and 2040, as well as the nine air quality modeling episodes in 1991-1995. All three emission reduction strategies project that SO2 and NOx emissions will decrease in 2010 and 2040; however, VOCs, fine particles, and NH3 show both increases and decreases in emissions depending on the specific strategy and year (Figure 9-1). A more detailed discussion of SO2 and NOx emissions follows. Annual SO2 emission rates in the eight SAMI states in 1990 and projected annual SO2 emission rates in 2010 and 2040 under the three SAMI strategies are illustrated in Figure 9-2. In the SAMI states, annual SO2 emissions are projected to decrease by 23% for the 2010 OTW strategy compared to the 1990 levels. Annual SO2 emissions in 2010 would be reduced by 49% and 86% under the BWC and BB strategies, respectively. The majority of the projected SO2 reductions in 2010 are attributed to the addition of scrubbers to coal-fired generating units. Utility SO2 emissions reductions in 2040 assume retirement and replacement or repowering with cleaner technologies, or adding emissions reduction equipment for most of the existing coal-fired power plants.

Million Tons/Year

Annual NOx emissions in the eight SAMI states are projected to be 20% lower in 2010 than in 1990 under the OTW reference strategy (Figure 9-3). Emissions from utilities and highway vehicles will be reduced in response to federal regulations, while emissions from non-road engines and area sources are projected to increase slightly. Under the OTW strategy, annual NOx emissions in the eight SAMI states are projected to be 3.3 million tons in 2010. Annual NOx emissions would be reduced by 45% and 72% under BWC and BB strategies, respectively. These reductions would be accomplished by a combination of advanced controls on utility and industrial sources, as well as cleaner fuels and improved engines in the highway vehicle, non-road engine, and area source sectors.

7.0

Emissions in the SAMI States

6.0 5.0 4.0 3.0 2.0 1.0 0.0

SO2 1990 Figure 9-1

2010 OTW

NOx 2010 BWC

VOC 2010 BB

PM2.5

2040 OTW

2040 BWC

NH3 2040 BB

Emissions from the eight SAMI states in 1990 and projections for 2010 and 2040 under the OTW, BWC, and BB strategies. VOC emissions reported here are anthropogenic and do not include biogenic emissions. 9-2

ODMAN,

ET AL.


Million Tons/Year

SO2 Emissions in the SAMI States 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0

1990

2010

Base OTW BWC Industrial Figure 9-2

Utility

Nonroad

2040

BB

OTW BWC

Highway Vehicle

BB Area

SO 2 emissions in 1990 and projected for 2010 and 2040 under the OTW, BWC, and BB strategies.

Million Tons/Year

NOx Emissions in the SAMI States 5.0

1990

2010

4.0 3.0 2.0 1.0 0.0 Base OTW BWC

Industrial Figure 9-3

2040

Utility

Nonroad

BB

OTW BWC

Highway Vehicle

NOx emissions in 1990 and projected for 2010 and 2040 under the OTW, BWC, and BB strategies.

9-3

BB Area

ODMAN,

ET AL.


Thousand Tons/Day

NOx Emissions in the SAMI States 16.0 14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0

1990

2010

Base OTW BWC Industrial Figure 9-4

Utility

Nonroad

2040

BB

OTW BWC

Highway Vehicle

BB Area

Summer day NOx emissions in 1990 and projected for 2010 and 2040 under the OTW, BWC, and BB strategies.

Ozone regulations require larger NOx emissions reductions during the “ozone season” (MaySeptember), when ozone levels are typically highest. Average summer-day NOx emissions in the eight SAMI states will be reduced by 44% in 2010 for the OTW strategy, compared to 1990 (Figure 9-4). Average summer day NOx emissions in the 2010 OTW strategy are projected to be 7.3 thousand tons per day. Summer day NOx emissions will be reduced by 52% and 75% for the BWC and BB strategies, respectively. For a more detailed discussion on the emissions developed for the SAMI episodes, refer to Pechan (2001). Figure 9-5 compares NOx emissions from the 8 SAMI states to the total NOx emissions in the modeling domain. Domain wide annual NOx emissions are projected to decrease between 24% (2010OTW) and 48% (2040-BB) from the base year (1990). NOx emissions from the SAMI states contribute between 23% (1990, 2010-OTW, and 2040-OTW) and 7% (2040-BB) to the domain wide NOx emissions; the majority of the NOx emissions are from outside the SAMI states. Figure 9-6 contains total emissions of nitrogen for each of the SAMI strategies. Total annual nitrogen emissions are calculated by summing the nitrogen contribution from the NOx and NH3 emissions. The NOx emissions are assumed to consist of 90% NO and 10% NO2. The majority of the nitrogen emissions are from outside the SAMI states; SAMI states contribute between 20% (1990, 2010-OTW, and 2040-OTW) and 5% (2040-BB) to the domain wide nitrogen emission totals. Note, the domain wide annual nitrogen emissions do not decrease as much as the domain wide annual NOx emissions due to increased NH3 emissions outside the SAMI states for all emission control strategies and inside the SAMI states for the OTW and BWC strategies. The contribution of ammonia emissions to the total domain wide emissions of nitrogen are approximately 25% for the base year (1990), 40% for the 2010 strategies, and 50% for the 2040 strategies. The reductions in domain wide annual nitrogen emissions from the base year are between 6% (2040-OTW) and 20% (2010-BB and 2040-BB).

9-4

ODMAN,

ET AL.


Million Tons/Year

Annual NOx Emissions 24.0

1990

2010

16.0 12.0 8.0 4.0 0.0

Base

A2

Domain Wide Figure 9-5

2040

20.0

B1

B3

A2

30 Non-SAMI States

B1

B3

SAMI States

NOx emissions in 1990 and projected for 2010 and 2040 under the OTW (A2), BWC (B1), and BB (B3) strategies.

Million Tons/Year

Annual Nitrogen Emissions 14.0 12.0

1990

2010

10.0 8.0 6.0 4.0 2.0 0.0

Base

A2

Domain Wide Figure 9-6

2040

B1

B3

A2

30 Non-SAMI States

B1

B3

SAMI States

Nitrogen emissions in 1990 and projected for 2010 and 2040 under the OTW (A2), BWC (B1), and BB (B3) strategies

9-5

ODMAN,

ET AL.

9.2.2


Emission Summary Tables

The total emissions were calculated both for the entire domain and for the 8 SAMI states. Emissions from different source categories were tabulated. Table 9-1 shows the 8 SAMI state total for an average day during the July-95 episode under the OTW strategy in the year 2010. Table 9-2 and Table 9-3 show the same totals for 2010 under the BWC and BB strategies. Finally, Table 9-4, Table 9-5, and Table 9-6 show the category totals for the year 2040 under the OTW, BWC and BB strategies, respectively. Similar tables for the other episodes can be found in Appendix A. Table 9-1

Daily average total 2010-OTW emissions from the 8 SAMI states for the July-95 episode (in tons per day).

Source Category

VOCs

NOX

CO

SO2

PM

PM2.5 2.5

NH3

Area

3,922

726

3,645

1,250

11,972

2,564

1,736

Biogenic Electric Generation Off-road mobile

54,069 29 1,193

520 1,128 1,546

0 249 12,607

0 8,423 406

0 348 489

0 97 227

0 0 2

On-road mobile Point Daily total

1,729 1,851 62,792

1,812 1,318 7,050

16,776 3,193 36,471

73 1,821 11,973

141 771 13,722

55 364 3,307

225 80 2,043

Table 9-2

Daily average total 2010-BWC emissions from the 8 SAMI states for the July-95 episode (in tons per day).

Source Category

VOCs

CO

SO2

PM

PM2.5 2.5

NH3

Area Biogenic

3,759 54,069

647 520

3,114 0

867 0

11,052 0

2,334 0

1,734 0

Electric Generation Off-road mobile On-road mobile Point

29 1,126 1,652 1,851

1,127 1,463 1,453 1,228

253 11,788 16,411 3,193

5,401 202 24 1,377

354 462 137 771

98 214 54 364

0 2 220 80

Daily total

62,486

6,439

34,759

7,871

12,776

3,064

2,037

Table 9-3

NOX

Daily average total 2010-BB emissions from the 8 SAMI states for the July-95 episode (in tons per day).

Source Category

VOCs

NOX

CO

SO2

PM

PM2.5 2.5

NH3

Area

1,274

197

1,548

340

6,319

1,269

472


54,069 29 874

520 572 810

0 253 9,009

0 892 142

0 354 346

0 98 160

0 0 1


1,178 1,851 59,274

655 863 3,617

11,771 3,193 25,774

16 912 2,301

77 771 7,866

28 364 1,920

161 80 714

9-6

ODMAN,

ET AL.

Table 9-4


Daily average total 2040-OTW emissions from the 8 SAMI states for the July-95 episode (in tons per day).

Source Category

VOCs

NOX

CO

SO2

PM

PM2.5 2.5

NH3

Area

5,382

814

4,165

1,276

14,634

3,129

2,270


54,069 43 1,308

520 1,012 1,319

0 209 16,726

0 2,609 594

0 277 471

0 79 216

0 0 2


2,569 2,231 65,602

1,810 1,164 6,639

26,802 3,523 51,425

123 1,243 5,846

219 662 16,263

85 320 3,829

349 105 2,725

Table 9-5

Daily average total 2040-BWC emissions from the 8 SAMI states for the July-95 episode (in tons per day).

Source Category

VOCs

CO

SO2

PM

PM2.5 2.5

NH3

Area Biogenic

3,758 54,069

480 520

2,854 0

59 0

12,326 0

2,554 0

1,197 0

Electric Generation Off-road mobile On-road mobile

53 1,087 1,256

1,085 912 534

296 13,683 12,370

1,637 316 19

286 405 62

82 185 21

0 2 165

Point Daily total

2,809 63,032

1,188 4,719

4,079 33,282

1,205 3,237

826 13,905

408 3,251

114 1,478

Table 9-6

NOX

Daily average total 2040-BB emissions from the 8 SAMI states for the July-95 episode (in tons per day).

Source Category

VOCs

NOX

CO

SO2

PM

PM2.5 2.5

NH3

Area

1,085

116

1,488

50

7,258

1,433

285

Biogenic Electric Generation

54,069 58

520 228

0 288

0 144

0 256

0 72

0 0

Off-road mobile On-road mobile Point Daily total

463 184 2,809 58,667

443 22 927 2,257

4,915 107 4,079 10,878

162 2 1,019 1,376

227 4 826 8,572

103 2 408 2,018

1 1 114 401

9.2.3 Spatial Plots for Selected Species and Days After gridding of strategy emissions, spatial plots (maps) were prepared for various categories over a portion of the domain covering the SAMI region. Figure 9-1 shows the ground-level NOx emissions for July 15, 2010 under the OTW strategy. The baseyear (i.e., 1995) emissions for the same day can be found in Figure 4-4. To facilitate the comparison, the difference (i.e., 2010-OTW minus 1995) is also mapped in Figure 9-2. A negative number in this figure corresponds to a reduction in ground-level NOx emissions. Note that the emissions are not reduced uniformly over the SAMI region and there are several locations of increase. Maps for other modeled days, as well as maps of other categories (elevated NOx , ground-level and elevated SO2 and ammonium (NH3) can be found in Appendix A. Difference maps can also be found in Appendix A. The differences are calculated from the baseyear for the 2010-OTW strategy, from 2010-OTW for the 2040-OTW strategy and the OTW strategy of the corresponding year for BWC and BB strategies. In equation form, these differences can be written as:

9-7

ODMAN,

ET AL.


Figure 9-1

Gridded daily total emissions of ground-level NOx over the SAMI region for July 15, 2010 under the OTW strategy.

Figure 9-2

Gridded change in daily total emissions of ground-level NOx for July 15, from 1995 to 2010 under the OTW strategy.

9-8

ODMAN,

ET AL.


2010-OTW change = (2010-OTW) – (basecase)

(9-1)

2010-BWC change = (2010-BWC) – (2010-OTW)

(9-2)

2010-BB change = (2010-BB) – (2010-OTW)

(9-3)

2040-OTW change = (2040-OTW) – (2010-OTW )

(9-4)

2040-BWC change = (2040-BWC) – (2040-OTW)

(9-5)

2040-BB change = (2040-BB) – (2040-OTW)

(9-6)

9.3 Initial and Boundary Conditions 9.3.1

Initial Conditions

Future changes in emissions will likely alter the background levels of precursors. Therefore, it is important to determine whether the future year simulations are sensitive to initial conditions used and, if so, what initial conditions should be used. To evaluate the sensitivity of future year results to the initial conditions, simulations were performed using the same initial conditions as the base year as well as using initial conditions changed in proportion with the emissions. Sensitivities of ozone, PM, and wet deposition to initial conditions are presented below. The initial conditions for SO2, NOx, particulate sulfate, and particulate ammonium were reduced for all future year simulations. The ICs were reduced because emission reduction in the modeling domain would likely result in lower initial conditions for these species. The reductions in ICs were accomplished by scaling the ICs proportionally to the reductions in emissions. SO2, sulfate, and ammonium ICs were reduced by the same percentage that the SO2 emissions were reduced. The NOx ICs were reduced by the same percentage that the NOx emissions were reduced. The percent reductions in initial conditions are shown in Table 9-1.

9.3.1.1 Sensitivity of Ozone to Initial NO x Concentrations Two simulations of the July-1995 episode were conducted using the 2010 emissions: first with the same initial conditions as 1995 and second with NOx initial concentrations reduced by 35% uniformly over the entire domain. The reduction amount (i.e., 35%) is the same as the domainwide reduction in NOx emissions from 1995 to 2010. The sensitivity of daily maximum ozone on July 11, 2010 to this change in initial NOx concentrations was analyzed. The reason for choosing July 11 is that, being the first day of the simulation (after the two ramp-up days), the sensitivity to initial conditions should be largest on this day. The daily maximum ozone difference between the simulations with reduced and unchanged (i.e., same as 1995) NOx initial concentrations are shown in Figure 9-1 (right panel). In the SAMI region, the changes in daily maximum ozone range from a decrease of 0.8 ppb to an increase of 0.3 ppb. Also shown is the change in daily maximum ozone from 1995 to 2010 (left panel). The magnitude of this change is between -29 ppb and zero. Thus, decreasing the initial NOx concentrations by the same amount as NOx emission reductions from 1995 to 2010 has a small effect on the daily maximum ozone concentrations. 9-9

ODMAN,

ET AL.

Table 9-1


Percent reductions in initial conditions of NOx, SO 2, sulfate and ammonium.

Episode and Control Scenario

NOx

SO2, Sulfate, and Ammonium

July_1995_2010_OTW

39.0%

32.0%

July_1995_2010_BWC

40.0%

39.0%

July_1995_2010_BB

45.0%

49.0%

July_1995_2040_OTW

49.0%

60.0%

July_1995_2040_BWC

50.0%

63.0%

July_1995_2040_BB

53.0%

67.0%

July_1991_2010_OTW

33.0%

40.0%

July_1991_2010_BWC

34.0%

46.0%

July_1991_2010_BB

39.0%

55.0%

July_1991_2040_OTW

42.0%

64.0%

July_1995_2040_BWC

45.0%

67.0%

July_1991_2040_BB

50.0%

70.0%

May_1995_2010_OTW

33.0%

17.0%

May_1995_2010_BWC

34.0%

25.0%

May_1995_2010_BB

41.0%

37.0%

May_1995_2040_OTW

32.0%

42.0%

May_1995_2040_BWC

43.0%

54.0%

May_1995_2040_BB

49.0%

58.0%

May_1993_2010_OTW

32.0%

30.0%

May_1993_2010_BWC

33.0%

37.0%

May_1993_2010_BB

40.0%

47.0%

May_1993_2040_OTW

39.0%

57.0%

May_1993_2040_BWC

42.0%

61.0%

May_1993_2040_BB

48.0%

65.0%

March_1993_2010_OTW

17.0%

27.0%

March_1993_2010_BWC

23.0%

35.0%

March_1993_2010_BB

32.0%

44.0%

March_1993_2040_OTW

35.0%

56.0%

March_1993_2040_BWC

40.0%

60.0%

March_1993_2040_BB

46.0%

64.0%

February_1994_2010_OTW

19.0%

25.0%

February_1994_2010_BWC

26.0%

35.0%

February_1994_2010_BB

30.0%

46.0%

February_1994_2040_OTW

36.0%

55.0%

February_1994_2040_BWC

40.0%

60.0%

February_1994_2040_BB

46.0%

63.0%

Daily average ozone on the same day was also analyzed. From 1995 to 2010, the daily average ozone concentrations are estimated to decrease in the SAMI region, except in Atlanta, GA and Birmingham, AL where nighttime ozone concentrations should increase as a result of reductions in NOx emission. In the region shown in Figure 9-1, the changes from 1995 to 2010 are estimated to range from -12ppb to +12 ppb. Decreasing the initial NOx concentrations increases this range by approximately 2% on either end. In summary, ozone estimates for the year 2010 are generally insensitive to the NOx initial concentrations used in the simulations.

9-10

ODMAN,

ET AL.

Figure 9-1


Change in daily maximum ozone for July 11 from 1995 to 2010 (left) and further change due to reducing NOx initial conditions by 35% (right).

9.3.1.2 Sensitivity of PM to Initial Conditions The change in daily average fine sulfate on July 11, 2010 when SO2 initial concentrations are reduced by 30% uniformly over the entire domain is shown in Figure 9-2 (right panel). The change normalized by 1995 sulfate concentrations, is a decrease between 0.1% and 6.4%. Also shown is the fractional change in daily average fine sulfate concentrations from 1995 to 2010 (left panel). The percent changes range from a 41% decrease to a 6% increase in the SAMI region. In an area over the Alabama-Georgia border, where the change from 1995 to 2010 is about -10% (yellow shaded on the left panel), reducing the initial SO2 concentrations by 30% may lead to a further 5% decrease (dark blue on the right panel) in fine sulfate. Thus, the estimated sulfate levels can be very sensitive to initial SO2

Figure 9-2

Fractional change in daily average SO42- for July 11 from 1995 to 2010 (left) and further change due to reducing SO 2 initial conditions by 30% (right). 9-11

ODMAN,

ET AL.


concentrations in certain locations. In general, the 30% reduced SO2 initial conditions increase the estimated reduction by another 10%. This sensitivity decreases on later days of the simulation. It should be noted that the 2-day ramp up period used here may not be sufficient to completely eliminate 2the effect of the SO2 initial conditions and, for this reason, the errors in SO4 -may be larger at the beginning of an episode. The effect of NOx initial concentrations on fine sulfate levels is negligible. Increasing the NH3 initial conditions from zero to 1 ppb can increase the estimated sulfate levels by as much as 5% but generally around 1%. This is not a large sensitivity compared to the sensitivity to SO2 initial conditions. The change in daily average fine nitrate on July 11, 2010 when NH3 initial concentrations are increased from zero to 1 ppb are shown in Figure 9-3 (right panel). The changes range from zero to an 3 increase of 0.8 µg/m in the SAMI region. Also shown in Figure 9-3 (left panel) is the change in daily average fine nitrate concentrations from 1995 to 2010. These changes range from a decrease of 1.1 3 3 µg/m to an increase of 0.8 µg/m (largest increases are expected in the Southern Appalachian Mountains). Thus, the estimated increase in nitrate levels can double if the initial NH3 concentrations are increased from zero to 1 ppb. The sensitivity of nitrate to initial SO2 concentrations is also large. In general, the 30% reduced SO2 initial conditions boost the estimated increase in nitrate by about 35%. The effect of NOx initial concentrations on fine nitrate levels is negligible compared to the sensitivities to NH3 and SO2 initial concentrations.

9.3.1.3 Sensitivity of Wet Deposition to Initial Conditions The fractional change in cumulative nitrate wet deposition for the week of July 11-18, 2010 when NOx initial concentrations are reduced by 35% is shown in Figure 9-1 (right panel). The changes normalized by the 1995 deposition fluxes range from -6% to +6% in the SAMI region. Also shown in Figure 9-1 (left panel) is the fractional change in weekly cumulative nitrate wet deposition from 1995 to 2010. These changes range from a decrease of 52% to an increase of 122%. The largest increases in nitrate wet deposition are expected in the Southern Appalachian Mountains. Reducing the initial NOx concentrations by 35% boosts the estimated increase approximately by 2%. In other words, cumulative nitrate wet deposition is not too sensitive to the initial concentrations of NOx .

Figure 9-3

Change in daily average NO3- for July 11 from 1995 to 2010 (left) and further change due to increasing NH3 initial conditions from zero to 1 ppb (right). 9-12

ODMAN,

ET AL.

Figure 9-1

9.3.2


Fractional change in cumulative NO3- wet deposition for the week of July 11-18 from 1995 to 2010 (left) and further change due to decreasing NOx initial conditions by 35% (right).

Boundary Conditions

Although it was obvious that the initial conditions should be reduced for the future year simulations, it was not so clear whether the boundary conditions should be reduced as well. To test the sensitivity of the pollutant levels to the boundary conditions, two test runs were made using the July 1995 episode with 2040-BWC emissions. The initial conditions for both runs were reduced proportionately to the reductions in domain-wide emissions of SO2 and NOx . This corresponded to IC 2+ reductions of approximately 50% for NOx and 63% for SO2, SO4 and NH4 . However, one test run kept the BCs the same as the basecase run, while the other reduced (scaled) the BCs by the same percentage that the ICs were reduced. Boundary condition sensitivity results for ozone, sulfate, nitrate and ammonium PM and wet depositions were compared. Figure 9-2 and Figure 9-3 show the sensitivity of ozone concentrations to boundary conditions at Look Rock, TN and Big Meadows, VA, respectively. The hourly ozone concentrations estimated for the year 2040 are compared to the baseyear for the July 11-19, 1995 episode. Two sets of 2040 estimates are shown: they both use the BWC emission strategy. However, one uses the same BCs as the baseyear (red line) and the other uses the scaled BCs (green line). Ozone concentrations decreased by another 5-8 ppb when the scaled BCs were used. This corresponds to an additional 2025% reduction in estimated 2040 ozone concentrations with respect to the baseyear levels.

9-13

ODMAN,

ET AL.


Look Rock (TN) - July 1995 Ozone (ppm)

0.090 0.080 0.070 0.060 0.050 0.040 0.030 0.020

11

12

13 O3_base

Figure 9-2

14

15

16

O3_2040_bwc

17

18

19

20

O3_2040_bwc_scaled_BC

Sensitivity of O3 at Look Rock, TN to boundary conditions during the July 1995 episode: baseyear (black), 2040BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (green)

Big Meadows (VA) - July 1995 Ozone (ppm)

0.090 0.080 0.070 0.060 0.050 0.040 0.030 0.020

11

12

13 O3_base

Figure 9-3

14

15

O3_2040_bwc

16

17

18

19

20

O3_2040_bwc_scaled_BC

Sensitivity of O3 at Big Meadows, VA to boundary conditions during the July 1995 episode: baseyear (black), 2040BWC emissions and 1995 to boundary conditions (red) and 2040-BWC emissions and scaled to boundary conditions (green)

9-14

ODMAN,

ET AL.


Figure 9-4 and Figure 9-5 show the sensitivity of fine sulfate concentrations to boundary conditions at Great Smoky Mountains and Shenandoah National Parks, respectively. When scaled 3 BCs are used (yellow bars), the estimated 2040 daily average sulfate concentrations are 2 -3 µg/m lower than those estimated by using the original BCs (red bars). This corresponds to an additional 1015% reduction in estimated 2040 sulfate concentrations with respect to the baseyear levels. Figure 9-6 and Figure 9-7 show the sensitivity of fine nitrate concentrations to boundary conditions at Great Smoky Mountains and Shenandoah National Parks, respectively. The daily average nitrate concentrations estimated by using the original BCs (red bars) or scaled BCs (yellow bars) do not 3 differ by more than 0.5 µg/m . Figure 9-8 and Figure 9-9 show the sensitivity of fine ammonium concentrations to boundary conditions at Great Smoky Mountains and Shenandoah National Parks, respectively. When scaled BCs are used (yellow bars), the estimated 2040 daily average ammonium concentrations are 0.5-1.0 3 µg/m lower than those estimated by using the original BCs (red bars). This corresponds to an additional 20-30% reduction of estimated 2040 ammonium concentrations with respect to the baseyear levels. Figure 9-10 shows the sensitivity of sulfate wet deposition to boundary conditions at Great Smoky Mountains and Shenandoah National Parks. When scaled BCs are used (yellow bars), the estimated 2040 sulfate wet deposition fluxes are reduced by an additional 15% with respect to the baseyear levels. Figure 9-11 shows the sensitivity of nitrate wet deposition to boundary conditions at Great Smoky Mountains and Shenandoah National Parks. There is little difference between the wet deposition of nitrate estimated by using the original BCs (red bars) and scaled BCs (yellow bars). Figure 9-12 shows the sensitivity of ammonium wet deposition to boundary conditions at Great Smoky Mountains and Shenandoah National Parks. When scaled BCs are used (yellow bars), the estimated 2040 ammonium wet deposition are reduced by an additional 5-10% with respect to the baseyear levels. It appears that reducing the boundary conditions proportionally to the emission reduction in the modeling domain can have a significant impact on the simulated pollutant values. However, there is no evidence that emission reductions in the eastern U.S. would impact these BCs. For the purpose of this study, four boundary sectors were considered: North (Canada), West (US), Southwest (Mexico), East and Southeast (marine). It is clear that the marine BC's should not be reduced since they are already low and there are no emission reductions planned over the ocean. The SW (Mexico) and North (Canada) boundaries are outside U.S. regulatory control and it is unclear whether pollutants from these boundaries will increase or decrease in the future. The West boundary is under U.S. regulatory control, but it is unclear whether emission reductions will be implemented. Because of the uncertainty in the magnitude and even the direction of future year boundary conditions, it was decided that the BCs will remain unchanged from the baseyear for all future year simulation.

9-15

ODMAN,

ET AL.


Great Smoky Mountains (TN) - Average Sulfate (µg/m3) (Basecase & BWC) 25.0 20.0 15.0 10.0 5.0 0.0 7/11/95

7/12/95

7/13/95

SULF_base Figure 9-4

7/14/95

7/15/95

SULF_2040

7/16/95

7/17/95

7/18/95

7/19/95

SULF_2040_scaled_BC

Sensitivity of fine SO42- at Great Smoky Mountains, TN to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

Shenandoah National Park (VA) - Average Sulfate (µg/m3) (Basecase & BWC) 25.0 20.0 15.0 10.0 5.0 0.0 7/11/95

7/12/95

7/13/95

SULF_base Figure 9-5

7/14/95

7/15/95

SULF_2040

7/16/95

7/17/95

7/18/95

7/19/95

SULF_2040_scaled_BC

Sensitivity of fine SO42- at Shenandoah National Park, VA to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

9-16

ODMAN,

ET AL.


Great Smoky Mountains (TN) - Average Nitrate (µg/m3) (Basecase & 2040 BWC) 2.0 1.6 1.2 0.8 0.4 0.0 7/11/95

7/12/95

7/13/95

NITF_base Figure 9-6

7/14/95

7/15/95

NITF_2040

7/16/95

7/17/95

7/18/95

7/19/95

NITF_2040_scaled_BC

Sensitivity of fine NO3- at Great Smoky Mountains, TN to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

Shenandoah National Park (VA) - Average Nitrate (µg/m3) (Basecase & 2040 BWC) 2.0 1.6 1.2 0.8 0.4 0.0 7/11/95

7/12/95

7/13/95

7/14/95

7/15/95

NITF_base NITF_2040 Figure 9-7

7/16/95

7/17/95

7/18/95

7/19/95

NITF_2040_scaled_BC

Sensitivity of fine NO3- at Shenandoah National Park, VA to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

9-17

ODMAN,

ET AL.


Great Smoky Mountains (TN) - Average Ammonium (µg/m3) (Basecase & 2040 BWC) 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 7/11/95

7/12/95

7/13/95

AMNF_base Figure 9-8

7/14/95

7/15/95

AMNF_2040

7/16/95

7/17/95

7/18/95

7/19/95

AMNF_2040_scaled_BC

Sensitivity of fine NH4+ at Great Smoky Mountains, TN to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

Shenandoah National Park (VA) - Average Ammonium (µg/m3) (Basecase & 2040 BWC) 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 7/11/95

7/12/95

7/13/95

AMNF_base Figure 9-9

7/14/95

7/15/95

AMNF_2040

7/16/95

7/17/95

7/18/95

7/19/95

AMNF_2040_scaled_BC

Sensitivity of fine NH4+ at Shenandoah National Park, VA to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

9-18

Sulfate Flux (mg/m2)

ODMAN,

ET AL.


Weekly Sulfate Wet Deposition (Basecase & 2040 BWC) 60.0 50.0 40.0 30.0 20.0 10.0 0.0

Great Smoky Mountains SO4_base

Nitrate Flux (mg/m2)

Figure 9-10

SO4_2040

Shenandoah National Park SO4_2040_scaled_BC

Sensitivity of SO42- wet deposition at Great Smoky Mountains and Shenandoah National Parks to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

Weekly Nitrate Wet Deposition (Basecase & 2040 BWC) 20.0 15.0 10.0 5.0 0.0

Great Smoky Mountains NO3_base Figure 9-11

NO3_2040

Shenandoah National Park NO3_2040_scaled_BC

Sensitivity of NO3- wet deposition at Great Smoky Mountains and Shenandoah National Parks to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

9-19

ODMAN,

ET AL.


NH4 Flux (mg/m2)

Weekly Ammonium Wet Deposition (Basecase & 2040 BWC) 20.0 15.0 10.0 5.0 0.0

Great Smoky Mountains NH4_base Figure 9-12

NH4_2040

Shenandoah National Park NH4_2040_scaled_BC

Sensitivity of NH 4+ wet deposition at Great Smoky Mountains and Shenandoah National Parks to boundary conditions during the July 1995 episode: baseyear (blue), 2040-BWC emissions and 1995 boundary conditions (red) and 2040-BWC emissions and scaled boundary conditions (yellow).

9-20

ODMAN,

ET AL.


10 RESPONSES TO FUTURE YEAR EMISSION STRATEGIES 10.1 Ozone Response to Future Year Emission Strategies 10.1.1 Spatial Plots of Daily Maximum Ozone Spatial plots (maps) of future year daily maximum ozone concentrations were prepared for each simulated day. The SAMI region rather than the entire modeling domain was focused on these maps. This allows direct comparison of the daily maximum ozone concentrations between the baseyear (see Section 5.2) and any future year strategy. Figure 10-1 shows the daily maximum ozone concentrations on July 12 when 2010-OTW emissions are used. Maps for other days and other strategies for the years 2010 and 2040 can be found in Appendix A. To facilitate visual comparisons, the change in daily maximum ozone concentrations were also mapped. Figure 10-1 shows the estimated change from July 12, 1995 to 2010 under the OTW strategy. For the 2010-OTW strategy, the mapped change is the difference between concentrations simulated using the future and baseyear emissions (i.e., 2010-OTW minus baseyear). A negative number indicates a decrease in daily maximum ozone. The change for the other future year strategies are defined in Equations 9-2 through 9-6. The change maps for all strategies and for all simulated days can be found in Appendix A.

Figure 10-1

Daily maximum ozone concentrations on July 12 when 2010-OTW emissions are used.

10-1

ODMAN,

ET AL.

Figure 10-1


Estimated change in daily maximum ozone concentrations from July 12, 1995 to 2010 under the OTW strategy.

10.1.2 Diurnal Plots of Hourly Ozone at Selected Stations One way to examine the effects of the different emission reduction strategies on ozone is to look at hourly plots of ozone at different sites for each episode. The example in Figure 10-2 compares the hourly ozone concentrations resulting from the three emission strategies for the years 2010 and

Great Smoky Mountains (TN) - July 1995

Ozone (ppm)

0.090 0.080 0.070 0.060 0.050 0.040 0.030

11 Base Figure 10-2

12 2010_OTW

13

14

2040_OTW

15 2010_BwC

16

17 2040_BwC

18 2010_BB

19

20 2040_BB

Diurnal variations of ozone using OTW, BWC, and BB emission strategies in the years 2010 and 2040 for the July 11-19, 1995 episode at Look Rock, TN. 10-2

ODMAN,

ET AL.

Table 10-1


List of stations for which diurnal plots of future year ozone concentrations were prepared. Number

Station

1 2 3 4 5 6 7 8 9 10 11 12

Sipsey, AL Washington, DC South DeKalb, GA Asheville, NC Frying Pan, NC Purchase Knob, NC Clingmans Dome, TN Look Rock, TN Knoxville, TN Big Meadows, VA Roanoke, VA Greenbrier Co, WV

2040 to those of July 11-19, 1995 episode at Look Rock, TN. The 2010-OTW, 2010-BWC, and 2040OTW emission strategies each show similar reductions in peak ozone ranging from approximately 10 to 15 ppb. The 2040 BB strategy shows peak ozone reductions between 10 and 25 ppb. The 2040BWC and 2010-BB strategies fall somewhere between the 2010-OTW and 2040-BB results. All three 2010 and 2040 strategies also show reductions in the nighttime ozone; however, there is little change in the magnitudes of nighttime reductions for different strategy. Similar plots for the 12 sites listed in Table 10-1 can be found in Appendix A. 10.1.3 Seasonal Cumulative Ozone W126 Response

Ozone W126 (ppm-hrs)

Figure 10-1 contains the modeled seasonal (May-September) cumulative reconstructed ozone W126 for the OTW, BWC, and BB emission strategies for the years 2010 and 2040 and compares

50.0

Seasonal Ozone W126

40.0 30.0 20.0 10.0 0.0 Basecae 2010 OTW 2010 BwC

2010 BB 2040 OTW 2040 BwC

Great Smoky Mountains Figure 10-1

2040 BB

Shenandoah

Simulated seasonal cumulative ozone (W126) for the years 2010 and 2040 under the OTW, BWC and BB strategies. 10-3

ODMAN,

ET AL.


them to the basecase at Great Smoky Mountains (Look Rock, TN) and Shenandoah (Big Meadows, VA). Significant reductions in ozone W126 are projected to occur at both GRSM (40%) and SHEN (48%) due to the emissions reductions under the 2010-OTW strategy. Additional reductions for 2010BWC are minimal, but there is approximately a 30% reduction in ozone W126 from the 2010-OTW strategy to the 2010-BB strategy. The ozone W126 for 2040-OTW is very similar to that for 2010-OTW at both GRSM and SHEN. Both BWC and BB strategies yield incremental benefits at GRSM (15% for BWC and 35% for BB) and SHEN (20% for BWC and 35% for BB) in the year 2040.

10.2 Aerosol Response to Future Year Emission Strategies 10.2.1 Spatial Plots of Daily Average PM Concentrations Spatial plots (maps) of future year daily average PM concentrations were prepared for each simulated day. These maps allow direct comparison of the daily average PM concentrations between the baseyear (see Section 6.2) and any future year strategy. Figure 10-1 shows the daily average PM2.5 concentrations on July 15, 1995 when 2010-OTW emissions are used. Maps for other days and other strategies for the years 2010 and 2040 can be found in Appendix A. In addition, maps can be found in Appendix A for sulfate, nitrate, ammonium, elemental and organic carbon and soil components of PM2.5 as well as PM10. Figure 10-1 can be compared with Figure 6-1 to see the estimated change in daily average PM2.5 on July 15 from 1995 to 2010 under the OTW strategy. Figure 10-1 shows the difference between the daily average concentrations of Figure 10-1 and Figure 6-1 (i.e., 2010-OTW minus 1995). Appendix A contains maps showing the difference in concentrations for all simulated days, all future year strategies (2010 and 2040) and all components of PM 2.5 as well as PM10. For the year 2010 and OTW strategy, the mapped change is the difference between concentrations simulated using the future

Figure 10-1

Daily average PM2.5 concentrations over the SAMI region on July 15, 1995 using the 2010-OTW emission strategy. 10-4

ODMAN,

ET AL.

Figure 10-1


Change in daily average PM2.5 concentrations from July 15, 1995 to 2010 under the OTW emission strategy.

and baseyear emissions (i.e., 2010-OTW minus baseyear). A negative number indicates a decrease in daily average PM concentration. The changes for the other future year strategies are defined in Equations 9-2 through 9-6. 10.2.2 Charts of Daily Average PM Concentrations for Selected Stations Daily averaged speciated fine PM concentrations were examined for each day contributing to the annual average for the basecase and three emission strategies for 2010 and 2040. Figure 10-1 shows the daily averaged sulfate concentrations for the basecase and the OTW, BWC, and BB strategies for 2010 at Great Smoky Mountains. Low sulfate days (Class 1 and 2) show minimal changes and can even show an increase in sulfate concentrations for some strategies. The increase in sulfate is thought to be due to either an increase in SO2 emissions or an increase in heterogeneous sulfate chemistry. Although SO2 emissions decrease in the SAMI states, there are some areas that show local increases in SO2 emissions. These increases may be due to new sources coming on-line in 2010, increased SO2 production by an existing source, or a discrepancy in the way the basecase and future emission inventories were developed. In the future scenarios, utility boilers are assumed to be operating all hours. However, in the basecase runs, actual hourly emissions (some being zero for nonoperating units) were used. An increase in heterogeneous sulfate chemistry could result from (1) an increase in hydrogen peroxide resulting from a decrease in NOx emissions ∗ and/or (2) an increase in the pH of the liquid droplets due to increased ammonia emissions, leading to greater oxidation by ozone. The days with higher sulfate concentrations (Classes 4 and 5) show significant reductions for each of the three control strategies, except for May 27, 1995, which shows an increase in sulfate under the 2010-OTW possibly due to the reasons mentioned above. Similar charts were prepared for all PM2.5 components at the six stations listed in Table 10-1 for the years 2010 and 2040 and can be found ∗

This is the case for the moderate-NOx regime. For the low- and high-NOx regimes, H2O2 decreases with decreasing NOx emissions (Stein and Lamb, 2000).

10-5

ODMAN,

ET AL.


Conc. (µ g/m 3 )

Great Smoky Mountains (TN) - Average Sulfate 20.0

Class 1

Class 3

Class 2

Class 4

Class 5

15.0 10.0 5.0 0.0 07/15/95

07/31/91

07/27/91

2010_BWC

07/12/95

05/27/95

08/11/93

2010_OTW

06/24/92

04/29/95

08/07/93

Figure 10-1

08/04/93

05/15/93

04/26/95

03/24/93

02/09/94

03/27/93

Basecase

2010_BB

Modeled sulfate concentrations at Great Smoky Mountains for the basecase and the OTW, BWC, and BB emission strategies in 2010.

in Appendix A. Nitrate concentrations change slightly for Class 1 and 2 days, but increase significantly for Class 4 and 5 days. The increase in nitrate is a result of more free ammonia being available to convert gas-phase nitric acid to the particulate nitrate form due to a decrease in particulate sulfate and an increase in ammonia emissions. Particulate ammonium tend to increase due to an increase in ammonia emissions for the OTW and BWC strategies, but show a significant decrease for the BB strategy. Sensitivity runs comparing the response of ozone and organic PM to anthropogenic and biogenic carbon emissions in the rural areas showed an order of magnitude higher response to biogenic emissions. As a result, organic PM show little change in the future year strategies because the biogenic emissions were assumed to remain unchanged. 10.2.3 Annual Average PM Response Figure 10-3 and Figure 10-2 demonstrate how various components of the annual PM2.5 at the Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks respond to the changes in emissions under the OTW, BWC, and BB strategies. The relative changes are dependent on the starting point for the changes (i.e., basecase levels). The reductions cannot reach 100% even if all the emissions were eliminated because the boundary conditions for the future scenarios are the same as the basecase. As emissions are reduced, the benefits get smaller because of the relative increase in the contribution of transport into the domain. Daily-averaged sulfate concentrations decrease by 6% (2010-OTW) to 60% (2040-BB) at GRSM and 16% (2010-OTW) to 58% (2040-BB) at SHEN. Nitrate Table 10-1

List of stations for which charts of daily PM concentrations were prepared. Number

Station

1 2 3 4 5 6

Dolly Sods (WV) Great Smoky Mnt. (TN) Jefferson (VA) Shenandoah (VA) Shining Rock (NC) Sipsey (AL)

10-6

ODMAN,

ET AL.


Concentration (µ g/m 3)

Annual Average Fine PM at Great Smoky Mountains 12.0 10.0 Soils EC ORG NH4 NO3 SO4

8.0 6.0 4.0 2.0 0.0 Basecase 2010_OTW 2010_BWC

Figure 10-3

2010_BB

2040_OTW 2040_BWC

2040_BB

Model estimates for annual average PM2.5 concentrations at Great Smoky Mountains National Park for the basecase and OTW, BWC, and BB emission strategies for 2010 and 2040.

increases for the OTW and BWC strategies at GRSM, but decreases slightly for the BB strategy. On the other hand, nitrate decreases for all the strategies at SHEN, except for 2040-OTW where nitrate increases slightly. This difference in the responses of the two sites is probably due to the proximity of SHEN to SO2 and NOx sources that are reduced and to the site’s relatively larger distance from the NH3 sources that are predicted to increase in 2010 and 2040. Ammonium concentrations increase slightly at GRSM for the 2010-OTW strategy, but decreased by 8% (2010-BWC) to 53% (2040-BB) for the other strategies. Ammonium decreased for all strategies at SHEN by 10% (2010-OTW) to 47% (2040-BB). The change in organic PM is small at both sites, except the BB strategy that shows approximately a 15% decrease for both 2010 and 2040. Elemental carbon was reduced under all strategies with decreases up to 50% (2040-BB) at both sites. Soil concentrations increased for the OTW and BWC strategies, but decrease for the BB strategies at GRSM. Large reductions in soils (up to 45%) were predicted at SHEN under the BB strategy.

Concentration (µ g/m 3)

Annual Average Fine PM at Shenandoah 12.0 10.0 Soils EC ORG NH4 NO3 SO4

8.0 6.0 4.0 2.0 0.0 Basecase 2010_OTW 2010_BWC

Figure 10-2

2010_BB

2040_OTW 2040_BWC

2040_BB

Model estimates for annual average PM2.5 concentrations at Shenandoah National Park for the basecase and OTW, BWC, and BB emission strategies for 2010 and 2040. 10-7

ODMAN,

ET AL.


10.3 Acid Deposition Response to Future Year Emission Strategies Weekly cumulative wet and dry deposition mass fluxes by species were calculated for each week contributing to the annual average for the basecase and three emission strategies for the years 2010 and 2040. Deposited species examined for the acid deposition assessment include: 2-

Dry: SO2 (gas), NO2 (gas), NO (gas), NH3 (gas), HNO3 (gas), HONO (gas), N2O5 (gas), SO4 + 2+ 2+ (PM), NO3 (PM), NH4 (PM), Mg (PM), Ca (PM) 2-

-

+

+

2+

2+

Wet: SO2 (very small), SO4 , NO3 , NH4 , H (hydrogen ion), Mg , Ca

The wet and dry deposition fluxes of sulfur, oxidized nitrogen, reduced nitrogen and total nitrogen were computed from the above, as shown in Table 10-1. A separate report was prepared (Odman et al, 2001) that discusses acid deposition estimates for future years in more detail. 10.3.1 Spatial Plots of Weekly Cumulative Deposition Fluxes Maps of future year weekly cumulative acid deposition fluxes were prepared for each episode. These maps allow direct comparison between the baseyear and any future year strategy. Figure 10-1 shows the estimated cumulative wet deposition flux of sulfate for the week of July 11-18, 2010 under the OTW emission strategy. To see the change in deposition from 1995, this figure can be compared to Figure 7-1. To facilitate the comparison, the difference was also mapped and is illustrated in Figure 10-2. Note that while the general trend is a decrease in wet deposition of sulfate, there may be some local increases. Figure 10-3 shows the cumulative dry deposition of SO2 for the same week in the year 2010 under the OTW strategy. The change in SO2 dry deposition is shown in Figure 10-4. While the general trend in deposition is a decrease there are several locations where an increase is estimated. This is because the 2010-OTW strategy does not reduce the SO2 emissions uniformly over the SAMI region and there are actually locations where the emissions increase. Maps for other days and other strategies of the years 2010 and 2040 can be found in Appendix A. Maps exist for the wet deposition of sulfate, nitrate, ammonium, calcium, magnesium, hydronium and the dry deposition of sulfate, nitrate, SO2 and HNO3. Difference maps also exist for all these species. The differences are calculated from the baseyear for the 2010-OTW strategy, from 2010-OTW for the 2040-OTW strategy and from the OTW strategy of the corresponding year for BWC and BB strategies (Equations 9-1-9-6). Table 10-1

Equations used to evaluate the sulfur and nitrogen wet and dry deposition.

Species

Equation used to determine deposition

S

dry

N dry oxidized N dry reduced

32 dry 32 SO + SO24 + dry 64 2 96 14 14 14 28 14 14 NOdry + NOdry NO3- dry + N Odry + HONOdry + HNOdry 2 + 3 30 46 62 108 2 5 47 63 14 14 NHdry NH+4 dry 3 + 17 18

N dry

dry N dry oxidized + N reduced

Swet

32 32 SOwet SO24 + wet 2 + 64 96 14 NO-3 wet 62 14 NH+4 wet 18

N wet oxidized N wet reduced N wet

wet N wet oxidized + N reduced

10-8

ODMAN,

ET AL.


Figure 10-1

Estimated weekly cumulative sulfate wet deposition for July 11-18, 2010 under the OTW strategy.

Figure 10-2

Estimated change in weekly cumulative sulfate wet deposition for July 11-18 from 1995 to 2010 under the OTW strategy.

10-9

ODMAN,

ET AL.


Figure 10-3

Estimated weekly cumulative SO2 dry deposition for July 11-18, 2010 under the OTW strategy.

Figure 10-4

Estimated change in weekly cumulative SO2 dry deposition for July 11-18 from 1995 to 2010 under the OTW strategy.

10-10

ODMAN,

ET AL.


2

Sulfate Flux (mg/m )

Sulfate Wet Deposition at Great Smoky Mountains Class 1

120.0

Class 2

Class 3

Class 4

90.0 60.0 30.0 0.0 April 1995 Basecase

Figure 10-1

May 1995 2010-OTW

June 1992 March 1993 Febr 1994 2010-BwC

2010-BB

May 1993 August 1993 July 1991

2040-OTW

2040-BwC

2010-BB

Modeled weekly cumulative sulfate wet deposition fluxes at Great Smoky Mountains for the basecase and the OTW, BWC, and BB emission strategies in 2010 and 2040.

10.3.2 Charts of Weekly Cumulative Deposition for Selected Stations Figure 10-1 shows weekly cumulative wet deposition of sulfate at Great Smoky Mountains for the basecase and the OTW, BWC, and BB strategies for 2010 and 2040. Each strategy shows decreases in sulfate from the basecase levels except for the 2010-OTW strategy for the May 1995 and February 1994 episodes. Significant decreases in wet sulfate deposition are seen for all other strategies. Appendix A contains similar charts for the wet deposition fluxes of sulfate, nitrate, ammonium, calcium, magnesium and hydronium at the six stations listed in Table 10-1. Although not shown here, nitrate, calcium, and magnesium wet deposition fluxes show minimal fluctuation from the basecase for all episodes and emission strategies. Ammonium deposition typically shows either a slight increase or decrease from the basecase for the OTW and BWC strategies, while the BB strategy always shows a significant decrease from the basecase. Hydrogen ion deposition decreases with each strategy due to the reduction in sulfate. Again, specific details for these species are not presented here, but can be found in Appendix A. Figure 10-1 shows nitric acid dry deposition for the basecase and for the OTW, BWC, and BB strategies for 2010 and 2040 at Shenandoah (SHEN). The gas phase nitric acid dry deposition decreased from the basecase for all future year emission strategies, not only at SHEN but GRSM as well (not shown). The sharp decrease from the basecase to the future years and the relative insensitivity to the future year strategy can be explained by examining the domain-wide emissions of NOx for the basecase and the future year emission scenarios (Figure 9-5). Typically, there is a much greater decrease in NOx emissions between the basecase and the future year than there is between each of the future year strategies. Also, recall that the boundary conditions were not reduced in any of the future year strategies. This minimizes the impact of emission reductions on the response of nitric

Table 10-1

List of stations for which charts of weekly cumulative acid deposition fluxes were prepared. Number 1 2 3 4 5 6

Station

State

Dolly Sods GSM-Noland Divide Joyce Kilmer James River Face SNP-White Oak Run Sipsey

WV TN NC VA VA AL

10-11

ODMAN,

ET AL.


Class 1

20.0

Class 2

2

HNO3 Flux (mg/m )

Nitric Acid Dry Deposition at Shenandoah

15.0 10.0 5.0 0.0 July 1991

Basecase

Figure 10-1

June 1992

2010_OTW

March 1993 August 1993

2010_BWC

2010_BB

Feb 1994

May 1995

2040_OTW

May 1993

2040_BwC

July 1995

2040_BB

Modeled weekly cumulative nitric acid dry deposition fluxes at Shenandoah for the basecase and the OTW, BWC, and BB emission strategies in 2010 and 2040.

acid dry deposition. In addition, the two main paths of forming nitric acid are through NO2 with OH radical during the day and NO2 with O3 during the night. Both ozone and OH are reduced significantly from the basecase, but this difference is much less between each of the future year scenarios leading to a much smaller response in nitric acid dry deposition. Recall that the particulate nitrate concentrations, hence nitrate dry deposition, typically increased due to the conversion of gaseous nitric acid to particulate nitrate as a result of increased free ammonia gas. The free ammonia increases were due to decreased sulfate formation and increased ammonia emissions. This is also the reason that ammonia gas dry deposition increased significantly for the OTW and BWC strategies. Ammonium dry deposition did not change much from the basecase except under the BB strategy, which resulted in larger decreases because of significant reductions in ammonia emissions. The other nitrogen containing species (NO, NO2, HONO, and N2O5) all showed decreased concentrations and dry deposition fluxes in all the future year strategies. Gas phase SO2 2and particle phase SO4 both showed significant decreases in dry deposition for all strategies. The charts for dry deposition of sulfate, nitrate, SO2 and HNO3 at the sites listed in Table 10-1 can be found in Appendix A. 10.3.3 Annual Acid Deposition Response Figure 10-2 and Figure 10-3 show the annual average wet and dry deposition fluxes of sulfate at the Great Smoky Mountains (GRSM) and Shenandoah (SHEN) National Parks for the OTW, BWC, and BB strategies for the years 2010 and 2040. The modeled wet deposition of sulfur is significantly higher than the dry deposition. However, it should be noted that the simulated episodes did not consider dry deposition during the selection process and were selected based on wet deposition only. Therefore, there may be a bias towards wet episodes in the annual average deposition fluxes. Both wet and dry sulfate deposition show significant reductions for each of the emission strategies. At GRSM, wet sulfate deposition decreases by 5% (2010-OTW) to 59% (2040-BB), while dry deposition decreases by 36% (2010-OTW) to 78% (2040-BB). Wet sulfate deposition decreases by 24% (2010OTW) to 73% (2040-BB) at SHEN, while dry deposition decreases by 41% (2010-OTW) to 88% (2040BB). These reductions are similar to the reductions in SO2 emissions for each strategy (Figure 9-2). Annual average wet and dry deposition mass fluxes of oxidized and reduced nitrogen at GRSM and SHEN are shown in Figure 10-5 and Figure 10-4. Modeled wet deposition of nitrogen is significantly greater than the dry deposition of nitrogen, especially at GRSM. Although there were large

10-12

ODMAN,

ET AL.


Mass Flux (mg/m 2)

Sulfur Deposition at Great Smoky Mountains 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 Basecase

2010_OTW 2010_BWC

2010_BB

Wet Sulfur Figure 10-2

2040_OTW 2040_BWC

2040_BB

Dry Sulfur

Annual average wet and dry deposition mass fluxes of sulfur at Great Smoky Mountains for the basecase and the three strategies in 2010 and 2040.

reductions in NOx emissions in the SAMI states for each of the three control strategies, the wet deposition of the oxidized nitrogen does not change significantly at either site. The low response in nitrate wet deposition is counterintuitive and contradicts with other studies (e.g., Shin and Carmichael, 1992). While the real reason is not known, this may be due to NOx that is transported to the Appalachian Mountains from outside the SAMI region. Recall that the boundary conditions were not changed for the future year control strategies and that NOx emissions were not reduced outside the SAMI states for the BWC and BB control strategies. Furthermore, NH3 emissions outside the SAMI states (accounting for 85% - 97% of the domain wide NH3 emissions) increased by 38% in 2010 and by 84% in 2040. This, in addition to SO2 reductions, resulted in more free ammonia

Mass Flux (mg/m 2)

Sulfur Deposition at Shenandoah 25.0 20.0 15.0 10.0 5.0 0.0 Basecase

2010_OTW 2010_BWC

2010_BB

Wet Sulfur Figure 10-3

2040_OTW 2040_BWC

2040_BB

Dry Sulfur

Annual average wet and dry deposition mass fluxes of sulfur at Shenandoah for the basecase and the three strategies in 2010 and 2040. 10-13

ODMAN,

ET AL.


Mass Flux (mg/m 2)

Nitrogen Deposition at Great Smoky Mountains 50.0 40.0 30.0 20.0 10.0 0.0 Basecase

2010_OTW 2010_BWC

Wet Oxidized N Figure 10-5

2010_BB

Wet Reduced N

2040_OTW 2040_BWC

Dry Oxidized N

2040_BB

Dry Reduced N

: Annual average wet and dry deposition mass fluxes of oxidized and reduced nitrogen at Great Smoky Mountains for the basecase and the three strategies in 2010 and 2040.

becoming available to convert gas-phase nitric acid to the particulate nitrate form. The dry deposition rates of particulate nitrate are much lower than for gas-phase nitric acid. Therefore, oxidized nitrogen in the form of particulate nitrate can be transported greater distances and in greater concentration before being deposited to the surface. The wet deposition module can scavenge mass from all layers in the model, especially the upper layers where the clouds are formed. The pollutant levels in these upper layers are influenced by long-range transport from regions outside the SAMI states and the boundary, so reductions in local NOx emissions may not significantly change the amount of nitrate wet deposition. Also, the wet deposition of the reduced nitrogen does not change significantly with either the OTW or BWC strategies, but does show decreases between 10% and 15% for the BB strategies.

Mass Flux (mg/m 2)

Nitrogen Deposition at Shenandoah 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 Basecase

2010_OTW 2010_BWC

Wet Oxidized N Figure 10-4

2010_BB

Wet Reduced N

2040_OTW 2040_BWC

Dry Oxidized N

2040_BB

Dry Reduced N

Annual average wet and dry deposition mass fluxes of oxidized and reduced nitrogen at Shenandoah for the basecase and the three strategies in 2010 and 2040. 10-14

ODMAN,

ET AL.


The dry deposition of oxidized nitrogen at GRSM decreased for all three strategies with reductions ranging from 8% - 10% (2010-OTW and 2040-OTW) to 37% (2040-BB), while reductions range from 25% (2010-OTW and 2040-OTW) to 66% (2040-BB) at SHEN. For ground-level sources, local emission reductions typically show a greater influence on the pollutant levels in the first layer than pollutant levels in the upper layers. Therefore, the reductions in oxidized nitrogen dry deposition are much greater than those for wet deposition (on a percentage basis) because dry deposition fluxes are a function of pollutant levels in the first layer of the model, whereas wet deposition fluxes are affected by pollutant levels in all layers of the model (especially the upper layers). The reduced nitrogen dry deposition increased for all three strategies at GRSM; these increases ranged from 34% (2010-BB) to 123% (2040-OTW). At SHEN, the reduced nitrogen dry deposition shows increases for the OTW and BWC strategies between 48% (2010-OTW) and 143% (2040-OTW), but shows reductions of 11% and 6% for the 2010-BB and 2040-BB strategies, respectively. The increase in reduced nitrogen deposition was a result of an increase in ammonia emissions for the OTW and BWC strategies. Typical annual observations indicate that the mass of sulfur and nitrogen removed from the atmosphere by wet deposition processes is approximately equal to the amount removed by dry deposition processes. Recall, that the modeled annual average wet deposition of ammonium (reduced nitrogen) was over predicted by approximately 600% at GRSM and 140% at SHEN. Decreasing the modeled reduced nitrogen by these percentages would make the ratio of wet to dry deposition much closer to values that are typically observed.

10-15

ODMAN,

ET AL.



10-16

ODMAN,

ET AL.


11 MODELING OF SENSITIVITIES 11.1 Introduction Sensitivity analysis is a powerful tool for identifying and quantifying the impacts of inputs (e.g. emissions) and system parameters (e.g. rate constants) on air quality modeling results. Sensitivity analyses are performed at various stages of the modeling process. They can be used for diagnostic evaluation, to determine areas of improvement, or to analyze the impacts of various control strategies (USEPA, 1999). Traditionally, a “brute-force” approach has been used to calculate sensitivities of pollutant concentrations to various parameters. This method involves running the model a number of different times, each time perturbing one parameter and comparing the results to the original run. On the other hand, embedding a direct sensitivity technique into the model allows the user to perform numerous sensitivity calculations in one model run. The Decoupled Direct Method (Dunker, 1981 and 1984) is one such technique that directly calculates sensitivity coefficients by taking the derivatives of the governing equations. The Decoupled Direct Method was successfully implemented into the CIT (California/Carnegie Institute of Technology) three-dimensional photochemical air quality model (Russell et al., 1988; Harley et al., 1993) by Yang et al. (1997) to estimate sensitivities of modeled ozone concentrations to initial conditions, dry deposition velocities, reaction rate constants, wind speeds, and emissions. The application of this technique is called the Decoupled Direct Method for three-dimensional models (DDM-3D). Most of the sensitivity analysis in the past has focused on the effect of NOx and VOC emission reductions on ambient ozone concentrations. Minimal work has been undertaken to address how fine particles and wet acid deposition levels respond to emission changes. Here, DDM-3D is incorporated into the Urban to Regional Multiscale - One Atmosphere (URM-1ATM) model and is extended to allow speciated PM and wet deposition sensitivities to be calculated in addition to ozone and other gas-phase sensitivities. This enhanced version of URM-1ATM produces multidimensional concentration and sensitivity fields that can be used to assess local and regional impacts from individual and distributed sources. Here, geographic sensitivity analysis will be used to examine the potential influence of SO2, NOx , and NH3 emissions from each SAMI state (AL, GA, NC, SC, TN, KY, VA, WV) and areas outside the SAMI region on fine particulate matter (sulfate, nitrate, and ammonium) and sulfate wet deposition levels in the Class I areas in order to make a first-order estimate of the source/receptor relationships.

11.2 Direct Sensitivity Analysis The decoupled direct method for three-dimensional models (DDM-3D) is a sensitivity analysis technique based upon solving a set of equations derived from differentiating the original set of equations governing the atmospheric pollutant dynamics. Using direct derivatives of the equations governing the evolution of species concentrations, the local sensitivities to a variety of model parameters and inputs are computed simultaneously along with the species concentrations. In equation form, the local sensitivity of a model output (e.g., PM2.5 concentration), c i, to a model input (e.g., emissions) or system parameter (e.g., rate constant), pj, can be described as:

sij ( x, t ) =

∂C i ( x, t ) ∂p j (x, t )

(11-1)

Since C is a function of p (the subscripts were dropped for convenience) it can be expanded in a Taylor series for small changes ∆p in parameter p around its original value p0 as:

11-1

ODMAN,

ET AL.


C ( p 0 + ∆p) = C ( p 0 ) +

∂C ∂ 2C ( p 0 ) × ∆p + 2 ( p0 ) × ∆p 2 + L . ∂p ∂p

(11-2)

The parentheses in Equation 11-2 denote functional relationships. As a first order approximation, if the terms with second and higher order derivatives are ignored and the definition of the sensitivity coefficient is used then

∆C = C ( p0 + ∆p ) − C( p 0 ) ≈ s ( p 0 ) × ∆p .

(11-3)

Thus, by using the sensitivity coefficient s (i.e., the first derivative of C with respect to p) at p = p0 and ∆p, one can estimate ∆C, or the change in C. If C is a linear function of p, then the approximation would be exact for any ∆p. Otherwise, Equation 11-3 is an approximation which is only valid for small ∆p. Note that using the higher-order derivatives would yield a more accurate approximation (or similar accuracy for relatively larger ∆p). In summary, when using the sensitivity coefficients presented below, keep in mind that the approximation in Equation 11-3 involves an error. This error would be larger for larger ∆p and more nonlinear relationships between C and p. As developed by Yang et al. (1997), the method takes full advantage of the numerical routines already incorporated into the AQM. Sensitivity coefficients determined by DDM-3D include the response to initial and boundary conditions, horizontal transport, vertical advection and diffusion, emissions, both homogeneous and heterogeneous chemical transformation, aerosol formation, and scavenging processes. The sensitivity coefficients are a function of the concentrations, but the concentrations are not dependent on the sensitivities and can be integrated separately, hence the decoupled nature of the method. Pollutant sensitivities to the emissions of individual species or lumped species (e.g., NOx and total VOC) can be calculated. Furthermore, these emissions can be assigned to the entire modeling domain, sub-regions of the domain (e.g., a state or county), or even individual grid cells. The numerical implementation of the sensitivity calculations for transport and gas-phase chemical reaction processes has been extensively documented by Yang et al. (1997) and will not be repeated here. Only the calculation of sensitivity coefficients for aerosols and wet deposition will be discussed. PM sensitivities for the inorganic equilibrium species are calculated in two steps. The first involves the condensation and evaporation of gas and particulate species, while the second step involves the growth of particles in each aerosol size bin. The derivative of each equilibrium equation in Table 3-1 is calculated to produce 15 sensitivity expressions. For example, the derivative of the sulfate concentration with respect to parameter pj for the first equilibrium expression in Table 3-1, i.e.,

[ SO42 − ] =

K1γ HSO − [ HSO − ] 4 4 γ H + γ SO 2 − [ H + ]

(11-4)

4

would yield:

 − 1 ∂[ H + ] ∂[ SO42 − ] ∂[ HSO4− ] γ H + γ SO42 − ∂  γ HSO4− 1 = [ SO42 − ] + + +  [ H ] ∂p j ∂p j [ HSO4− ] ∂p j γ HSO4− ∂p j  γ H + γ SO42 −  11-2

   (11-5)  

ODMAN,

ET AL.


Functionally, the last term involving the activities is complex. However, the activities are constrained by the relative humidity, reducing the sensitivity to other parameters. When pj represents an emission, the relative humidity does not change with pj, therefore this last term is neglected. To complete the set of sensitivity equations, the derivative of the mass conservation and charge balance expressions are required. Table 11-1 shows the mass conservation and the charge balance expressions (12 expressions). For example, the derivative of the total nitrate (TN) expression yields:

∂ [TN ] ∂[ AN ] ∂ [HNO 3 ] = + ∂p j ∂p j ∂p j

(11-6)

where, [TN] is the total nitrate concentration, [AN] is particulate nitrate concentration, and [HNO3] is the nitric acid concentration. The resulting 27 (=15+12) sensitivity equations are solved simultaneously to produce the desired sensitivities for particulate sulfate, nitrate, ammonium, sodium, chloride and hydrogen, in addition to the sensitivities for gas-phase HNO3, NH3 and HCl. Because the production of condensable organics is calculated in the gas-phase module of URM-1ATM, the sensitivity coefficients for the organic PM are computed in the same way as the gas-phase species. However, extension to using more thermodynamically comprehensive models for semivolatile organic PM (e.g. using partitioning coefficients) is straight forward. Finally, the sensitivities for the PM species are apportioned to each size bin according to the individual species’ concentration in each size bin. Table 11-1

Mass conservation and charge balance expressions.a

[TS ] = [ AS ] [TN ] = [ AN] + [HNO3 ] [TA] = [ AA] + [NH 3 ] [TNa ] = [ANa] [TCl] = [ACl] + [HCl] [AS ] = [SO42− ] + [HSO4− ] + [NaHSO4 ] + [NH 4 HSO4 ] + [Na2 SO4 ] + [( NH 4 ) 2 SO4 ] + 2[( NH 4 ) 3 H ( SO4 ) 2 ] [AN ] = [NO3− ] + [NaNO3 ] + [NH 4 NO3 ] [ AA] = [NH 4+ ] + [NH 4 NO3 ] + [ NH 4 HSO4 ] + [NH 4 Cl] + 2[( NH 4 ) 2 SO4 ] + 3[( NH 4 ) 3 H ( SO4 ) 2 ] [ANa ] = [Na + ] + [NaHSO4 ] + [NaNO3 ] + 2[Na2 SO4 ] + [ NaCl] [ACl ] = [Cl − ] + [NH 4 Cl] + [ NaCl] [AH ] = 2[AS ] + [ AN ] + [ ACl] − [ AA] − [ANa ]

[H ] = 2[SO ] + [HSO ]+ [NO ]+ [Cl ] + [OH ] − [NH ] − [Na ] +

2− 4

− 4

− 3

−

−

a

+ 4

+

Definition of abbreviations: TS = total sulfate, TN = total nitrate, TA = total ammonium, TNa = total sodium, TCl = total chloride, AS = particulate sulfate, AN = particulate nitrate, AA = particulate ammonium, ANa = particulate sodium, ACl = particulate chloride, AH = particulate hydrogen.

11-3

ODMAN,

ET AL.


Next, the “growth” of aerosol sensitivities is calculated using the aerosol droplet currents (Friedlander, 1977). The aerosol droplet current can be represented by:

∂M ∂(mI ) =− +ρI ∂t ∂V

(11-7)

-3

where M is the aerosol mass distribution function (µg m ), I is the particle current (number of particles -1 -3 per unit time; sec ), m is the particle mass (µg), V is the particle volume (m ), and ρ is the particle -3 density (µg m ). The particle current calculated from the aerosol growth module is used to “grow” the sensitivity coefficients through the particle space. Taking the derivative of Equation 11-7 results in:

∂  ∂M ∂t  ∂p j

 =− ∂  ∂V 

 ∂I m  ∂p j 

  + ρ ∂I  ∂p j 

(11-8)

This sensitivity equation is solved for each size bin assuming that growth sensitivity by size is proportional to the original growth distribution, along with a stability criterion to produce the sensitivity coefficients for particle growth. The stability criteria is:

∆t ≤

1 I

(11-9)

where, ∆t is the integration time step. The total PM sensitivity results for each species in each size bin are then determined by following the sensitivity coefficients through both the condensation/evaporation and particle growth sensitivity calculations. The gas and PM sensitivities are impacted by wet deposition and scavenging processes that are simulated by the Reactive Scavenging Module (Berkowitz et al., 1989). Column mass fluxes before and after RSM are calculated and used to scale the sensitivity coefficients for each species:

∑ (c )(∆z ) l

s tk =

k =1 l

t k

k

∑ (c )(∆z ) k =1

o k

s ko

(11-10)

k

where, sk and ck are the sensitivities and concentrations at layer k, l is the number of layers, and ∆ zk is the thickness of layer k. The wet deposition mass flux sensitivity (Swet ) for each species is calculated as:

S wet = ∑ (s ko )(∆z k ) − ∑ (s tk )(∆z k ) l

l

k =1

k =1

(11-11)

Combining the sensitivity calculations from each module results in a set of integrated sensitivity coefficients. The sensitivity coefficients for the gas and particulate species are in units of ppm and µg m 3 per percent increase in the parameter of interest, respectively. The wet deposition sensitivity -2 coefficients are in units of mg m per percent increase in the parameter of interest. 11-4

ODMAN,

ET AL.


11.3 Comparison of DDM-3D and Brute Force Methods A “brute-force” approach has been traditionally used to calculate the response of a model to changes in various input parameters. This method involves running the model a number of different times, each time perturbing one parameter and comparing the results to the basecase run. The brute force approach has been used extensively to study the response of ozone (e.g. McNair et al.,1992; Winner et al.,1995; Bergin et al., 1998) and PM 2.5 (e.g. Seigneur et al., 2000) in air quality models to various model inputs including emissions. However, if the perturbation is small, the brute-force method may not yield accurate sensitivities due to numerical errors propagated in the model. Most of the time, the response of the models to small perturbations would be within the error bounds of the basecase estimates. Moreover, as the number of perturbed parameters increases, the feasibility of the approach is hampered by computational resource limitations. On the other hand, the DDM-3D technique uses direct derivatives of the governing equations and allows the user to perform numerous sensitivity calculations in a single model run. The sensitivity is defined as the first derivative with respect to the parameter (i.e., response to an infinitesimal change in the parameter) therefore, DDM-3D is, by definition, most accurate for small perturbations. However, DDM-3D has the disadvantage of only providing first-order sensitivity coefficients; inaccurate sensitivities may result for large changes in independent variables if the response is non-linear. Both DDM-3D and the brute force sensitivity techniques were applied to the SAMI modeling domain for the July 11-19, 1995 episode using emissions for the 2010-OTW control strategy (discussed in Section 9.2.1). A qualitative comparison was made for the sensitivity coefficients produced for each pollutant to various emissions. Specifically, the sensitivity of ozone, fine sulfate PM, fine nitrate PM, fine ammonium PM, fine organic PM, PM2.5, wet sulfate deposition, wet nitrate deposition, and wet ammonium deposition to a 30% reduction in domain-wide emissions of SO2, NOx , VOCs, and NH3 were evaluated. Figure 11-1 shows a comparison of particulate sulfate sensitivities to a 30% reduction in domain-wide SO2 emissions using DDM-3D and the brute force techniques. The results shown are 24hour averaged sensitivities for July 15. Both methods show a reduction in sulfate concentrations in the -3 SAMI region ranging from approximately 1.2 – 4.7 µg m and very similar spatial patterns.

Figure 11-1

Change in particulate sulfate concentrations due to 30% reduction in SO2 emissions using DDM-3D (left) and brute force (right). 11-5

ODMAN,

ET AL.

Figure 11-2


Change in sulfate wet deposition concentrations due to 30% reduction in SO2 emissions using DDM-3D (left) and brute force (right).

Figure 11-2 shows a comparison of sulfate wet deposition sensitivities to a 30% reduction in domain-wide SO2 emissions using DDM-3D and the brute force techniques. The results are 7-day cumulative sensitivities for the week of July 11-18. Both methods show reductions in sulfate wet -2 deposition of up to ~60 mg m and show very similar spatial patters. Both the DDM-3D and brute force methods provide nearly identical results. However, in the time it took to produce the basecase and one sensitivity using the brute force method, a basecase and thirteen sensitivities were produced using the DDM-3D technique. Table 11-1 contains a subjective description of how well the sensitivities produced by the DDM-3D technique matched the brute force results for each pollutant evaluated. Spatial plots containing sensitivity results using the DDM-3D technique and the brute force method were produced for July 12 and 15, 1995. These plots were presented to the SAMI modeling subcommittee and they rated each sensitivity as “GOOD”, “FAIR”, or “POOR” depending on how well the DDM-3D sensitivity matched the brute force sensitivity.

Table 11-1

Comparison of DDM-3D sensitivity results to brute force results for a 30% reduction in emissions Emissions

Species Ozone

SO2 a

NOx

NH3

VOCs

a

n/a

GOOD

n/a

GOOD

SO42- (aerosol) NO3- (aerosol) NH4+ (aerosol) OC (aerosol) PM2.5

GOOD FAIR GOOD n/aa GOOD

n/aa GOOD GOOD GOOD GOOD

GOOD POOR FAIR n/aa POOR

n/aa FAIR FAIR GOOD GOOD

SO42- (wet dep.) NO3- (wet dep.) NH4+ (wet dep.)

GOOD POOR POOR

n/aa GOOD POOR

n/aa POOR POOR

n/aa GOOD POOR

a

species shows an insignificant response to emission reductions

11-6

ODMAN,

ET AL.


Although the DDM-3D sensitivity results were rated "GOOD" for many species-emission pairs + (e.g., sulfate PM to SO2 emissions), there were some pairs that did not match very well (e.g. NH4 wet deposition to SO2, NOx, VOCs, and NH3 emissions). Also, some species showed an insignificant response to emission reductions (e.g. ozone to SO2 or NH3 emissions) and a comparison is not applicable (N/A). Other species such as elemental carbon, soils, and calcium and magnesium wet deposition showed no response to any emission reduction and will not be discussed. More details on the range, direction, spatial agreement, magnitude of changes, and additional discussion can be found in Appendix D.

11.4 Important Remarks about DDM-3D The limitations have been indicated throughout the report. It is extremely important to understand the limitations of DDM-3D before interpreting the results presented below (Chapter 12). Though these limitations have been stated above (as well as in Chapter 12) they will be summarized here for the readers’ convenience. 1.

The sensitivity coefficient (or sensitivity) calculated by DDM-3D is the local slope (a point value for the level of emissions used in the simulation) of the curve that defines a pollutant concentration as a function of a single emission parameter (a certain type of emission from a given geographic region). For convenience, the sensitivity coefficients were presented above as estimated changes in pollutant concentrations due to a 30% reduction in the emission parameter. These are not actual responses to the emission reductions. In other words, no model simulation was conducted with reduced emissions but the slope was extrapolated from the simulation level to 90% of this level. If the concentration is not a linear function of the emission parameter, there is an error involved in this extrapolation.

2.

A limited analysis was performed to estimate the level of emission reduction at which the extrapolation error becomes significant. Note that the analysis was performed only for the July 1995 episode and only using domain-wide reductions. Among the pollutant-emission pairs discussed above, the differences between the actual model response and the extrapolation of DDM-3D sensitivities were observed qualitatively and were deemed not to be significant up to about 30% reductions. The differences may be larger for other episodes, reductions from an individual state or region, or for reductions beyond 30%.

3.

The sensitivities are calculated for reductions in a single emission parameter at a time. Here the emission parameter was the SO2 emissions from a single sub-domain (a state or a region). For convenient comparison of the sensitivities to the emission reductions from different sub-domains, they were stacked in bar charts. Limited analysis has shown that the error involved in superposing sensitivities from different sub-domains was not significant for the level of reductions discussed here. However, note that this error may compound the extrapolation error.

4.

A more significant error may result if the sensitivities to reductions in different emission source types but from the same sub-domain were superposed. For example, the response of nitrate PM to simultaneous reductions of NOx and NH3 from the same state may be substantially different then the superposed sensitivities to individual reductions.

5.

The grid used did not have the same resolution everywhere. The finest resolution was placed over the SAMI region. Some of the outer sub-domains (e.g., Central region) were covered with very coarse grid resolution. Therefore, the sensitivities to emission reductions from regional subdomains are less reliable than those from SAMI states.

6.

Finally, DDM-3D may be used in initial stages of control strategy design but the effectiveness of the design should always be checked with full-scale modeling. 11-7

ODMAN,

ET AL.



11-8

ODMAN,

ET AL.


12 SENSITIVITY ANALYSIS 12.1 Definition of Regions for Sensitivity Analysis Sensitivity analysis has been performed for the nine episodes using the 2010-OTW emission estimates and meteorological data from each 1991 – 1995 basecase episode. The year 2010 was chosen because the sensitivity results were initially to be used by SAMI to guide the design of control strategies. In the end, the results were used to provide information on relative impact of different geographic regions on Class I areas. The modeling domain was divided into 13 sub-domains. The regional sub-domains are shown in Figure 12-1 and are also listed in Table 12-1. Each of the eight SAMI states is treated as an individual sub-domain. The other 5 sub-domains consist of the Midwest, northeast, central, southeast, and a region that contains all other emissions not represented by the other 12 sub-domains. Sensitivities of all modeled species (gas phase, particle phase, and acid deposition) to SO2, ground-level NOx, elevated point NOx, and NH3 emissions (Table 12-2) from each of these 13 regions were calculated.

Figure 12-1

Source regions for which emission sensitivities were calculated. 12-1

ODMAN,

ET AL.

Table 12-1


List of states in each source region used for sensitivity analysis.

Region

States

SAMI

Alabama (AL), Georgia (GA), South Carolina (SC), North Carolina (NC), Tennessee (TN), Kentucky (KY), Virginia (VA), West Virginia (WV)

Midwest (MW) Northeast (NE)

Ohio, Michigan, Indiana, Illinois, Wisconsin Maryland, Delaware, District of Columbia, Pennsylvania, New Jersey, New York, Connecticut, Massachusetts, Rhode Island, Vermont, New Hampshire, Maine

Central (CN)

Louisiana, Arkansas, Missouri, Iowa, Minnesotaa, Texas a, Oklahomaa, Kansas a, Nebraskaa

Southeast (SE) All Others (AO)

Florida a, Mississippi South Dakotaa, North Dakota a, Canada, Major Bodies of Water

a

Entire state is not contained in the modeling domain

Recall from Figure 4-1 that the grid resolution coarsened outside the SAMI states. There were very few fine cells where the Northeast region borders the SAMI states. In general, grid-cell sizes over the regional sub-domains were 48-km or larger. Among all regional sub-domains, the Central region was modeled using the largest cell sizes. Large grid sizes dilute emissions by instantaneous mixing of plumes inside the grid cell leading, in general, to higher local concentrations of secondary pollutants and less transport of precursors. Because of this grid-resolution issue, the most reliable sourcereceptor relationships are likely those originating from the SAMI region and ending in the SAMI region, followed by those originating from the SAMI region and ending in the regional sub-domains. The least accurate would be the source-receptor relationships originating from the regional sub-domains and ending in the SAMI states. Since no special point-source treatment was used in this study, the impact of point-source plumes originating from regional sub-domains may be highly inaccurate. For example, elevated SO2 and NOx sources in the Central region may actually have a larger impact on sulfate and ozone concentrations at the receptors within the SAMI region than what is projected here. For small perturbations, one can assume linearity and use the DDM sensitivities in combination. For example, the sensitivities to emissions from 8 SAMI states were added to find the sensitivity to emissions from the SAMI region. Similarly, the sensitivities to emissions from 6 regions in Table 12-1 were added to find the sensitivity to domain-wide emissions. The sensitivities to different source types can also be added. For example, the sensitivities of ozone to elevated and ground-level NOx sources can be added to find the sensitivity to total NOx sources. One can attempt to calculate the sensitivities for more complex combinations such as the sensitivity of fine sulfate PM to different amounts of different emission types (e.g., SO2, NOx , NH3) from different states and regions. However, the result would be valid only if the linearity assumption still holds under the selected combination.

Table 12-2

Emission types and species list for which the sensitivities to the emissions were analyzed.

Emission Type

Analyzed Species

SO2

Particulate sulfate, nitrate, ammonium, and PM2.5 Wet sulfate

Elevated NOx

Ozone Particulate nitrate, ammonium, organics, and PM2.5 Wet nitrate and ammonium

Ground-level NOx

Ozone Particulate nitrate, ammonium, organics, and PM2.5 Wet nitrate and ammonium

Ammonia

Particulate sulfate, nitrate, ammonium, and PM2.5 Wet ammonium

12-2

ODMAN,

ET AL.


Providing validity bounds for the linearity assumption is beyond the scope of this report. Note that large or combinatory emission perturbations would eventually lead to nonlinearities at which point DDM results would cease to be reliable. Here, after some limited comparisons with the “brute-force” method, DDM results were scaled and presented for a 10% reduction in SO2 emissions. The sensitivity results here should not be used as a replacement for full-scale modeling in evaluating the effectiveness of a control strategy.

12.2 Annual and Seasonal Sensitivities Recall from Section 8.1 that seasonal or annual average metrics at each site is calculated using the following equation:

N

m average = ∑ i =1

wi mi 100

(12-1)

where, maverage is the seasonal or annual average pollutant value, N is the number of days or periods contributing to the metric, wi is the percent contribution to the seasonal and annual metric (from Table 8-2 and Table 8-3), and mi is the daily or weekly value of the metric. Similarly, the seasonal or annual average sensitivity of each metric at each site can be calculated as:

N

saverage = ∑ i =1

wi si 100

(12-2)

where, saverage is the seasonal or annual average pollutant sensitivity and s i is the daily or weekly value of the sensitivity. Also recall, that annual weights are only defined at two sites (GRSM and SHEN). However, these weights are applied to calculate seasonal ozone, annual average PM, and deposition at other Class I areas by assuming that sites south of the Virginia and Kentucky borders are represented by the weights at GRSM and sites north of the North Carolina and Tennessee borders are represented by the SHEN weights.

12.3 Aerosol Sensitivities 12.3.1 PM Sensitivities to SO 2 Emissions The sensitivity of domainwide daily average PM2.5 concentrations to a 10% reduction in total (elevated plus ground-level) SO2 emissions from each of the eight SAMI states and surrounding regions was mapped for each day contributing to the annual metric (i.e., IMPROVE days). Figure 12-1 shows the daily average PM2.5 (first frame in second row) and its change due to reductions from the SAMI states for July 15, 1995. On this particular day, the impact of emission reductions from many states are local except for Alabama and Georgia whose reductions benefit the neighboring states to the northwest. This is, in part, due to a high pressure system over the SAMI region and the anticyclone 2around it. Similar maps for other days both for PM2.5 and fine SO4 can be found in Appendix A along with maps of percentage change. The percentage change maps are helpful in determining the significance of the benefit relative to the 10% reduction. Although not shown here, maximum reduction 2in SO4 concentration anywhere in the SAMI region was less than 10% for a 10% SO2 emission reduction from any SAMI state.

12-3

ODMAN,

ET AL.

Figure 12-1


Daily average PM2.5 and its change on July 15, 1995 for a 10% reduction of the 2010-OTW SO2 emissions from SAMI states.

In addition to the spatial plots (maps), the changes in PM2.5 concentrations at a number of Appalachian Mountain sites were illustrated with greater detail. These sites are listed in Table 12-1 and their geographic locations are shown in Figure 12-2 (SAMI, 2001). Reductions in pollutant levels at each site resulting from a 10% reduction in SO2 emissions from each of the 13 sub-domains are discussed below. By examining sensitivities at a specific station, it is possible to determine the sub-domain from which emission reductions would have the greatest effect. Specifically, this discussion will focus on the sensitivity of fine sulfate, nitrate, ammonium, and PM2.5 at the ten Class I sites listed in Table 12-1 to SO2 emission reductions from the thirteen sub-domains (five regions and 8 SAMI states) listed in Table 12-1. Figure 12-3 and Figure 12-4 contain the daily averaged sulfate concentrations (right axis) and the daily average sulfate sensitivity (left axis) by class for each weighted IMPROVE day at the Great Smoky Mountains National Park (GRSM). In Figure 12-3, the sulfate sensitivities represent the absolute 3 change (µg/m ) in sulfate concentrations due to a 10% reduction in SO2 emission from each geographic sub-domain. The sensitivity to a 10% domain-wide emission reduction can be calculated by summing the sensitivities for each of the sub-domains. For example, if the meteorology of July 15, 1995 3 reoccurs in 2010, a 10% reduction in domain-wide SO2 emissions is estimated to result in a 1.07 µg/m 3 3 reduction in the sulfate concentration at GRSM. Of this, 0.34 µg/m is connected to TN, 0.19 µg/m to

12-4

ODMAN,

ET AL.

Table 12-1


Labels, and locations for receptors used in aerosol and wet deposition sensitivity analysis.

Site Name

Label

State

Sipsey Cohutta Joyce Kilmer Look Rocka, Great Smoky Mountains Elkmontb, Great Smoky Mountains Clingmans Dome b, Great Smoky Mountains Shining Rock Linville Gorge Jefferson/James River Face Shenandoah Otter Creek Dolly Sodds

SIPS COHU JOKM GRSM ELKM CLND SHRO LIGO JEFF SHEN OTRC DOSO

AL GA NC TN TN TN NC NC VA VA WV WV

a

Location of the IMPROVE monitor in the Great Smoky Mountains. This site will only be used to look at aerosol sensitivities. These sites are located in the Great Smoky Mountains and will only be used to look at wet deposition sensitivities. ELKM is a low elevation site and CLND is a high elevation site. b

3

GA, 0.15 µg/m to NC, and smaller fractions to the other ten sub-domains assuming they all reduced their SO2 emissions by 10%. In Figure 12-4, the sulfate sensitivities represent the percent change (%) in sulfate concentrations due to a 10% SO2 emission reduction from each geographic sub-domain. Again, the sensitivity to a 10% domain-wide emission reduction can be calculated by summing the sensitivities for each of the sub-domains. For example, if the meteorology of July 15, 1995 reoccurs in 2010, a 10%

Figure 12-2

Geographic location of the Class I receptor sites where aerosol and wet deposition sensitivity analysis was performed (reproduced from SAMI, 2001). 12-5

ODMAN,

ET AL.


Sulfate Sensitivity at GRSM to 10% SO2 Emission Reductions Class 2

Class 3

Class 4

Class 5

-1.6

16.0

-1.4

14.0

-1.2

12.0

-1.0

10.0

-0.8

8.0

-0.6

6.0

-0.4

4.0

-0.2

2.0

0.0

0.0 7/15/95

7/31/91

7/27/91

7/12/95

5/27/95

8/11/93

6/24/92

4/29/95

8/07/93

8/04/93

5/15/93

4/26/95

3/24/93

2/09/94

3/27/93

Figure 12-3

Sulfate Concentration ( g/m 3 )

Sulfate Sensitivity (µ g/m 3)

Class 1

AO SE NE MW CN WV VA TN SC NC KY GA AL SO4

Daily average fine sulfate concentrations (* ) and absolute sensitivities for each classified day at Great Smoky Mountains to a 10% reduction in SO2 emissions from each geographic sub-domain.

reduction in domain-wide SO2 emissions is estimated to result in an 8.8% reduction in the sulfate concentration at GRSM. Of this, 2.8% is connected to TN, 1.5% to GA, 1.2% to NC, and smaller fractions to the other ten sub-domains assuming they all reduced their SO2 emissions by 10%. At GRSM, it can be seen that different sub-domains can have varying contributions to the overall reduction of sulfate from day-to-day depending on the specific meteorology. Charts for other locations and other particulate species can be found in Appendix A. Figure 12-6 contains the annual average and class average sulfate concentrations and sensitivity contributions to sulfate reductions at GRSM. The contribution from TN is somewhat constant across classes, but the sensitivity to domain-wide reductions vary depending on the roles of the other

Sulfate Sensitivity at GRSM to 10% SO2 Emission Reductions Class 2

Class 3

Class 4

Class 5

-10.0

16.0

-9.0

14.0

-8.0

12.0

-7.0 -6.0

10.0

-5.0

8.0

-4.0

6.0

-3.0

4.0

-2.0 -1.0

2.0

0.0

0.0 7/15/95

7/31/91

7/27/91

7/12/95

5/27/95

8/11/93

6/24/92

4/29/95

8/07/93

8/04/93

5/15/93

4/26/95

3/24/93

2/09/94

3/27/93

Figure 12-4


Sulfate Sensitivity (%)

Class 1


Daily average fine sulfate concentrations (* ) and sensitivities for each classified day at Great Smoky Mountains to a 10% reduction in SO2 emissions from each geographic sub-domain. 12-6

ODMAN,

ET AL.


Sulfate Sensitivity at GRSM to 10% SO2 Emission Reductions

-8.0

12.0

-6.0

9.0

-4.0

6.0

-2.0

3.0

Sulfate Sensitivity (% )

15.0

0.0

0.0 Class 1

Figure 12-6


-10.0

Class 2

Class 3

Class 4

Class 5

Annual


Annual average fine sulfate concentrations (* ) and sensitivities for each class at Great Smoky Mountains to a 10% reduction in SO2 emissions from each geographic sub-domain.

12 sub-domains. It is clear that there are greater responses to SO2 emission reductions on days with the highest pollutant concentrations (Class 4 and 5 days), both in an absolute and relative sense. Charts for other locations can be found in Appendix A. Figure 12-5 contains the annual average fine sulfate concentrations and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from different sub-domains. The stations are geographically ordered from southwest to northeast. Sipsey, AL (SIPS) shows the greatest response to

Annual Sulfate Sensitivity to 10% SO2 Emission Reductions AO

8.0

-6.0

6.0

-4.0

4.0

-2.0

2.0

SE NE MW CN WV VA TN SC NC KY GA

3

Sulfate Concentration ( g/m )


-8.0

0.0

0.0 SIPS

Figure 12-5

COHU JOKM GRSM SHRO LIGO

JEFF

AL SO4

SHEN OTRC DOSO

Annual average fine sulfate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic sub-domain. 12-7

ODMAN,

ET AL.


Annual Nitrate Sensitivity to 10% SO2 Emission Reductions AO

1.2

5.00

1.0

4.00

0.8

3.00

0.6

2.00

0.4

1.00

0.2

3

Nitrate Concentration ( g/m )

Nitrate Sensitivity (% )

6.00

0.00

0.0 SIPS

Figure 12-7


JEFF

SE NE MW CN WV VA TN SC NC KY GA AL NO3

SHEN OTRC DOSO

Annual average fine nitrate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic sub-domain. Note, in this case nitrate increases as SO2 is decreased.

reductions in emissions from Alabama. The sites in North Carolina (JOKM, SHRO, LIGO) and Tennessee (GRSM) show the greatest response to emission reductions in TN. The sites in Virginia and West Virginia (JEFF, SHEN, OTRC, DOSO) show the greatest response to emission reductions in West Virginia and the Midwest sub-domain. Figure 12-7, Figure 12-8 and Figure 12-9, show the response of nitrate, ammonium, and PM2.5

Annual Ammonium Sensitivity to 10% SO2 Emission Reductions

-4.00

1.6

-3.00

1.2

-2.00

0.8

-1.00

0.4

0.00

0.0 SIPS

Figure 12-8

AO

2.0


JEFF

Ammonium Concentration ( µg/m3)

Ammonium Sensitivity (% )

-5.00

SE NE MW CN WV VA TN SC NC KY GA AL NH4

SHEN OTRC DOSO

Annual average fine ammonium concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic sub-domain. 12-8

ODMAN,

ET AL.


Annual Fine PM Sensitivity to 10% SO2 Emission Reductions AO

15.0

-2.0

10.0

-1.0

5.0

3

Fine PM Concentration ( g/m )

Fine PM Sensitivity (% )

-3.0

0.0

0.0 SIPS

Figure 12-9


JEFF

SE NE MW CN WV VA TN SC NC KY GA AL PMF

SHEN OTRC DOSO

Annual average fine PM concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from each geographic sub-domain. 3

3

to reductions in SO2 emissions. Nitrate increases between 1.5% and 4.0% (0.01 µg/m to 0.04 µg/m ) 3 3 and ammonium decreases between 2.0% and 4.0% (0.03 µg/m to 0.08 µg/m ). The increase in nitrates is due to the increase in free ammonia gas that becomes available when sulfate concentrations decrease. Ammonium does not decrease as much as sulfate because of the increase in of ammonium 3 3 nitrate. Typical PM2.5 reductions range from 1.7% to 2.7% (0.15 µg/m to 0.30 µg/m ) because approximately half of the PM2.5 contains species that are not affected by reductions in SO2 emissions. A normalized sensitivity can be calculated by dividing the percent reduction from a specific sub-domain by the total percent reduction from the entire domain, indicating the fractional contribution of the sub-domain to the domain-wide sensitivity. Figure 12-1 shows the normalized annual average fine PM concentrations and sensitivities for ten Class I areas to a 10% reduction in SO2 emissions from different sub-domains. Although not shown here (see the charts in Appendix A), the normalized annual average fine sulfate, nitrate, and ammonium sensitivities show nearly identical regional distributions. 12.3.2 PM Sensitivities to Elevated NOx Emissions The sensitivity of domainwide daily average PM2.5 concentrations to a 10% reduction in elevated NOx emissions from each of the eight SAMI states and surrounding regions was mapped for each simulated IMPROVE day. Figure 12-2 shows the percent change in daily average PM2.5 concentrations due to reductions from all SAMI states for July 15, 1995. The impact of reductions is very small and on this particular day it does not exceed 0.14%. Even though this is a reduction from the entire region its impact is minute in comparison to the impact of SO2 reductions from any single state. This is due to the fact that particulate nitrate is not as affected by NOx reductions as particulate sulfate is from SO2 reductions and that particulate nitrate constitutes a small portion of the PM2.5 in the region. Similar maps exist for all simulated IMPROVE days; they can be found in Appendix A along with maps of absolute change.

12-9

ODMAN,

ET AL.


20.0

-0.9

18.0

-0.8

16.0

-0.7

14.0

-0.6

12.0

-0.5

10.0

-0.4

8.0

-0.3

6.0

-0.2

4.0

-0.1

2.0

0.0

0.0 SIPS COHU JOKM GRSM SHRO LIGO

Figure 12-1

AO

3

-1.0


Fine PM Sensitivity

Annual Fine PM Sensitivity to 10% SO2 Emission Reductions SE NE MW CN WV VA TN SC NC KY GA AL PMF

JEFF SHEN OTRC DOSO

Annual average fine PM concentrations (* ) and normalized sensitivities for ten Class I areas to a 10% reduction in SO 2 emissions from each geographic sub-domain.

In addition to maps that show the spatial distribution of the changes over the domain, the changes in daily average concentrations of fine particulate nitrate, fine particulate ammonium, and PM2.5 concentrations at the 10 sites listed in Table 12-1 were illustrated by charts similar to those in

Figure 12-2

Percent change in daily average PM2.5 on July 15, 1995 for a 10% reduction of the 2010-OTW elevated NOx emissions from eight SAMI states. 12-10

ODMAN,

ET AL.


Figure 12-3 and Figure 12-4. In these charts, the sensitivities to 10% reductions in elevated NOx emissions from the thirteen sub-domains (5 regions and 8 SAMI states) are stacked in columns. This way, one can see how large each sub-domain’s contribution is to the domain wide sensitivity or compare one sub-domain’s contribution to another. Two sets of charts were prepared: one for absolute sensitivities and another for percentage sensitivities. These charts can be found in Appendix A. Figure 12-3 contains the annual average PM2.5 concentrations and sensitivities for ten Class I areas to a 10% reduction in elevated NOx emissions from different geographic sub-domains. The receptors in AL, GA, NC, and TN showed reductions of between 0.15% and 0.25%, while receptors in WV and VA show reductions between 0.30% and 0.45%. Although not shown here, annual average 3 3 PM concentrations showed a decrease of 1.5% to 2.5% (0.01 µg/m to 0.03 µg/m ) for nitrate, 0.2% to 3 3 3 3 0.4% (0.003 µg/m to 0.007 µg/m ) for ammonium, 0.15% to 0.25% (0.003 µg/m to 0.005 µg/m ) for organics, and an insignificant response for sulfate. 12.3.3 PM Sensitivities to Ground-Level NO x Emissions The sensitivity of domainwide daily average PM2.5 concentrations to a 10% reduction in ground-level NOx emissions from each of the eight SAMI states and surrounding regions was mapped for each simulated IMPROVE day. Figure 12-1 shows the percent change in daily average PM2.5 concentrations due to reductions from all SAMI states for July 15, 1995. Note that the scale of the change shown on this map is five times larger than the scale of the change shown in Figure 12-2, yet the perceived change is about the same (actually larger). This means that a 10% reduction in groundlevel NOx emissions is at least five times more beneficial in reducing PM2.5 than a 10% reduction in elevated NOx emissions. However, on this particular day, the impact of a 10% emission reduction still does not exceed a 0.8% reduction in PM2.5. Maps exist for all simulated IMPROVE days; they can be found in Appendix A along with maps of absolute change. In addition to maps that show the spatial distribution of the changes over the domain, the changes in daily average concentrations of fine particulate nitrate, fine particulate ammonium, and PM2.5 concentrations at the 10 sites listed in Table 12-1 were illustrated by charts similar to those in

Annual Fine PM Sensitivity to 10% Elevated NOx Emission Reductions AO

15.0

-0.4

12.0

-0.3

9.0

-0.2

6.0

-0.1

3.0

3



-0.5

0.0

0.0 SIPS

Figure 12-3


JEFF


SHEN OTRC DOSO

Annual average fine PM2.5 concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domain. 12-11

ODMAN,

ET AL.

Figure 12-1


Percent change in daily average PM2.5 on July 15, 1995 for a 10% reduction of the 2010-OTW ground-level NOx emissions from eight SAMI states.

Figure 12-3 and Figure 12-4. In these charts, the sensitivities to 10% reductions in ground level NOx emissions from the thirteen sub-domains (5 regions and 8 SAMI states) are stacked in columns. This way, one can see how large each sub-domain’s contribution is to the domain wide sensitivity or compare one sub-domain’s contribution to another. Two sets of charts were prepared: one for absolute sensitivities and another for percentage sensitivities. These charts can be found in Appendix A. Figure 12-2 contains the annual average PM2.5 concentrations and sensitivities for ten Class I areas to a 10% reduction in ground level NOx emissions from different geographic sub-domains. The 3 3 receptors show reductions of between 0.30% and 0.45% (0.03 µg/m to 0.05 µg/m ). Although not 3 shown here, annual average PM concentrations showed a decrease of 2.0% to 3.0% (0.020 µg/m to 3 3 3 0.025 µg/m ) for nitrate, 0.3% to 0.5% (0.005 µg/m to 0.008 µg/m ) for ammonium, 0.35% to 0.50% 3 3 (0.005 µg/m to 0.015 µg/m ) for organics, and an insignificant response for sulfate. 12.3.4 PM Sensitivities to NH 3 Emissions No spatial maps were prepared for the sensitivity of PM to NH3 emission. The changes in daily average concentrations of fine particulate sulfate, fine particulate nitrate, fine particulate ammonium, and PM2.5 at the 10 sites listed in Table 12-1 were illustrated by charts similar to those in Figure 12-3 and Figure 12-4. In these charts, the sensitivities to 10% reductions in NH3 emissions from the thirteen sub-domains (5 regions and 8 SAMI states) are stacked in columns. This way, one can see how large each sub-domain’s contribution is to the domain wide sensitivity or compare one sub-domain’s contribution to another. Two sets of charts were prepared: one for absolute sensitivities and another for percentage sensitivities. These charts can be found in Appendix A.

12-12

ODMAN,

ET AL.


Annual Fine PM Sensitivity to 10% Ground Level NOx Emission Reductions AO

15.0

-0.4

12.0

-0.3

9.0

-0.2

6.0

-0.1

3.0

3



-0.5

0.0

0.0 SIPS

Figure 12-2


JEFF


SHEN OTRC DOSO

Annual average fine PM concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in ground level NOx emissions from each geographic sub-domain.

Figure 12-1, Figure 12-3, Figure 12-4, and Figure 12-1 contain the annual average sulfate, nitrate, and ammonium, and PM2.5 concentrations and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from different geographic sub-domains. Annual average PM concentrations 3 3 showed a decrease of up to 1.4% (0.06 µg/m ) for sulfate, between 1.5% to 3.5% (0.015 µg/m to 3 3 3 0.030 µg/m ) for nitrate, between 1.0% and 2.0% (0.015 µg/m to 0.035 µg/m ) for ammonium, and

Annual Sulfate Sensitivity to 10% NH3 Emission Reductions AO

5.0

-1.6

4.0

-1.2

3.0

-0.8

2.0

1.0

0.0

0.0


3

-0.4

Sulfate Concentration ( g/m )


-2.0

SIPS

Figure 12-1


JEFF

AL SO4

SHEN OTRC DOSO

Daily average fine sulfate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH 3 emissions from each geographic sub-domain. 12-13

ODMAN,

ET AL.


Annual Nitrate Sensitivity to 10% NH3 Emission Reductions AO

1.2

-3.00

0.9

-2.00

0.6

-1.00

0.3

3

Nitrate Concentration ( g/m )


-4.00

0.00

0.0 SIPS

Figure 12-3


JEFF

SE NE MW CN WV VA TN SC NC KY GA AL NO3

SHEN OTRC DOSO

Daily average fine nitrate concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domain. 3

3

between 0.4% and 1.1% (0.040 µg/m to 0.12 µg/m ) for PM2.5. Although not shown here, organics show an insignificant response for NH3 emission reductions.

Annual Ammonium Sensitivity to 10% NH3 Emission Reductions

-2.00

1.5

-1.00

1.0

0.00

0.5

1.00

0.0 SIPS

Figure 12-4

AO

2.0


JEFF

Ammonium Concentration ( µg/m3)


-3.00


SHEN OTRC DOSO

Daily average fine ammonium concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domain. 12-14

ODMAN,

ET AL.


16.0

-1.2

14.0

-1.0

12.0

-0.8

10.0

-0.6

8.0

-0.4

6.0

-0.2

4.0

0.0

2.0

0.2

0.0 SIPS

Figure 12-1


JEFF

AO

3

-1.4



Annual Fine PM Sensitivity to 10% NH3 Emission Reductions SE NE MW CN WV VA TN SC NC KY GA AL PMF

SHEN OTRC DOSO

Daily average fine PM concentrations (* ) and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domain.

12.4 Wet Deposition Sensitivities 12.4.1 Wet Deposition Sensitivities to SO 2 Emissions The sensitivity of domainwide cumulative wet deposition of sulfate to a 10% reduction in total (ground-level plus elevated) SO2 emissions from each of the eight SAMI states and surrounding regions was mapped for each episode. Figure 12-2 shows the cumulative wet deposition of sulfate and its change due to reductions from the SAMI states during the week of July 11-18, 1995. On this particular episode, the impact of emission reductions on wet deposition is mostly local. In other words, the greatest benefits are observed within the boundaries of the state reducing its SO2 emissions. Similar maps for other episodes can be found in Appendix A. No percentage change maps were prepared because the percentage change is usually greatest for locations receiving the smallest amount of rain. Figure 12-2 shows the changes in weekly cumulative wet deposition flux of sulfate at Elkmont, which is in the Great Smoky Mountains National Park, for all the episodes contributing to the annual metric. In this figure, the changes due to 10% reductions in SO2 emissions from the thirteen subdomains (5 regions and 8 SAMI states) are stacked in columns. This way, each sub-domain’s contribution to the domain wide sensitivity can be seen easily and compared to another state’s. For example, the contribution of SO2 emission reductions from Tennessee is the largest for the July 91 episode, when sulfate wet deposition was the largest. For the May 1993 and May 1995 episodes, on the other hand, the contribution of SO2 emission reductions from Georgia is the largest. Similar charts for sulfate, nitrate, and ammonium wet deposition sensitivities for the 11 sites listed in Table 12-1 can be found in Appendix A. Two sets of charts exist: one for absolute sensitivities and another for percentage sensitivities.

12-15

ODMAN,

ET AL.

Figure 12-2


Cumulative wet deposition of SO42- and its change for the week of July 11-18, 1995 for a 10% reduction of the 2010-OTW total SO2 emissions from SAMI states.

Annual average sulfate wet deposition sensitivities to SO2 emission reductions at eleven Class I areas (listed in Table 12-1) were calculated. As shown in Figure 12-1, a 10% reduction in SO2 2 emissions results in sulfate wet deposition decreasing between 4.0% and 9.2% (2.0 mg/m to 6.0 2 mg/m ). There were some notable differences between the sensitivities of particulate sulfate and sulfate wet deposition sensitivities in the way they responded to SO2 emission reductions from different geographic regions. In general, the sulfate wet deposition responses were more localized (compare Figure 12-1 to Figure 12-5). For example, the sites in Alabama and Georgia (SIPS, COHU) show relatively larger responses to SO2 reductions in AL and GA. A similar trend is observed for the VA and WV sites. However, the sites in North Carolina (JOKM, SHRO, LIGO) and Tennessee (CLND, ELKM) show significantly larger responses to emission reductions from AL and GA. In fact, the fraction of sulfate wet deposition attributed to SO2 emissions from AL and GA exceed those connected to SO2 emissions from TN. Recall that SO2 emissions from TN had the dominant impact on particulate sulfate concentrations. This suggests that most of the rain at the TN and NC sites can be attributed to air masses coming from the Gulf of Mexico. Although not shown here, annual average wet deposition mass fluxes show an insignificant response for nitrate and ammonium.

12-16

ODMAN,

ET AL.

Figure 12-2


Weekly cumulative sulfate wet deposition fluxes (* ) and sensitivities for each classified episode at Elkmont (Great Smoky Mountains) to a 10% reduction in SO2 emissions from each geographic sub-domain.

12.4.2 Wet Deposition Sensitivities to Elevated NOx Emissions The sensitivity of domainwide cumulative wet deposition of nitrate to a 10% reduction in elevated NOx emissions from each of the eight SAMI states and surrounding regions was mapped for each episode. Figure 12-3 shows the cumulative wet deposition of nitrate and its change due to reductions from the SAMI states during the week of July 11-18, 1995. On this particular episode, the greatest benefits (i.e., reductions in NO3 wet deposition) are observed within the boundaries of the

Annual Sulfate Sensitivity to 10% SO2 Emission Reductions AO

100.0

-8.0

80.0

-6.0

60.0

-4.0

40.0

20.0

0.0

0.0


2

-2.0

Sulfate Concentration (mg/m )


-10.0

AL SO4

SIPS COHU JOKM CLND ELKM SHRO LIGO JEFF SHEN OTRC DOSO

Figure 12-1

Annual weekly average sulfate wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in SO 2 emissions from each geographic sub-domains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park. 12-17

ODMAN,

ET AL.

Figure 12-3


Cumulative wet deposition of NO3- and its change for the week of July 11-18, 1995 for a 10% reduction of the 2010OTW elevated NOx emissions from SAMI states.

state reducing its NOx emissions with a few exceptions. For example, the impacts of reductions in Kentucky, West Virginia and Tennessee are large along the eastern flank of the Appalachians. The impact of Alabama is primarily over southern Georgia and the Florida panhandle. This behavior is due to the spatially scattered nature of the summertime thunderstorm activities. Similar maps for other episodes can be found in Appendix A. No percentage change maps were prepared because the percentage change is usually greatest for locations receiving the smallest amount of rain. In addition to maps that show the spatial distribution of the changes over the domain, the changes in weekly cumulative wet deposition fluxes of nitrate and ammonium at the 11 sites listed in Table 12-1 were illustrated by charts similar to the one in Figure 12-2. In these charts, the sensitivities to 10% reductions in elevated NOx emissions from the thirteen sub-domains (5 regions and 8 SAMI states) are stacked in columns. This way, one can see how large each sub-domain’s contribution is to the domain wide sensitivity or compare one sub-domain’s contribution to another. Two sets of charts were prepared: one for absolute sensitivities and another for percentage sensitivities. These charts can be found in Appendix A. Figure 12-1 contains the annual average nitrate wet deposition fluxes and sensitivities for eleven Class I areas to a 10% reduction in elevated NOx emissions from different geographic subdomains. The receptors in GA, NC, and TN show reductions between 0.6% and 1.0%, while receptors 12-18

ODMAN,

ET AL.


Annual Nitrate Sensitivity to 10% Elevated NOx Emission Reductions

-1.50

50.0


60.0

-1.20

40.0

-0.90

30.0

-0.60

20.0

10.0

0.00

0.0 SIPS COHU JOKM CLND ELKM SHRO LIGO

Figure 12-1


2

-0.30

AO

Nitrate Concentration (mg/m )

-1.80

AL NO3

JEFF SHEN OTRC DOSO

Annual weekly average nitrate wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park.

in WV and VA show reductions between 1.2% and 1.7%. Compared to particulate nitrate sensitivities (see Appendix A), the relative benefits (i.e., reductions in nitrate wet deposition) attributed to the emissions reductions in Alabama and Georgia are greater than those attributed to TN, at the TN and NC sites. There is also a significant increase in the relative benefit at SIPS and COHU that is attributed to elevated NOx emission reductions in the Central region. Although not shown here, annual average wet deposition fluxes of sulfate and ammonium show an insignificant response to elevated NOx emission reductions. 12.4.3 Wet Deposition Sensitivities to Ground-Level NOx Emissions The sensitivity of domainwide cumulative wet deposition of nitrate to a 10% reduction in ground-level NOx emissions from each of the eight SAMI states and surrounding regions was mapped for each episode. Figure 12-2 shows the cumulative wet deposition of nitrate and its change due to reductions from the SAMI states during the week of July 11-18, 1995. Note that the scale of the change shown in this figure is twice the scale of the change shown in Figure 12-3 but that the areas of impact are very similar. This suggests that the impact of 10% reductions in ground-level emissions on NO3 wet deposition is almost twice the impact of 10% reductions in elevated emissions on this particular episode. There are some notable exceptions such as West Virginia where reducing elevated emissions by 10% seems to have a greater impact, and Virginia where the benefit from ground-level emissions may be more than twice as beneficial as reducing the elevated emissions by the same fraction. However, recall that these reductions do not correspond to the same amount in tons of NOx and that a 10% reduction in elevated emissions from West Virginia may correspond to a much larger amount than a 10% reduction in ground-level emissions. Similar maps for other episodes can be found in Appendix A. Once again, no percentage change maps were prepared because the percentage change is usually greatest for locations receiving the smallest amount of rain.

12-19

ODMAN,

ET AL.

Figure 12-2


Cumulative wet deposition of NO3- and its change for the week of July 11-18, 1995 for a 10% reduction of the 2010OTW ground-level NOx emissions from SAMI states.

In addition to maps that show the spatial distribution of the changes over the domain, the changes in weekly cumulative wet deposition fluxes of nitrate and ammonium at the 11 sites listed in Table 12-1 were illustrated by charts similar to the one in Figure 12-2. In these charts, the sensitivities to 10% reductions in ground level NOx emissions from the thirteen sub-domains (5 regions and 8 SAMI states) are stacked in columns. This way, one can see how large each sub-domain’s contribution is to the domain wide sensitivity or compare one sub-domain’s contribution to another. Two sets of charts were prepared: one for absolute sensitivities and another for percentage sensitivities. These charts can be found in Appendix A. Figure 12-3 contains the annual average nitrate wet deposition fluxes and sensitivities for eleven Class I areas to a 10% reduction in ground level NOx emissions from different geographic subdomains. The receptors in GA, NC, and TN show reductions of between 1.1% and 1.5%, while receptors in WV and VA show reductions between 1.3% and 1.8%. Note that these are larger reductions than those obtained as a result of reducing the elevated NOx emissions by the same percentage. The contributions from different geographic regions are also different (compare Figure 12-3 to Figure 12-1). Comparing the responses to ground level NOx emission reductions with those to elevated NOx emission reductions at the AL, GA, TN and NC sites, the fractions attributed to the Central region shrunk in favor of the fractions attributed to TN and GA. At the VA and WV sites, the responses to a 10% reduction in ground level NOx emissions from WV were much smaller than those 12-20

ODMAN,

ET AL.


Annual Nitrate Sensitivity to 10% Ground Level NOx Emission Reductions

-1.50

50.0


60.0

-1.20

40.0

-0.90

30.0

-0.60

20.0

10.0

0.00

0.0 SIPS COHU JOKM CLND ELKM SHRO LIGO

Figure 12-3


2

-0.30

AO

Nitrate Concentration (mg/m )

-1.80

AL NO3

JEFF SHEN OTRC DOSO

Annual weekly average nitrate wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in ground level NOx emissions from each geographic sub-domains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park.

to a reduction by the same percentage amount of elevated NOx emissions from the same state. A 10% reduction in elevated emissions from WV corresponds to a much larger mass reduction (tons of NOx ) than a 10% reduction in ground-level emissions in this state. Although not shown here, annual average wet deposition mass fluxes of sulfate and ammonium show an insignificant response to elevated NOx emission reductions. -

The response of NO3 wet deposition to a 10% reduction in NOx emissions is generally smaller than what is found in other studies (e.g. Shin and Carmichael, 1992). The reason for this low response is currently being investigated. 12.4.4 Wet Deposition Sensitivities to NH3 Emissions No spatial maps were prepared for the sensitivity of wet deposition to NH3 emission. The changes in weekly cumulative wet deposition fluxes of nitrate and ammonium at the 11 sites listed in Table 12-1 were illustrated by charts similar to the one in Figure 12-2. In these charts, the sensitivities to 10% reductions in NH3 emissions from the thirteen sub-domains (5 regions and 8 SAMI states) are stacked in columns. This way, one can see how large each sub-domain’s contribution is to the domain wide sensitivity or compare one sub-domain’s contribution to another. Two sets of charts were prepared: one for absolute sensitivities and another for percentage sensitivities. These charts can be found in Appendix A. Figure 12-1 contains the annual average ammonium wet deposition fluxes and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from different geographic sub-domains. Annual 2 2 average wet deposition fluxes show a decrease between 1.0% and 1.3% (0.2 mg/m to 0.5 mg/m ) for ammonium. Although not shown here, sulfate and nitrate wet deposition show an insignificant response for NH3 emission reductions. However, recall from Table 11-1 that the sensitivities of nitrate wet deposition to NH3 emissions was rated “POOR”. This subjective rating was applied because of a severe underestimation, by a factor of 20, of the sensitivity by the DDM-3D method (see Appendix D). Therefore, the actual nitrate wet deposition response to NH3 emissions may be comparable in 12-21

ODMAN,

ET AL.


Annual Ammonium Sensitivity to 10% NH3 Emission Reductions AO

50.0

-1.20

40.0

-0.90

30.0

-0.60

20.0

-0.30

10.0

0.00

0.0

Ammonium Concentration (mg/m2 )


-1.50


SIPS COHU JOKM CLND ELKM SHRO LIGO JEFF SHEN OTRC DOSO

Figure 12-1

Annual weekly average ammonium wet deposition levels and sensitivities for ten Class I areas to a 10% reduction in NH3 emissions from each geographic sub-domains. Note, Clingmans Dome (CLND) and Elkmont (ELKM) are both in the Great Smoky Mountains National Park.

magnitude to its response to NOx emissions reductions. Further, if NH3 emissions are increased, the benefit from NOx emissions reductions may reduce or even cancel out.

12.5 Ozone Sensitivities The sensitivity of ozone W126 to 10% reductions in elevated and ground level NOx emissions from each of the eight SAMI states and surrounding regions were calculated for each day contributing to the seasonal metric. Recall from Section 8.2 that W126 is a non-linear function of ozone concentration. For this reason, the following procedure was used to calculate ozone W126 sensitivities. unperturbed First, using the unperturbed hourly ozone concentrations, c i , in Equations 8-3 and 8-4, unperturbed W126 , the unperturbed daily cumulative ozone W126, was calculated. Then using the hourly ozone sensitivities, s i , that are estimated by the DDM-3D module, perturbed hourly ozone concentrations were calculated as:

c iperturbed = ciunperturbed + s i

(12-3) perturbed

Using the perturbed hourly ozone concentrations in Equations 8-3 and 8-4, W126 , the perturbed daily cumulative ozone W126, was calculated. The sensitivity of daily cumulative ozone W126 was defined as:

S Wi 126 = W 126 perturbed − W 126 unperturbed Finally, using Si was calculated.

W126

(12-4)

in Equation 12-2, a seasonal average sensitivity of daily cumulative ozone W126

12-22

ODMAN,

ET AL.

Figure 12-1


O3 (W126) and its change on July 12, 1995 for a 10% reduction of the 2010-OTW elevated NOx emissions from SAMI states.

12.5.1 Ozone Sensitivities to Elevated NOx Emissions The sensitivity of daily cumulative ozone W126 to a 10% reduction in elevated NOx emissions from each of the eight SAMI states and surrounding regions was mapped for each day contributing to the seasonal metric. Figure 12-1 shows the cumulative ozone W126 and its change due to reductions from the SAMI states for July 12, 1995. On this particular day, the impact of emission reductions are mostly local (i.e. the greatest benefits are observed within the boundaries of the state reducing its NOx emissions). Similar maps for other days can be found in Appendix A along with maps of percentage change. The percentage change maps are helpful in determining the significance of the benefit relative to the actual metric, i.e., daily cumulative ozone W126. In addition to the spatial plots (maps), the changes in cumulative ozone W126 at a number of Appalachian Mountain sites were illustrated with greater detail. For each site listed in Table 12-1, charts were prepared showing the sensitivities of daily cumulative W126 for each day contributing to the seasonal metric. The sensitivities to a 10% reduction in elevated NOx emissions from the thirteen subdomains (5 regions and 8 SAMI states) were stacked in bars representing the domainwide sensitivity. For convenience, days were grouped by ozone class (Table 8-2). Since there are too many ozone days, the charts were broken into two groups: one for Class 1 and 2 days and another for Class 3 and

12-23

ODMAN,

ET AL.


Table 12-1

Labels, and locations for receptors used in ozone sensitivity analysis.

Site Name

Label

State

Sipsey Cohutta Dawsonville Coweeta Cranberry Joyce Kilmer Linville Gorge Shining Rock Table Rock Long Creek Look Rock Speedwell Big Meadows Horton Station Jefferson Sawmill Run Bearden Knob Parsons

SIPS COHU DAWS COWT CRAN JOKM LIGO SHRO TBLR LONG LOOK SPDW BIGM HORT JEFF SAWM BEAR PARS

AL GA GA NC NC NC NC NC NC SC TN TN VA VA VA VA WV WV

4 days. No example of these charts is given here but the interested reader is referred to Appendix A. There are two sets of charts: one for absolute sensitivities and another for percentage sensitivities. Figure 12-2 contains the seasonal average ozone W126 and sensitivities for 18 Class I areas to a 10% reduction in elevated level NOx emissions from different geographic sub-domains. Overall, there is a reduction of 3.0% to 3.5% (0.002 ppm-hr to 0.004 ppm-hr) at all sites. It should be noted that many geographic sub-domains (e.g. Midwest and Central) can have a significant contribution to multiple sites. This would indicate that ozone W126 is less local than PM and wet deposition and that transport is more important.

AO

-5.0

0.150

-4.0

0.120

-3.0

0.090

-2.0

0.060

-1.0

0.030

0.0

0.000 BRKB, WV

PARS, WV

SNP2, VA

SNP1, VA

JEFF, VA

HORT, VA

SPED, TN

GRSM, TN

CRAN, NC

LIGO, NC

TBLR, NC

SHRO, NC

COWE, NC

JOKM, NC

LONG, SC

COHU, GA

DAWS, GA

SIPS, AL

Figure 12-2

Ozone W126 (ppm-hr)

W126 Sensitivity (%)

Seasonal Ozone W126 to 10% Elevated NOx Emission Reductions

SE NE MW CN WV VA TN SC NC KY GA AL O3

Seasonal average ozone W126 (* ) and sensitivities for 18 Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domain. 12-24

ODMAN,

ET AL.

Figure 12-1


O3 (W126) and its change on July 12, 1995 for a 10% reduction of the 2010-OTW ground-level NOx emissions from SAMI states

12.5.2 Ozone Sensitivities to Ground-Level NO x Emissions The sensitivity of daily cumulative ozone W126 to a 10% reduction in ground-level NOx emissions from each of the eight SAMI states and surrounding regions was also mapped for each day contributing to the seasonal metric. Figure 12-1 shows the cumulative ozone and its change due to reductions from the SAMI states for July 12, 1995. Note that the scale of the change in this figure is three times the scale of Figure 12-1. In comparison, the impact of an equal percentage reduction in ground-level NOx emissions is larger than the impact of elevated NOx emissions. The difference between the impacts of ground-level and elevated NOx is smallest in West Virginia. Note, however, that while the reduction is the same on a percentage basis, the amount of reduction in tons of NOx is not the same. Recall that under the OTW strategy, the fraction of electric generating utilities and other point sources was only 37% of anthropogenic NOx emissions from the eight SAMI states (Table 9-1). The greatest impacts are once again observed in the states where NOx emissions are reduced. Absolute sensitivity maps for other days as well as percentage change maps can be found in Appendix A. Although not shown here, the maximum reduction in cumulative ozone (W126) is less than 10% for any location as a result of 10% NOx emission reductions from any state. Figure 12-1 contains the seasonal average daily cumulative ozone W126 and its sensitivities at 18 Class I areas to a 10% reduction in ground level NOx emissions from different geographic sub12-25

ODMAN,

ET AL.


AO

-10.0

0.150

-8.0

0.120

-6.0

0.090

-4.0

0.060

-2.0

0.030

0.0

0.000 BRKB, WV

PARS, WV

SNP2, VA

SNP1, VA

JEFF, VA

HORT, VA

SPED, TN

GRSM, TN

CRAN, NC

LIGO, NC

TBLR, NC

SHRO, NC

COWE, NC

JOKM, NC

LONG, SC

COHU, GA

DAWS, GA

SIPS, AL

Figure 12-1

Ozone W126 (ppm-hr)

W126 Sensitivity (%)

Seasonal Ozone W126 to 10% Ground Level NOx Emission Reductions

SE NE MW CN WV VA TN SC NC KY GA AL O3

Seasonal average ozone W126 (* ) and sensitivities for 18 Class I areas to a 10% reduction in elevated NOx emissions from each geographic sub-domain.

domains. Overall, there is a reduction of 7.0% to 8.0% (0.005 ppm-hr to 0.009 ppm-hr) at all sites. Again, it should be noted that outer regions, in particular CN, MW and NE, have a significant contribution at many of the sites. However, the relative contribution from distant sub-domains are not as large as those of elevated NOx emissions, indicating that the transport of ground level NOx is not as significant.

12.6 Summary of Sensitivity Analysis The following is a summary of the response of the various pollutants to reductions of SO2, elevated NOx, ground level NOx, and NH3 emissions. Table 12-1 and Table 12-2 contain the average absolute and percent changes in pollutant levels at Class I areas of the SAMI region due to a 10% reduction in SO2, elevated NOx, ground level NOx, and NH3 emissions. These tables can be useful in Table 12-1

Average sensitivity (absolute) of ozone W126, PM, and wet deposition at Class I areas to a 10% reduction in SO2, elevated NOx, ground level NOx, and NH 3 emissions. Emissions

Species

SO2

Elevated NOx

Ground NOx

NH3

Ozone (ppm-hr)

0.000

-0.003

-0.007

0.000

-0.200

+0.001

-0.001

-0.030

Particulate NO3 (µg/m )

+0.025

-0.020

-0.022

-0.020 *

Particulate NH4+ (µg/m3)

-0.050

-0.005

-0.006

-0.025

Particulate OC (µg/m3)

0.000

-0.004

-0.008

0.000

PM2.5 (µg/m3)

-0.250

-0.025

-0.035

-0.080 *

Wet Deposition SO42(mg/m2)

-4.00

+0.01

+0.02

+0.05

Wet Deposition NO3(mg/m2)

+0.05 *

-0.35

-0.50

-0.04 *

Wet Deposition NH4+ (mg/m2)

-0.07 *

-0.01 *

-0.01 *

-0.32 *

Particulate SO42- (µg/m3) -

3

* DDM-3D was rated “POOR” for these sensitivities. Actual (i.e., brute-force) sensitivities may be significantly larger.

12-26

ODMAN,

ET AL.

Table 12-2


Average sensitivity (%) of ozone W126, PM, and wet deposition at Class I areas to a 10% reduction in SO2, elevated NOx, ground level NOx, and NH 3 emissions. Emissions

Species

SO2

Elevated NOx

Ground NOx

Ozone W126 (%)

0.0

-3.0

-7.0

0.0

-5.00 +2.50 -3.00 0.00 -2.20

+0.03 -2.00 -0.30 -0.20 -0.28

-0.03 -2.20 -0.40 -0.40 -0.40

-0.80 -2.50 * -1.50 0.00 -0.80 *

-7.00 +0.25 * -0.22 *

+0.01 -1.00 -0.03 *

+0.03 -1.30 -0.04 *

+0.08 -0.10 * -1.10 *

SO42- – PM (%) NO3- – PM (%) NH4+ – PM (%) OC – PM (%) PM2.5 (%) SO42- – wet dep. (%) NO3- – wet dep. (%) NH4+ – wet dep. (%)

NH3

* DDM-3D was rated “POOR” for these sensitivities. Actual (i.e., brute-force) sensitivities may be significantly larger.

evaluating the relative importance of emission reductions on various pollutant levels. SO2 emission reductions can have a significant impact on the concentration of sulfate, nitrate, and ammonium PM and sulfate wet deposition. As sulfate decreases, more ammonia becomes available to react with HNO3 and form nitrate, hence particulate nitrate concentrations increase. SO2 emission reductions do not significantly impact the concentrations of ozone, organic PM, nitrate wet deposition, or ammonium wet deposition. Reductions of elevated and ground level NOx emissions can impact the concentration of ozone, particulate nitrate, particulate ammonium, organic PM, and nitrate wet deposition. NOx emission reductions do not significantly impact the concentrations of particulate sulfate, sulfate wet deposition, or ammonium wet deposition. In general, ground level NOx emission reductions have a greater impact on the pollutant levels than the elevated NOx emission reductions. For example, reductions in ground-level NOx emissions result in a 7% reduction of ozone W126 while reductions in elevated NOx emissions result in a 3% reduction of ozone W126. The differences are not as large for the PM and wet deposition sensitivities. Note that the average response of ozone W126 at class I areas to a 10% reduction in total (elevated plus ground level) NOx emissions is a 10% (= 7% + 3%) reduction. Reductions of NH3 emission can have a significant impact on the concentration of particulate sulfate, particulate nitrate, particulate ammonium, and ammonium wet deposition. NH3 emission reductions do not significantly impact the concentrations of ozone, organic PM, sulfate wet deposition, or nitrate wet deposition. However, note that the actual response of nitrate wet deposition can be larger than indicated by the DDM results. It can also be seen that reductions of SO2 emissions will give the greatest reduction in PM2.5 concentrations when compared to reductions in NOx and NH3 emissions. There are many different ways of processing the information provided by the sensitivity analysis. One possibility is to look at the impact of various emission reductions from a single subdomain onto pollutant levels at Class I areas. Figure 12-2 contains the percent responses (minus sign indicates a decrease) in particulate sulfate and wet sulfate due to 10% reductions of SO2 emissions from Tennessee. The decreases in ozone W126 due to reductions in elevated and ground level NOx emissions are also shown. In general, Class I areas closer to Tennessee (distances between the sites and Tennessee increase from left to right on the abscissa) show larger responses to emission reductions than those further away. However, the relative changes in response, as one looks at Class I areas further away from Tennessee, can be significantly different across the various pollutant-emission pairs. The response of sulfate wet deposition to SO2 emission reductions displays the largest decrease as the distance increases, followed by particulate sulfate to SO2 emissions, ozone W126 to groundlevel NOx emissions, and ozone W126 to elevated NOx emission. This indicates that the sulfate wet deposition responses to SO2 emission reductions from Tennessee are more local than other responses. The ozone W126 responses to elevated NOx emission reductions from Tennessee are more regional than other responses. 12-27

ODMAN,

ET AL.


Sensitivity to 10% Emission Reductions in TN

Sensitivities (%)

-5.0 -4.0 -3.0 -2.0 -1.0 0.0 GRSM

LIGO

SHRO JOKM COHU

Aerosol Sulfate to SO2 Ozone W126 to Ground NOx Figure 12-2

SIPS

JEFF

OTRC SHEN DOSO

Wet Sulfate to SO2 Ozone W126 to Elevated NOx

Sensitivity of various pollutants to a 10% reduction in SO2 and NOx emissions from Tennessee.

Figure 12-3 shows the ratio of the response that can be attributed to emission reductions from Tennessee to the response to domain wide reductions. The largest fractions of the responses that can be attributed to the emission reductions from Tennessee are seen at GRSM (48% of the particulate sulfate response to SO2 emission reductions). The fractions become smaller for stations further away (i.e., moving from left to right on the abscissa). Again, this figure indicates that the sulfate wet deposition responses to SO2 emission reductions from Tennessee are more local than the responses of other pollutants and that the ozone W126 responses to elevated NOx are more regional.

Fraction of Sens. (%)

TN Fraction of Sensitivity (TN Sensitivity/Domain Sensitivity)

50.0 40.0 30.0 20.0 10.0 0.0 GRSM

LIGO SHRO JOKM COHU

Aerosol Sulfate to SO2 Ozone W126 to Ground NOx Figure 12-3

SIPS

JEFF OTRC SHEN DOSO

Wet Sulfate to SO2 Ozone W126 to Elevated NOx

Fraction of domain wide sensitivities attributed to 10% reductions in SO2 and NOx emissions from Tennessee. 12-28

ODMAN,

ET AL.


13 CONCLUSION The Urban-to-Regional Multiscale (URM) model, a gas-phase photochemical model (Odman and Russell, 1991; Kumar et al., 1994), was successfully updated to include aerosol dynamics (Ansari and Pandis, 1999), aqueous sulfate formation, and wet deposition scavenging processes (Berkowitz et al., 1989). The model’s DDM-3D (Yang et al., 1997) sensitivity analysis module was enhanced to directly calculate aerosol and wet deposition sensitivities. The new version of URM, called URM-1ATM, is an integrated “one atmosphere” air quality model that can simulate all the important physical and chemical processes that govern the formation, transport, and removal of gas and particulate-phase pollutants in the atmosphere. It can also simultaneously calculate the sensitivities of pollutant levels to small changes in SO2, NOx, and NH3 emissions. The RAMS/EMS-95/URM-1ATM atmospheric modeling system was applied to nine weeklong episodes between 1991 and 1995. Model performance was evaluated by comparing model results to observations for ozone, particulate matter, and acid deposition in the Southern Appalachian Mountains. The ability of the modeling system to estimate ozone was typically within EPA’s guidance criteria for urban-scale modeling (USEPA, 1991) with normalized biases generally less than ±15% (except for the April-1995 episode) and normalized errors less than 35%. The major components of PM 2.5 (sulfate, ammonium, elemental and organic carbon) each had a mean normalized error of approximately 40%. The errors for less abundant components such as nitrate and crustal material were larger (98% and 187% respectively). Wet sulfate and nitrate deposition was biased low with mean normalized errors less than 25%, and wet ammonium deposition was biased high (112%). There is insufficient information for evaluating the ability of the model to simulate dry deposition fluxes. In general, pollutant levels were overestimated when observed levels were low and underestimated when they were high. Major sources of model uncertainty include: 1) coarse vertical and horizontal grid resolution that can smooth meteorological and concentration fields, 2) the inability of the modeling system to accurately reproduce some meteorological conditions (e.g. location and magnitude of precipitation), 3) highly uncertain emission inventories, especially ammonia and PM emissions of elemental and organic carbon, and crustal minerals, and 4) the assumption of linear decline of concentrations made for the initial and boundary conditions aloft. It was observed that the modeling results were sensitive to the boundary conditions and that long-range transport of both gas and particulate species from the boundary could bias the simulated pollutant levels in the Appalachian Mountains. Next, the RAMS/EMS-95/URM-1ATM atmospheric modeling system was used to assess how various emission strategies would affect seasonal ozone, annual PM2.5, and annual acid deposition in the Southern Appalachian Mountains in the years 2010 and 2040. Specific attention was given to the Class I areas, especially the Great Smoky Mountains and Shenandoah National Parks. The uncertainty in the model simulated response to emission changes can be expected to be proportional to the errors in the basecase simulations (i.e., differences between the modeled and observed values). Findings indicate that air quality will improve in the future due to regulations mandated under the Clean Air Act Amendments of 1990 and other recently promulgated regulations. In some cases, more stringent emission controls will provide additional improvements. However, the magnitude (and sometimes the direction) of responses differ depending on the specific pollutant and the location of the receptor. It was found that peak ozone concentrations in the Class I areas could be reduced with NOx controls. Sulfate PM and sulfate wet deposition decreased significantly in the Class I areas in response to SO2 emission controls. In general, the largest sulfate reductions (on a percent basis) occurred when the sulfate levels were highest. However, SO2 controls and increasing ammonia emissions may result in an increase of nitrate PM levels due to more ammonia becoming available. The response of organic PM to emission 13-1

ODMAN,

ET AL.


strategies is small since the biogenic emissions do not change in the future strategies. The decrease in oxidized nitrogen deposition was approximately equal to the increase in reduced nitrogen deposition. Therefore, changes in total nitrogen deposition were minimal, except when ammonia emissions are controlled. The sensitivity analysis module of URM-1ATM was used to quantify source-receptor relationships that are important for the SAMI region. DDM-3D is an efficient tool that allows calculation of many sensitivity coefficients, simultaneously, in a single simulation. The sensitivities of ozone, PM and wet deposition levels to SO2, elevated and ground level NOx, and NH3 emission reductions from the eight SAMI states and five surrounding regions were estimated. Assuming linear relationships between pollutant levels and emissions, local sensitivity coefficients were scaled and presented as responses to 10% emission reductions. Prior analysis, though limited, has shown that this assumption is generally valid up to about 30% reductions for most of the pollutant-emission pairs studied here. Sulfate wet deposition showed the largest percentage response to SO2 emission reductions. In response to 10% emission reductions along the trajectories of air masses that bring precipitation to the SAMI region, the average decrease of sulfate wet deposition at Class I areas was 7%. Particulate sulfate concentrations were reduced by 5%, on average, and this response was attributed to SO2 reductions from the home states of Class I areas and/or their adjacent states. Nitrate increased by an average of 2.5% and ammonium decreased by 3% in response to 10% SO2 emission reductions. As sulfate decreases, more ammonia becomes available to react with nitric acid, therefore particulate nitrate levels increase.The response of cumulative ozone W126 to a 10% reduction of total NOx emissions showed a 10% decrease at Class I areas of the SAMI region. The sensitivity to ground level emissions was approximately twice as large as the sensitivity to elevated NOx emissions. A large fraction of this response was attributed to elevated NOx emissions from the surrounding regions. This results indicates that "transport" is significant for cumulative ozone W126 at Class I areas of the SAMI region. Reductions of NOx emissions also impacted nitrate, ammonium and organic PM, but had little effect on nitrate wet deposition. In general, the ground level NOx emission reductions had a greater impact than elevated NOx emission reductions though the differences were small. Particulate nitrate showed the largest decrease, an average of 4.4% at Class I areas, in response to 10% reduction of total NOx emission (elevated plus ground level) reductions. The nitrate wet deposition decreased by an average of 2.3%. The responses to 10% reductions in NH3 emissions were mostly local except in Class I areas of Virginia and West Virginia that responded to reductions from the Midwest and Northeast regions. Nitrate and ammonium PM decreased by an average of 2.5% and 1.5%, respectively. Wet deposition of ammonium decreased by an average 1.1%. at Class I areas of the SAMI region. The sensitivity of nitrate wet deposition to NH3 emissions was not well quantified by the DDM-3D method. This is probably due to a potentially large dependence on NH3 emissions from distant regions, which would be poorly quantified due to the coarse (96 km or larger) grid resolution over those regions. The actual response is believed to be comparable in magnitude to the response to NOx emissions. Therefore, increases in NH3 emission may reduce the benefits of NOx emission reductions on nitrate wet deposition. The sensitivity analysis results presented here provide qualitative guidance that can be very useful in the design of emission strategies . They can be used to identify the location and type of emission sources that should be controlled to most effectively reduce ozone, PM, and acid deposition levels. However, these results cannot, by themselves, provide reliable information on air quality changes expected from emission reductions when multiple emission species are involved or when changes are large (larger than 30%). Explicitly modeling emissions strategies is still necessary to obtain robust estimates of air quality benefits from emission reductions. Some questions could not be answered by the time this report has been written: 13-2

ODMAN,

ET AL.


1) Why is the model estimating too much ammonium wet deposition? 2) Why is the nitrate wet deposition not responding to NOx emission reductions? 3) Why is DDM-3D is performing poorly for ammonium and nitrate wet deposition? These questions are the topic of ongoing research. The papers listed at the end of Section 2.4 and the follow-on papers should be consulted for the forthcoming answers. The results of this air quality modeling study can be used in evaluating the environmental and socioeconomic implications of current federal and state regulations and SAMI emissions reduction strategies. This information can be used by policy-makers to help them make better informed decisions.

13-3

ODMAN,

ET AL.



13-4

ODMAN,

ET AL.


14 REFERENCES Andrews E., Saxena P., Musarra S., Hildemann L.M., Koutrakis P., McMurry P.H., Olmez I. and White W.H., 2000. Concentration and composition of atmospheric aerosols from the 1995 SEAVS experiment and a review of the closure between chemical and gravimetric measurements. J. Air Waste Management Assoc. 50, 648-664. Ansari A.S. and Pandis S.N., 1999. Prediction of multicomponent inorganic atmospheric aerosol behavior. Atmospheric Environment 33, 745-757. ARS (2001) Air Resource Specialists, Inc., Southern Appalachian Mountains Initiative Visibility Analyses, Draft Final Report: December 21, 2001. Fort Collins, CO. Bergin M.S., Boylan, J.W., Wilkinson J.G., Odman M.T. and Russell A.G. (2001) Regional multiscale atmospheric modeling: comparison of model performance for aerosol simulations using two grids. Poster presentation. American Association for Aerosol Research Conference. Portland, OR. Bergin M.S., Russell A.G. and Milford J.B. (1998) Effects of chemical mechanism uncertainties on the reactivity quantification of volatile organic compounds using a three-dimensional air quality model. Environ. Sci. Technol . 32, 694-703. Berkowitz C.E., Easter R.C. and Scott B.C., 1989. Theory and results from a quasi-steady-state precipitation-scavenging model. Atmospheric Environment 23, 1555-1571. Byun D.W. and Ching J.K.S., 1999. Science Algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System. EPA/600/R-99/030, U. S. Environmental Protection Agency, Office of Research and Development, Washington, DC. Carter W.P.L., 1990. A detailed mechanism for the gas-phase atmospheric reaction of organic compounds. Atmospheric Environment 24, 481-518. Carter W.P.L., 1994. Development of ozone reactivity scales for volatile organic compounds. J. Air Waste Management Assoc. 44, 881-899. Carter W.P.L., 1995. Computer modeling of environmental chamber measurements of maximum incremental reactivities of volatile organic compounds. Atmospheric Environment 29, 2513 – 2527. Cowling E.B. (1998) Recent changes in chemical climate and related effects on forests in North America and Europe. AMBIO 18, 167-171. Deuel H.P. and Douglas, S.G. (1998) Episode selection for the integrated analysis of ozone, visibility and acid deposition for the Southern Appalachian Mountains; Systems Applications International, Inc.: San Rafael, CA. SYSAPP-98/07r1. Dockery D.W., Pope C.A., Xu X., Spengler J.D., Ware J.H., Fay M.E., Ferris Jr. B.G., Speizer F.E. (1993) An association between air pollution and mortality in six U.S. cities, N. Engl. J. Med. 329:1753. Doty K.G., Tesche T.W., McNally D.E, Timin B. and Mueller S.F. (2001). Meteorological Modeling for the Southern Appalachian Mountains Initiative (SAMI). Final Report, Southern Appalachian Mountain Initiative. Asheville, NC. Dunker A.M. (1981) Calculation of sensitivity coefficients for complex atmospheric models. Atmos. Environ. 15, 1155-1161. Dunker A.M. (1984) The decoupled direct method for calculating sensitivity coefficients in chemical kinetics. J. Chem. Phys. 81, 2385-2393.

14-1

ODMAN,

ET AL.


EPRI, 1998. Southeastern Aerosol and Visibility Study (SEAVS): Concentration and Composition of Atmospheric Aerosols at Look Rock, Tennessee, July-August 1995. Electric Power Research Institute. Report TR-111063. Palo Alto, CA. Friedlander S.K., 1977. Smoke, Dust, and Haze. Wiley. New York, NY. GIT

(2001) The SAMI Atmospheric http://environmental.gatech.edu/SAMI

Modeling

Home

Page

at

Georgia

Institute

of

Technology.

Green P.J. and Sibson R., 1978. Computing Dirichlet tessellations in the plane. The Computer Journal 21, 168-173. Harley R.A., Russell A.G., McRae G.J., McNair L.A., Winner D.A., Odman M.T., Dabdub D., Cass G.R. and Seinfeld J.H., 1992. Continued development of a photochemical model and application to the Southern California Air Quality Study (SCAQS) intensive monitoring period: Phase I. Final Report to Coordinating Research Council. Harley R.A., Russell A.G., McRae G.J., Cass G.R. and Seinfeld J.H. (1993) Photochemical modeling of the Southern California Air Quality Study. Environ. Sci. Technol . 27, 378-388. Heck W.W., Furiness C.S., Cowling E.B. and Sims C.K. (1998) Effects of ozone on crop, forest, and natural ecosystems: assessment of research needs. EM. October, 11-22. Heck W.W., Yaylor O.C., Adams R., Bingham G., Miller J., Preston E., and Weinstein L. (1982) Assessment of crop loss from ozone. J. Air Pollut. Control Assoc. 32:353. IIrving, P.M., Ed. (1990) Evaluation of Regional Acidic Deposition Models (Part I). Acid Deposition: State of Science and Technology, Report 5, Vol. 1, 118-122. National Acid Precipitation Assessment Program, Office of the Director: Washington, DC. Kalnay E., Kanamitsu M., Kistler R., Collins W., Deaven D., Gandin L., Iredell M., Saha S., White G., Woollen J., Zhu Y., Chelliah M., Ebisuzaki W., Higgins W., Janowlak J., Mo K.C., Ropelewski C., Wang, J., Leetmaa A., Reynolds R., Jenne R. and Joseph D., 1996. The NCEP/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc. 77, 437-471. Kim Y.P. and Seinfeld J.H. (1989) Simulation of multicomponent aerosol condensation by the moving sectional method. Journal of Colloid and Interface Sci. 135, 185-199. Kumar N. and Russell A.G., 1996. Multiscale air quality modeling of the northeastern United States. Atmospheric Environment 30, 1099-1116. Kumar N., Odman M.T. and Russell A.G., 1994. Multiscale air quality modeling: Application to southern California. J. Geophys. Res. 99, 5385-5397. Lippmann M. (1989) Health effects of ozone: A critical review . J. Air & Waste Manage. Assoc. 39:672. Lefohn A.S. and Runeckles V.C. (1987) Establishing a standard to protect vegetation - ozone exposure/dose considerations. Atmospheric Environment 21, 561-568. McBride J.L. and Ebert E.E. (2000) Verification of quantitative precipitation forecasts from operational numerical weather prediction models over Australia. Wea. Forecasting. 15, 103-121. McNally D.E. and Tesche T.W. (1991): Modeling Analysis and Plotting System User Manual; Alpine Geophysics: Pittsburgh, PA. McNair L., Russell A. and Odman M.T. (1992) Airshed calculation of the sensitivity of pollutant formation to organic compound classes and oxygenates associated with alternative fuels. J. Air & Waste Manage. Assoc . 42, 174-178.

14-2

ODMAN,

ET AL.


Mendoza-Dominguez A. and Russell A.G. (2001) Estimation of emission adjustments from the application of four-dimensional data assimilation to photochemical air quality modeling. Atmospheric Environment 35, 2879-2894. Mueller P. K., 1998. NARSTO 1998 Model-Intercomparison Study Verification Data: NARSTO-Northeast 1995 Surface Ozone, NO, and NOX . Available online from the Langley Atmospheric Sciences Data Center (http://eosweb.larc.nasa.gov/), NASA Langley Research Center, Hampton, Virginia, U.S.A. NAPAP (1990) National Acid Precipitation Assessment Program. Effects of acid deposition on materials. NAPAP Report 19, pp 3280. In Acid Deposition: State of Science and Technology, Vol III. Terrestrial, Materials, Health and Visibility Effects . P.M. Irving, Ed., NAPAP: Washington, D.C. National Center for Atmospheric Research, 1982. Regional Acid Deposition: Models and Physical Processes, Boulder, CO. Nenes A., Pilinis C. and Pandis S.N., 1998. ISORROPIA: A new thermodynamic equilibrium model for multiphase multicomponent inorganic aerosols. Aquatic Geochem. 4, 123-152. NPS, 2000. National Park Service Air Quality Research Division Fort Collins. Anonymous ftp at ftp://alta_vista.cira.colostate.edu in /data/improve. NRC (1993) National Research Council. Protecting visibility in national parks and wilderness areas. National Academy Press: Washington, D.C. Odman M.T. and Russell A.G., 1991a. Multiscale modeling of pollutant transport and chemistry. J. Geophys. Res. 96, 7363-7370. Odman M.T. and Russell A.G., 1991b. A multiscale finite element pollutant transport scheme for urban and regional modeling. Atmospheric Environment 25A, 2385-2394. Odman M.T. and Russell A.G., 1993. A nonlinear filtering algorithm for multi-dimensional finite element pollutant advection schemes. Atmospheric Environment 27A, 793-799. Odman M.T. and Russell A.G., 2000. Mass conservative coupling of non-hydrostatic meteorological models with air quality models. Air Pollution Modeling and Its Application XIII, pp. 651-660. Gryning S.-E. and Batchvarova E. (Eds.), Kluwer Academic/Plenum Publishers. New York, NY. Odman M.T., Hakami A., Boylan J.W. and Russell A.G., 2001. Acid Deposition Linkage, Final Report to SAMI, Georgia Institute of Technology, Atlanta, GA, 20 pp. Odum J.R., Hoffmann T., Bowman F., Collins D., Flagan R.C. and Seinfeld J.H. (1996) Gas/particle partitioning and secondary organic aerosol yields. Environ. Sci. Technol.30, 2580-2585. Pandis S.N., Harley R.A., Cass G.R. and Seinfeld J.H., 1992. Secondary organic aerosol formation and transport. Atmospheric Environment 26A, 2269-2282. Pandis S.N., Wexler S.W. and Seinfeld J.H., 1993. Secondary organic aerosol formation and transport-II. Predicting the ambient secondary organic aerosol size distribution. Atmospheric Environment 27A, 2403-2416. Pechan /Avanti Group, 2001. Southern Appalachian Mountains Initiative (SAMI) Emissions Projections to 2010 and 2040: Growth and Control Data and Emission Estimation Methodologies. Draft Final Report # 01.07.002/9405.000. Pielke R.A., Cotton W.R., Walko R.L., Tremback C.J., Lyons W.A., Grasso L.D., Nicholls M.E., Moran M.D., Wesley D.A., Lee T.J. and Copeland J.H., 1992. A comprehensive meteorological modeling system - RAMS. Meteor. Atmos. Phys. 49, 69-91. Pierce T.E. and Geron C.D., 1996. The personal computer version of the Biogenic Emissions Inventory System (PCBEIS2.2). AAREADME, anonymous ftp at monsoon.rtpnc.epa.gov located in /pub/beis2/pcbeis22.

14-3

ODMAN,

ET AL.


Pierce T.E., 1996. Documentation for BEIS2 located at ftp://monsoon.rtpnc.epa.gov/pub/beis2/SOS/AAREADME. Pierce T.E., Lamb B.K. and Van Meter A.R., 1990. Development of a Biogenic Emissions Inventory System for Regional Scale Air Pollution Models. In Proceedings of 83 rd Annual Meeting of the Air and Waste Management Association. Pittsburgh, PA. Preparata and Shamos, 1985. Computational Geometry: An Introduction, Springer-Verlag. Rogers E., Deaven D.G. and DiMego G. J. (1995) The regional analysis system for the operational "Early" Eta model: Original 80km configuration and recent changes. Wea. Forecasting. 10, 810-825. Russell A.G. and Dennis R., 2000. NARSTO critical review of photochemical models and modeling. Atmospheric Environment 34, 2283-2324. Russell A.G., Winter D.A., McCue K.F. and Cass G.R., 1990. Mathematical modeling and control of dry deposition flux of nitrogencontaining air pollutants. Rep. CARB A6-188-32. California Air Resources Board, CA. Russell A.G., McCue K.F. and Cass G.R. (1988) Mathematical modeling of the formation of nitrogen-containing air pollutants. I. Evaluation of an Eulerian photochemical model. Environ. Sci. Technol. 22, 263-271. SAMI (2001) Southern Appalachian Mountains Initiative 2001 Interim Report. Saxena P. and Hildemann L.M. (1996) Water-soluble organics in atmospheric particles: a critical review of the literature and application of thermodynamics to identify candidate compounds. J. Atmos. Chem. 24, 57-109. Schaefer J.T. (1990) The critical success index as an indicator of warning skill. Wea. Forecasting 5, 571-575. Scott B.C., 1987. User’s Manual for the Convective Cloud Module Version 1.0. PNL-6188. Pacific Northwest Laboratory. Richland, WA. Seigneur C., Pai P., Hopke P.K. and Grosjean D. (1999) Modeling atmospheric particulate matter. Environ. Sci. Technol. 33:80A86A. Seigneur C., Tonne C., Krishnakumar V., Pai P. and Levin L. (2000) The sensitivity of PM2.5 source-receptor relationships to atmospheric chemistry and transport in a three-dimensional air quality modeling. J. Air & Waste Manage. Assoc . 50, 428-435. Seigneur C. (2001) Current status of air quality models for particulate matter. J. Air & Waste Manage. Assoc. 51, 1508-1521. Seinfeld J.H. (1986) Atmospheric Physics and Chemistry of Air Pollution, Wiley Interscience: New York, NY. Seinfeld, J. and Pandis, S., 1998. Atmospheric Chemistry and Physics. John Wiley & Sons, Inc. New York, NY. Shin, W.-C. and Carmichael, G.R. (1992) Sensitivity of acid production/deposition to emission reductions. Environ. Sci. Technol., 26, 715-725. Sisler J.F. and Malm W.C. (2000) Interpretation of Trends of PM2.5 and Reconstructed Visibility from the IMPROVE Network. J. Air & Waste Manage. Assoc. 50, 775-789.

.

Stein, A.H. and Lamb, D. (2000) The sensitivity of sulfur wet deposition to atmospheric oxidants Atmos. Environ., 34, 1681-1690. Turpin B.J. and Lim H-J (2001) Species contributions to PM2.5 mass concentrations: revisiting common assumptions for estimating organic mass. Aerosol Sci. Technol.35, 602-610.

14-4

ODMAN,

ET AL.


USEPA (1990) Section 112 of the Clean Air Act Amendments. http://www.epa.gov/ttn/atw/mactfnl.html. USEPA (1991) Guidance for Regulatory Application of the Urban Airshed Model (UAM), Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. USEPA (1994). User’s Guide To MOBILE5 (Mobile Source Emission Factor Model). EPA-AA-TEB-94-01. U.S. Environmental Protection Agency, Office of Air and Radiation, Office of Mobile Sources, Emissions Planning and Strategies Division, Air Quality Analysis Branch, Ann Arbor, MI. USEPA (1996) Air Quality Criteria for Ozone and Related Photochemical Oxidants. U.S. Environmental Protection Agency, Office of Research and Development. EPA/600/P-93/004aF USEPA (1998) 40 CFR Parts 51, 72, 75, and 96. Federal Register Vol. 63, No. 207, 57356-57538. USEPA (1999) Draft Guidance on the Use of Models and Other Analyses in Attainment Demonstrations for the 8-Hour Ozone NAAQS. EPA-454/R-99-004. USEPA (2001a). EPA AIRS Data. U.S. Environmental Protection Agency, Office of Air Quality Planning & Standards, Information Transfer & Program Integration Division, Information Transfer Group. http://www.epa.gov/airsdata. USEPA (2001b) 40 CFR Parts 80 and 86. Federal Register Vol. 66, No. 72, 19295-19311. USEPA (2002) Air Quality Criteria for Particulate Matter. U.S. Environmental Protection Agency, Office of Research and Development. EPA/600/P-99/002aC. Wark K. and Warner C.F. (1981) Air Pollution – Its Origin and Control. Harper & Row, Publishers: New York, NY. Warneck P. (1988) Chemistry of the Natural Atmosphere. International Geophysics Series, Volume 41, p. 523. Academic Press: San Diego, CA. Wesely M.L. (1989) Parameterization of surface resistances to gaseous dry deposition in regional-scale numerical models. Atmospheric Environment 23, 1293-1304. Wexler A.S., Lurmann F.W., and Seinfeld J.H. (1993) Modeling Urban Aerosols. In proceedings of 1992 A&WMA/CARB/SCAQMD International Symposium. Los Angeles, CA. Wilkinson J.G., Loomis C.F., McNally D.E., Emigh R.A. and Tesche T.W. (1994). Technical Formulation Document: SARMAP/LMOS Emissions Modeling System (EMS-95). AG-90/TS26 & AG-90/TS27. Alpine Geophysics, Pittsburgh, PA. Winner D.A., Cass G.R. and Harley R.A. (1995) Effect of alternative boundary conditions on predicted ozone control strategy performance: A case study in the Los Angeles area. Atmos. Environ. 29, 3451-3464. Yang Y.J., Wilkinson J.W. and Russell A.G. (1997) Fast, direct sensitivity analysis of multidimensional photochemical models. Environ. Sci. Technol. 31, 2859-2868. Zhang Y., Seigneur C., Seinfeld J.H., Jacobson M.Z. and Binkowski, F.S., 1999. Simulation of aerosol dynamics: A comparative review of algorithms used in air quality models. Aerosol Sci. Tech. 31, 487-514.

14-5

ODMAN,

ET AL.



14-6

ODMAN,

ET AL.


15 APPENDIX A: Compact Disk (CD) Over 25,000 illustrations were prepared during this project. At present, the easiest way to access these illustrations is through the Georgia Tech (http://environmental.gatech.edu/SAMI) or TVA/AG (http://www.tva.gov/sami) web sites. A compact disk (CD) was prepared as a permanent record of the illustrations mentioned throughout this report.

15-1

ODMAN,

ET AL.



15-2

ODMAN,

ET AL.


16 APPENDIX B: CD File Structure and Naming Conventions In this section the directory structure and the file naming conventions used in the compact disk (CD) are described. Colors are used to identify the depth of directories in the structure as follows: st

1 level Blue nd

2 level Pink rd

3 level Red th

4 level Green The file names are italicized and a particular notation is used. The portion of a file name demarcated by curly brackets can be substituted by any of the options separated by vertical bars. For example, a{nitr|sulf}.gif represents two files: anitr.gif and asulf.gif The illustrations on the CD are grouped in four directories: 1) Emissions (Chapters 4 and 9) 2) Model Results (Chapters 5, 6, 7 and Appendix C) 3) Future Years (Chapter 10) 4) Regional Sensitivities (Chapter 12)

16.1 Directory Structure of "Emissions" There are eight directories under "Emissions": 1) 2010-OTW 2) 2010-BwC 3) 2010-BB 4) 2040-OTW 5) 2040-BwC 6) 2040-BB 7) Basecase 8) Summary Tables The first six directories are for future year emissions and contain sub-directories for different episodes. The "2010-OTW " and "2040-OTW " directories have sub-directories for each one of the nine episodes: 1) Apr27-May3,1995 2) Aug3-11,1993 3) Feb8-13,1994 4) July11-19,1995 5) July23-31,1991 6) June24-29,1992 7) March23-31,1993 8) May11-17,1993 9) May24-29, 1995

16-1

ODMAN,

ET AL.


The "2010-BwC" , "2010-BB", "2040-BwC" and "2010-BB" directories have only six subdirectories: "Feb8-13,1994", "July11-19,1995", "July23-31,1991", "March23-31,1993", "May11-17,1993" and "May24-29,1995". Under these six sub-directories there are two directories: 1) projections 2) comparisons The "projections" directory contains the future year emissions and the "comparisons " directory contains the differences between future year emissions and baseyear or OTW strategy. The files in these directories are named as: {point|area}_{nox|so2}_mmmDD.gif and area_nh3_mmmDD.gif where mmm is a string of characters identifying the month, for example "feb" for the "Feb8-13,1994" episode, and DD is the day of the month as a number. The structure and file naming convention is different for the "Apr27-May3, 1995", "Aug3-11, 1993", and "June24-29, 1992" episodes. There is only a "comparisons " directory for these episodes (no "projections " directory) and the files are named as: {pt|gl}_{NOX|SO2}_YYMMDD.gif and gl_NH3_YYMMDD.gif where YY is the year, MM is the month and DD is the day, all as numbers. The "Basecase" directory contains a separate directory for each one of the nine episodes: 1) Apr27-May3, 1995 2) Aug3-11, 1993 3) Feb8-13,1994 4) July11-19,1995 5) July23-31,1991 6) June24-29, 1992 7) March23-31, 1993 8) May11-17,1993 9) May24-29, 1995 The files under these sub-directories are named with the same conventions as those under the future year sub-directories. The "Summary Tables" directory contains files named as: {april95|august93|february94|july91|july95|june92|march93|may93|may95}_{all|sami}.htm

16-2

ODMAN,

ET AL.


16.2 Directory Structure of "Model Results" There are nine directories under “Model Results”: 1) Apr27-May3, 1995 2) Aug3-11, 1993 3) Feb8-13,1994 4) July11-19,1995 5) July23-31,1991 6) June24-29, 1992 7) March23-31, 1993 8) May11-17,1993 9) May24-29, 1995 10) Drydep The directories for the nine episodes contain two sub-directories: 1) maps 2) MPE The "maps" sub-directories contain the spatial plots for aerosols, ozone, dry deposition and wet deposition. For the "Feb8-13,1994", "July11-19,1995", "July23-31,1991", "March23-31,1993", "May11-17,1993" and "May24-29,1995" episodes the aerosol files are named as: a{amon|elmc|nitr|orgc|pm10|pm25|soil|sulf}_mmmDD.gif where mmm is the month in letters, for example "february" for the "Feb8-13,1994" episode, and DD is the day of the month as a number. The ozone files are named as: ozon_mmmDD.gif Spatial plots are also available for SO2 in files named as: so2_mmmDD.gif The dry and wet deposition files are respectively named as: dd{hno3|nitr|so2|sulf}_episode.gif and wd{amon|calc|hion|magn|nitr|sulf}_episode.gif where "episode" is a qualifier of the episode, for example "february8-13" for the the "Feb8-13,1994" episode. The precipitation files are named as: precip_episode.gif The file naming conventions are slightly different for the "Apr27-May3,1995", "Aug3-11,1993" and "June24-29,1992" episodes. The aerosol files are named as: {AMNF|CARF|NITF|ORGF|PM10|PM25|SOIL|SULF}mmmDD.gif

16-3

ODMAN,

ET AL.


The dry and wet deposition files are respectively named as: conc_dd_{HNO3|NITF|SO2|SULF}_mmmDD_base.gif and conc_wd{CA|HION|MG|NH4|NO3|PREC|SO4}mmmdd_base.gif The "MPE" sub-directories contain the illustrations for model performance evaluation. They each have three directories: 1) aerosols 2) deposition 3) ozone (except for "Feb8-13,1994" and " March23-31, 1993" episodes) First the file naming conventions for the "Feb8-13,1994", "July11-19,1995", "July23-31,1991", " March23-31, 1993", "May11-17,1993" and "May24-29, 1995" episodes will be described. In the "aerosols" directory, the files are named as follows: a{amon|elmc|nitr|orgc|pm10|pm25|soil|sulf}NN.gif where NN is a number identifying IMPROVE stations as listed in Table 16-1. The stacked bar charts showing the comparison to observations for all aerosol components at all IMPROVE sites are named all_DD.gif where DD is the day of the month as a number.

Table 16-1

IMPROVE monitoring station identifiers used in the file names.

IMPROVE Station

Station ID

Brigantine, NJ

1

Dolly Sods/Otter Creek, WV

2

Great Smoky Mountains, TN

3

Jefferson/James River Face, VA

4

Lye Brook, VT

5

Mammoth Cave, KY

6

Okefenokee, GA

7

Cape Romain, SC

8

Shenandoah, VA

9

Shining Rock, NC

10

Sipsey, AL

11

Upper Buffalo, AR

12

Washington, DC

13

16-4

ODMAN,

ET AL.


The "deposition" directory contains the mass-flux files named as wet_{ca|hion|mg|nh4|no3|so4}.gif The concentrations in the rainwater also exist and they are named as cwet_{ca|hion|mg|nh4|no3|so4}.gif In the "ozone" directory the files are named as: ozonMM.gif where MM is a number identifying AIRS stations as listed in Table 16-2.

Table 16-2

AIRS monitoring station identifiers used in the file names.

AIRS Station

Station ID

Clay Co, AL

1

Huntsville, AL

2

Sipsey Wilderness area, Lawrence Co, AL Shelby Co, AL

3 4

Washington, DC

5

Dawsonville, Dawson Co, GA

6

South Decalb, Decalb Co, GA

7

Yorkville, Paulding Co, GA

8

Lexington, Fayette Co, KY

9

Taylorsville, Alexander Co, NC

10

Asheville , Buncombe Co, NC

11

Lenoir, Caldwell Co, NC

12

Winston-Salem, Forsyth Co, NC

13

Frying Pan, Haywood Co, NC

14

Purchase Knob, Haywood Co, NC

15

Crouse, Lincoln Co, NC

16

Co Line, Mecklenburg Co, NC

17

Swain Co, NC

18

Powdersville, Anderson Co, SC

19

Oconee Co, SC

20

Look Rock, Blount Co, TN

21

Bradley Co, TN

22

Nashville, Davidson Co, TN

23

Chattanooga, Hamilton Co, TN

24

Knoxville, Knox Co, TN

25

Cove Mountain, Sevier Co, TN

26

Clingmans Dome, Sevier Co, TN

27

Blountville, Sullivan Co, TN

28

Big Meadows, Madison Co, VA

29

Roanoke Co, VA

30

Whitetop Mt, Smyth Co, VA

31

Greenbrier Co, WV

32

Charleston, Kanawha Co, WV

33

16-5

ODMAN,

ET AL.


The "ozone" directories also contain files for the normalized bias and error files for the 12-km (biasn12.gif and erorn12.gif) as well as the 24- and 48-km resolution grids (biasn24.gif and erorn24.gif). Finally, there are also files for the daily mean bias and error in the 12-km grid for all simulated days (biasm12.gif and erorm12.gif). The file naming conventions are different for the "Apr27-May3, 1995", "Aug3-11, 1993" and "June24-29, 1992" episodes. For these three episodes, the "aerosols" directory contains tables showing the bias and error. These files are named as: aer_{ammonium|carbon|nitrate|organics|pm10|pm25|soils|sulfate}_mpe.gif The illustrations for IMPROVE stations are grouped by species and the files are named as {ammonium|carbon|nitrate|organics|pm10|pm25|soils|sulfate}_perf_scat.pdf Similarly, in the "deposition" directory there are error and bias tables named as wet_{ca|hion|mg|nh4|no3|so4}_mpe.gif as well as files showing performance at all NADP station. Deposition flux and concentration in rainwater share the same file named as: wet_dep_{ca|hion|mg|nh4|no3|so4}.pdf In the "ozone"directory the file names are ozonNN.gif but the station identifiers, NN, are different. The "Drydep" directory contains comparisons of the dry deposition to observations at AIRMoN sites. These files are named as: dd{hno3|nitr|so2|sulf}NN.gif where NN is 1 for Panola, GA and 2 for Oakridge, TN.

16.3 Directory Structure of "Future Years" The “Future Years” includes spatial plots for each future year strategy (total of six) of all 9 episodes in directories named as: 1) Apr27-May3,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 2) Aug3-11,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 3) Feb8-13,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 4) July11-19,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 5) July23-31,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 6) June24-29,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 7) March23-31,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 8) May11-17,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} 9) May24-29,{2010-BB|2010-BWC|2010-OTW|2040-BB|2040-BWC|2040-OTW}} Each one of these directories has two sub-directories: 16-6

ODMAN,

ET AL.


1) comparisons 2) results The "results" directory contains the future year pollutant levels and the "comparisons" directory contains the differences between future year and baseyear (or OTW strategy) levels. For the "Feb813,1994", "July11-19,1995", "July23-31,1991", " March23-31, 1993", "May11-17,1993" and "May24-29, 1995" episodes, the files in these directories are named as: a{amon|elmc|nitr|orgc|pm10|pm25|pmc|soil|sulf}_mmmDD.gif {ozon|so2}_mmmDD.gif wd{amon|calc|hion|magn|nitr|sulf}_episode.gif where mmm is the month in letters, for example "february" for the "Feb8-13,1994" episode, DD is the day of the month as a number, and "episode" is a qualifier of the episode, for example "february8-13" for the "Feb8-13,1994" episode. For the "Apr27-May3, 1995", "Aug3-11, 1993" and "June24-29, 1992" episodes, files in the “results” directory are named as: conc_{AMNF|CARF|NITF|ORGF|PM10|PM25|SOIL|SULF}_strategy_mmmDD.gif conc_o3_strategy_mmmDD.gif conc_dd_{HNO3|NITF|SO2|SULF}_strategy.gif conc_wd_{CA|HION|MG|NH4|NO3|SO2}_strategy.gif and those in the “comparisons ” directories are named as: dif_{AMNF|CARF|NITF|ORGF|PM10|PM25|SOIL|SULF}_strategy_mmmDD.gif dif_o3_strategy_mmmDD.gif dif_dd_{HNO3|NITF|SO2|SULF}_strategy.gif dif_wd_{CA|HION|MG|NH4|NO3|SO2}_strategy.gif There are three more directories under “Future Years”: 1) Ozone 2) Stationwise_2010 3) Stationwise_2040 The "Ozone" directory contains four sub-directories: 1) July11-19,1995 2) July23-31,1991 3) May11-17,1993 4) May24-29,1995 These directories contain files that compare future-year ozone levels at various sites to the basecase levels. The files are named as: 16-7

ODMAN,

ET AL.


ozonNN.gif where NN is a numeric identifier for the site as listed in Table 10-1. For the other ozone episodes, the comparisons at various stations are merged in a single file. (Recall that "Feb8-13,1994" and "March23-31, 1993" are not ozone episodes.) The files are named as {Apr95|Aug93|Jun92}_ozone.pdf The “Stationwise_2010” contains comparisons of the aerosol levels for 2010 strategies to the basecase levels. These files are named as a{amon|elmc|nitr|orgc|pm25|pmc|soil|sulf}NN.gif where NN is a numeric site station identifier as listed in Table 10-1. The comparisons for dry and wet deposition fluxes can be respectively found in files named as: dd{amon|hno3|hono|n2o5|nh3|nitr|no|no2|o3|so2|sulf}MM.gif wd{amon|calc|hion|magn|nitr|sulf}MM.gif where MM is a numeric site station identifier as listed in Table 10-1. The “Stationwise_2040” contains comparisons between 2040 strategies and the basecase. These files are named with the same convention as the one for “Stationwise_2010”

16.4 Directory Structure of "Regional Sensitivities" Under “Regional Sensitivities” there are directories for all nine episodes: 1) Apr27-May3, 1995 2) Aug3-11, 1993 3) Feb8-13,1994 4) July11-19,1995 5) July23-31,1991 6) June24-29, 1992 7) March23-31, 1993 8) May11-17,1993 9) May24-29, 1995 These directories contain spatial plots of sensitivities to emissions from 8 SAMI states and the neighboring regions. The sensitivities of PM 2.5, its sulfate component, and ozone can be found in files respectively named as: apm25_{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{anox|pnox|so2}_mmmDD.gif asulf_{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{so2}_mmmDD.gif o3_{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{anox|pnox}_mmmDD.gif

16-8

ODMAN,

ET AL.


These files contain absolute sensitivities. A second set of files exists where the sensitivities are presented as percentages. These files follow the same conventions above but their names are prefixed with “p” The sensitivities of nitrate and sulfate wet deposition can be found in files named as: wdnitr_{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{anox|pnox}_episode.gif wdsulf_{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{so2}_episode.gif For the "Apr27-May3, 1995", "Aug3-11, 1993" and "June24-29, 1992" episodes, the file names for the sensitivities of PM2.5, ozone and nitrate wet deposition are slightly different: apm25_{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{nll|npt|so2}_mmmDD.gif o3__{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{nll|npt}_mmmDD.gif wdnitr_{AL|AO|CN|DW|GA|KY|MW|NC|NE|SAMI|SC|SE|TN|VA|WV}_{nll|npt}_episode.gif For the "Feb8-13,1994", "July11-19,1995", "July23-31,1991", "March23-31,1993", "May1117,1993" and "May24-29,1995" episodes, there are also files containing the spatial plots for all 5 subdomains (regions) and all 8 SAMI states along with the results of the simulation with the 2010-OTE emissions. These files are named as: a{pm25|sulf}_{all5|all8}_so2_mmmDD.gif ozone_{all5|all8}_{anox|pnox}_mmmDD.gif wdnitr_{all5|all8}_{anox|pnox}_episode.gif wdsulf_{all5|all8}_so2_episode.gif The matching illustrations for the "Apr27-May3,1995", "Aug3-11,1993" and "June24-29,1992" episodes can be found in the “TVA Episodes all5 & all8 (ppt)” directory. The file names are self explanatory. The “Stationwise” directory contains the episodic sensitivities at various sites. The files are named as: {aamon|anitr|apm25|wdamon|wdnitr}_{anox|nh3|pnox|so2}_site.GIF {aorgc}_{anox|pnox}_site.GIF asulf_{nh3|so2}_site.GIF wdsulf_so2_site.GIF where site is a 4 letter site identifier. The ozone sensitivities are broken into two groups: one for class 1 and 2 ozone days and a second for class 3 and 4 ozone days. The file names are as follows: {ozon12|ozon34}_{anox|pnox}_site.GIF 16-9

ODMAN,

ET AL.


These files contain absolute sensitivities. A second set of files exist where the sensitivities are presented as percentages. These files follow the same conventions above but their names are prefixed with “p” The “Annual_Seasonal” directory contains annual or seasonal (ozone) sensitivities at various sites. The files are named as {aamon|anitr|apm25|wdamon|wdnitr}_{anox|nh3|pnox|so2}_site.GIF {aorgc|ozon}_{anox|pnox}_site.GIF asulf_{nh3|so2}_site.GIF wdsulf_so2_site.GIF where site is a 4 letter site identifier. There are also illustrations of sensitivities at all sites. These files are named with “all” as a site identifier. In addition to these files that contain absolute sensitivities, there are also files that present sensitivities as percentages. The names of these files are prefixed with “p”. A third set of files contain normalized sensitivities; their names are prefixed with “n”.

16-10

ODMAN,

ET AL.


17 APPENDIX C: Dry Deposition Model Performance Observations of dry deposition are scarce and not as accurate as the observations for ozone PM or wet deposition. Dry deposition fluxes are inferred indirectly from other measurements using another model. In addition, dry deposition fluxes are, by their nature, very site-specific and cannot be readily extrapolated to other locations or larger geographic areas surrounding the points of measurement. Thus, we have no expectation that the inferred dry deposition fluxes will be similar to those modeled as an average for grid cells. Consequently, the dry deposition performance evaluation in this section should not be given as much credence as the ozone, PM or wet deposition performance evaluations. For this reason, the evaluation of dry deposition performance was not included in the body of this report.

17.1 AIRMoN Observation Database Dry deposition performance was evaluated by comparing modeling results to values computed from the measurements at the Atmospheric Integrated Research Monitoring Network (AIRMoN) monitoring sites. The dry deposition mass fluxes reported by AIRMoN are based on weekly average concentrations. This approach uses observations of ambient pollutant concentrations, known site characteristics and meteorological conditions as input to an inferential model that estimates speciated dry deposition mass fluxes. The results are very site specific and tend to be prone to errors because the approach ignores the short term fluctuations in conditions that affect deposition rates. The species that were compared include sulfur dioxide (gas), nitric acid (gas), sulfate (PM), and nitrate (PM). AIRMoN measurements are taken once each week (Tuesday) and the mass deposition fluxes are reported as a seven-day cumulative. There are twelve AIRMoN monitoring sites in the modeling domain. However, only data from the two stations in the 12 km grids are used to determine model

Figure 17-1

Dry deposition fluxes (mg/m2) of SO2 for the week of July 11-18, 1995. 17-1

ODMAN,

ET AL.


Dry Deposition Flux (mg/m2)

Oak Ridge - Nitric Acid Dry Deposition 40.0 30.0 20.0 10.0 0.0 Jul 1991 Jun 1992 Mar 1993 May 1993 Aug 1993 Feb 1994 Apr 1995 May 1995 Jul 1995

AIRMoN Figure 17-1

Model

Nitric acid dry deposition mass flux at Oak Ridge, TN

performance. These stations are Oak Ridge, TN (near Great Smoky Mountains) and Panola, GA (outside Atlanta). The model estimated values for the grid cells that contain each station were used for model performance evaluation.

17.2 Spatial Plots of 7-Day Cumulative Dry Deposition Spatial plots or maps of simulated weekly cumulative dry deposition fluxes were prepared for each episode. Only the portion of the modeling domain over the SAMI region is shown in these maps. Figure 17-1 shows the map of sulfur dioxide (SO2) dry deposition for the week of July 11-18, 1995. The 2 maximum dry deposition flux of SO2 is 136 mg/m and is located in West Virginia. Maps of dry deposition fluxes of sulfate, nitrate, SO2 and nitric acid (HNO3) for all episodes can be found in Appendix A.

Dry Deposition Flux (mg/m2)

Oak Ridge - Sulfur Dioxide Dry Deposition 40.0 30.0 20.0 10.0 0.0 Jul 1991 Jun 1992 Mar 1993 May 1993 Aug 1993 Feb 1994 Apr 1995 May 1995 Jul 1995

AIRMoN Figure 17-2

Sulfur dioxide dry deposition mass flux at Oak Ridge, TN. 17-2

Model

ODMAN,

ET AL.

Table 17-1


Performance statistics for the dry deposition species (all episodes).

Species

Mean Flux (mg m-2)

MB (mg m-2 )

NMB (%)

ME (mg m -2 )

NME (%)

Sulfur Dioxide (gas)

10.484

7.662

73.080

10.940

104.349

Nitric Acid (gas)

17.016

4.915

28.886

10.164

59.733

Sulfate (PM)

3.361

-0.506

-15.050

1.422

42.306

Nitrate (PM)

0.068

0.818

1201.433

0.818

1201.433

17.3 7-Day Cumulative Dry Deposition at AIRMoN Stations Figure 17-2 and Figure 17-1 contain modeled and observed sulfur dioxide and nitric acid dry deposition mass fluxes at Oak Ridge. These figures include results for all the episodes that AIRMoN measurements were taken. Similar charts for other species as well as charts of deposition at Panola, GA can be found in Appendix A.

17.4 Discussion of Performance Results Performance statistics were calculated using the same definitions for normalized mean bias and error as were used for wet deposition performance in Section 7.5. A total of 17 observations at two stations were used to calculate the mean mass flux, mean bias, normalized mean bias, mean error, and normalized mean error for the species listed in Table 17-1. It should be noted that the nitric acid measurements are highly uncertain due to the filter pack system used by AIRMoN. The front filter, intended to collect particles, can also capture nitric acid. In addition, nitrate can volatilize off the front filter and be recaptured by the nitric acid filter. Sulfur dioxide and nitric acid are the primary species for the dry deposition of sulfur and oxidized nitrogen, respectively. Dry deposition of sulfur dioxide, nitric acid, and nitrate PM are typically over estimated, while sulfate PM is underestimated. The normalized mean error for sulfur dioxide is around 100% and the normalized mean error for nitric acid is around 60%. The normalized mean error for nitrate PM is over 1000% due to an over prediction in nitrate concentrations and the observations being extremely small.

17-3

ODMAN,

ET AL.



17-4

ODMAN,

ET AL.


18 APPENDIX D: Comparison of DDM-3D and Brute-Force Methods The sensitivities calculated by using DDM-3D and "brute-force" methods were compared and the findings were summarized in Section 11.3. Recall that this comparison was performed for the July 11-19, 1995 episode using 30% domain-wide emission reductions with respect to the 2010-OTW control strategy (discussed in Section 9.2.1). The sensitivities of ozone, NH3, PM2.5 and its sulfate, nitrate, ammonium and organic components and wet deposition of sulfate, nitrate, ammonium, calcium and magnesium to the emissions of SO2, NOx , VOCs, and NH3 were calculated using both DDM-3D and "brute-force" methods. Spatial plots were prepared for July 12 and 15, 1995 and presented to the SAMI atmospheric modeling subcommittee. The subcommittee rated the agreement between DDM-3D and "brute-force" sensitivities as "Excellent", “Good”, “Fair”, or “Poor”. The results of this subjective analysis are presented below. Ozone to SO2 Emissions: N/A Range: 0.5 ppb decrease Direction: Poor Spatially: Poor Magnitude: Poor Notes: Small change compared to NOx emissions. Ozone to NOx Emissions: Good Range: 10 ppb decrease to 10 ppb increase (urban) Direction: Excellent Spatially: Excellent Magnitude: Excellent Notes: Ozone to NH3 Emissions: N/A Range: 0.2 ppb decrease to 0.2 ppb increase Direction: Poor Spatially: Poor Magnitude: Poor Notes: Small change compared to NOx emissions. Ozone to VOC Emissions: Good Range: 2 ppb increase Direction: Excellent Spatially: Excellent Magnitude: Good Notes: SO2 to SO2 Emissions: Good Range: 14 ppb decrease Direction: Excellent Spatially: Excellent Magnitude: Excellent Notes:

18-1

ODMAN,

ET AL.


NH3 to NH3 Emissions: Good Range: 18 ppb decrease Direction: Excellent Spatially: Excellent Magnitude: Fair - DDM shows slightly less sensitivity than Brute Force Notes: Particulate Sulfate to SO2 Emissions: Good 3 Range: 5 µg/m decrease Direction: Excellent Spatially: Excellent Magnitude: Excellent Notes: Particulate Sulfate to NOx Emissions: N/A 3 3 Range: 0.1 µg/m decrease to 0.1 µg/m increase Direction: Fair Spatially: Fair Magnitude: Fair Notes: Brute Force has numerical noise – sensitivity small compared to SO2 emission reductions. Particulate Sulfate to NH3 Emissions: Fair 3 Range: 0.75 µg/m decrease Direction: Good Spatially: Fair Magnitude: Good Notes: By comparing the 30% reduction sensitivity results to the central difference results, it can be seen that the Brute Force has numerical noise. Particulate Sulfate to VOC Emissions: N/A 3 Range: 0.5 µg/m decrease Direction: Poor Spatially: Poor Magnitude: Poor Notes: DDM shows little sensitivity – Brute Force sensitivities may be the result of numerical noise - sensitivity small compared to SO2 emission reductions. Particulate Nitrate to SO2 Emissions: Fair 3 Range: 1.5 µg/m increase Direction: Good Spatially: Good Magnitude: Fair – DDM less sensitive than Brute Force Notes: Particulate Nitrate to NOx Emissions: Good 3 Range: 0.9 µg/m decrease Direction: Good Spatially: Good Magnitude: Good Notes:

18-2

ODMAN,

ET AL.


Particulate Nitrate to NH3 Emissions: Poor 3 Range: 0.8 µg/m decrease Direction: Good Spatially: Good Magnitude: Poor – Brute Force sensitivity is 2 - 3 times larger than DDM Notes: Particulate Nitrate to VOC Emissions: Fair 3 Range: 0.9 µg/m increase Direction: Good Spatially: Good Magnitude: Fair – DDM shows less sensitivity than Brute Force Notes: Particulate Ammonium to SO2 Emissions: Good 3 Range: 1.4 µg/m decrease Direction: Good Spatially: Good Magnitude: Good – DDM shows slightly less sensitivity than Brute Force Notes: Particulate Ammonium to NOx Emissions: Good 3 Range: 0.3 µg/m decrease Direction: Good Spatially: Good Magnitude: Good – DDM shows slightly less sensitivity than Brute Force Notes: Particulate Ammonium to NH3 Emissions: Fair 3 Range: 0.8 µg/m decrease Direction: Good Spatially: Good Magnitude: Fair – Brute Force sensitivity is 2 times larger than DDM Notes: DDM and Brute Force match better when compared to central difference results – may be non-linear response. Particulate Ammonium to VOC Emissions: Fair 3 Range: 0.3 µg/m decrease Direction: Good Spatially: Good Magnitude: Fair – DDM shows less sensitivity than Brute Force Notes: Particulate Organic Carbon to SO2 Emissions: N/A 3 Range: 0.05 µg/m increase Direction: Fair Spatially: Poor Magnitude: Fair Notes: Sensitivity small compared to VOC and NOx emission reductions.

18-3

ODMAN,

ET AL.


Particulate Organic Carbon to NOx Emissions: Good 3 Range: 0.8 µg/m decrease Direction: Good Spatially: Good Magnitude: Good Notes: Particulate Organic Carbon to NH3 Emissions: N/A 3 Range: 0.04 µg/m increase Direction: Fair Spatially: Fair Magnitude: Fair Notes: Sensitivity small compared to VOC and NOx emission reductions. Particulate Organic Carbon to VOC Emissions: Good 3 Range: 3.5 µg/m decrease Direction: Good Spatially: Good Magnitude: Good Notes: Particulate Elemental Carbon to SO2, NOx, NH3, and VOC Emissions: N/A 3 3 Range: 0.005 µg/m decrease to 0.005 µg/m increase Direction: N/A Spatially: N/A Magnitude: N/A 3 Notes: Small sensitivity compared to EC concentrations (0.5 µg/m ) – DDM shows no sensitivity – Brute Force sensitivities could be numerical noise. Particulate Soils to SO2, NOx, NH3, and VOC Emissions: N/A 3 3 Range: 0.02 µg/m decrease to 0.02 µg/m increase Direction: N/A Spatially: N/A Magnitude: N/A 3 Notes: Small sensitivity compared to Soils concentrations (0.5 – 2.0 µg/m ) – DDM shows no sensitivity – Brute Force sensitivities could be numerical noise. PM2.5 to SO2 Emissions: Good 3 Range: 5.5 µg/m decrease Direction: Good Spatially: Good Magnitude: Good 2+ Notes: Good agreement between DDM and Brute Force for SO4 and NH4 sensitivities. PM2.5 to NOx Emissions: Good 3 Range: 1.7 µg/m decrease Direction: Good Spatially: Good Magnitude: Good + Notes: Good agreement between DDM and Brute Force for NO3 , NH4 , and Organic Carbon sensitivities.

18-4

ODMAN,

ET AL.


PM2.5 to NH3 Emissions: Poor 3 Range: 1.5 µg/m decrease Direction: Good Spatially: Good Magnitude: Poor – Brute Force sensitivity is 2 - 4 times larger than DDM + Notes: Poor agreement between DDM and Brute Force for NO3 and NH4 sensitivities. PM2.5 to VOC Emissions: Good 3 Range: 4.6 µg/m decrease Direction: Good Spatially: Good Magnitude: Good + Notes: Good agreement between DDM and Brute Force for NO3 , NH4 , and Organic Carbon sensitivities. Wet Sulfate to SO2 Emissions: Good 2 Range: 60 mg/m decrease Direction: Good Spatially: Good Magnitude: Good Notes: Wet Sulfate to NOx Emissions: N/A 2 2 Range: 1.0 mg/m decrease to 1.0 mg/m increase Direction: N/A Spatially: N/A Magnitude: Poor – DDM shows very little sensitivities Notes: Change in wet sulfate is small compared to SO2 emission reductions. Wet Sulfate to NH3 Emissions: N/A 2 Range: 3.4 mg/m decrease Direction: N/A Spatially: N/A Magnitude: Poor – DDM shows very little sensitivities Notes: By comparing the 30% reduction sensitivity results to the central difference results, it can be seen that the Brute Force has numerical noise – system may not be linear - change in wet sulfate is small compared to SO2 emission reductions. Wet Sulfate to VOC Emissions: N/A 2 2 Range: 18 mg/m decrease to 2 mg/m increase Direction: N/A Spatially: N/A Magnitude: Poor 2 2 Notes: Brute Force shows little sensitivity (1 mg/m decrease to 1 mg/m increase) except at a couple of nodes – sensitivity at these nodes could be due to numerical noise. Wet Nitrate to SO2 Emissions: Poor 2 Range: 15 mg/m decrease Direction: Good Spatially: Good Magnitude: Poor - Brute Force sensitivity is 10 times larger than DDM Notes:

18-5

ODMAN,

ET AL.


Wet Nitrate to NOx Emissions: Good 2 Range: 8 mg/m decrease Direction: Good Spatially: Good Magnitude: Good Notes: Wet Nitrate to NH3 Emissions: Poor 2 Range: 12 mg/m decrease Direction: Good Spatially: Good Magnitude: Poor - Brute Force sensitivity is 20 times larger than DDM Notes: Wet Nitrate to VOC Emissions: Good 2 Range: 9 mg/m decrease Direction: Good Spatially: Good Magnitude: Good – DDM shows slightly less sensitivity than Brute Force Notes: Wet Ammonium to SO2 Emissions: Poor 2 Range: 10 mg/m decrease Direction: Good Spatially: Good Magnitude: Poor - Brute Force sensitivity is 2 times larger than DDM Notes: Wet Ammonium to NOx Emissions: Poor 2 Range: 3 mg/m decrease Direction: Fair Spatially: Fair Magnitude: Poor - Brute Force sensitivity is 5 times larger than DDM Notes: Wet Ammonium to NH3 Emissions: Poor 2 Range: 13 mg/m decrease Direction: Good Spatially: Good Magnitude: Poor - Brute Force sensitivity is 5 times larger than DDM Notes: Wet Ammonium to VOC Emissions: Poor 2 Range: 2.3 mg/m increase Direction: Fair Spatially: Fair Magnitude: Poor - Brute Force sensitivity is 4 times larger than DDM Notes:

18-6

ODMAN,

ET AL.


Wet Calcium to SO2, NOx, NH3, and VOC Emissions: N/A 2 Range: 0.03 mg/m decrease to 0.03 mg/m2 increase Direction: N/A Spatially: N/A Magnitude: N/A 2 Notes: Small sensitivity compared to wet calcium deposition flux (0.5 – 4.0 mg/m ) – DDM shows no sensitivity – Brute Force sensitivities could be numerical noise. Wet Magnesium to SO2, NOx, NH3, and VOC Emissions: N/A 2 2 Range: 0.003 mg/m decrease to 0.003 mg/m increase Direction: N/A Spatially: N/A Magnitude: N/A 2 Notes: Small sensitivity compared to wet magnesium deposition flux (0.1 – 1.0 mg/m ) – DDM shows no sensitivity – Brute Force sensitivities could be numerical noise.

18-7