&EPA
United S*s
Eiwiroimerilfll Protection
Agancy
Technical Report on Ozone Exposure, Risk,
and Impact Assessments for Vegetation
-------
EPA 452/R-07-002
January 2007
Technical Report on Ozone Exposure, Risk, and Impact Assessments for
Vegetation
By:
Jonathan A. Lehrer
Melanie Bacou
Brian Blankespoor
Don McCubbin
Jason Sacks
C. Robert Taylor
David A. Weinstein
Abt Associates, Inc.
Bethesda, MD
Prepared for:
Nancy Riley, Project Officer
Victoria Sandiford, Work Assignment Manager
Jeffrey Herrick, Alternate Work Assignment Manager
Health and Environmental Impacts Division
Contract No. 68-D-03-002
Work Assignment 3-38, TDF #2
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Ambient Standards Group
Research Triangle Park, NC
-------
Table of Contents
Table of Contents i
List of Tables iv
List of Figures vi
List of Maps vii
Introduction 1-1
1. Definition of O3 Metrics 1-1
1.1 Daily SUM06 and Annual Maximum 3-Month SUM06 1-2
1.2 Daily Maximum 8-hour Average and Annual 4th Highest Daily Maximum 8-hour Average 1-
2
1.3 Daily W126 and Annual Maximum 3-month 12-hour W126 1-2
1.4 Terminology 1-3
2. Input Data 2-1
2.1 Monitor Data 2-1
2.2 Model Data 2-2
3. Generating a National Potential O3 Exposure Surface (POES) 3-1
3.1 Composite CMAQ Grid 3-1
3.2 Interpolation Approaches 3-1
3.2.1 Voronoi Neighbor Averaging (VNA) 3-1
3.2.2 Enhanced Voronoi Neighbor Averaging (eVNA) 3-1
3.2.3 Four Types of Condition-Specific Adjustment Factors 3-2
3.2.4 Interpolating Hourly Data vs. Metrics 3-3
3.2.5 Summary of Approaches 3-4
3.3 Evaluation of Interpolation Approaches 3-4
3.4 Generation of the National Potential O3 Exposure Surface 3-12
4. Rollback 4-1
4.1 Scenarios 4-1
4.2 Rollback Methodologies 4-1
4.2.1 8-hour Maximum Quadratic Rollback 4-1
4.2.2 SUM06 Rollback 4-2
4.2.3 W126 Rollback 4-2
4.3 Aggregate Results 4-3
4.4 Air Quality Maps For Alternative Ozone Standards 4-3
5. Crop and Tree Exposure 5-1
5.1 Ozone Concentration-Response Functions 5-5
5.2 Derivation of Seasonal Ozone Indices for Crop and Tree Species 5-9
5.2.1 Ten Percent Adjustment 5-9
5.2.2 Characterization of Crop and Tree Growing Seasons 5-10
5.3 Characterization of Crop and Tree Species Growing Ranges 5-11
Abt Associates Inc. Contents i
-------
5.4 Estimation of Crop Yield and Tree Seedling Biomass Loss 5-12
5.4.1 Yield Impact on Cotton and Soybean 5-12
5.4.2 Yield Response Maps for Cotton and Soybean 5-15
5.4.3 Biomass Impact on Tree Seedlings 5-30
5.4.4 Seedling Biomass Response Maps for Selected Tree Species 5-33
6. Economic Benefits 6-1
6.1 AGSIM© 6-1
6.2 Methodology 6-1
6.3 Results 6-1
7. Tree Growth Simulation 7-1
7.1 Summary of Results 7-1
Appendix A Detailed Results from Dropout Monitor Investigation A-l
Choice of Interpolation Approaches A-l
Choice of Dropout Monitor Sites A-l
Results A-l
Appendix B Memorandum Regarding the Quadratic Rollback B-l
Appendix C Numerical Examples of VNA and eVNA C-l
Appendix D Air Quality Maps D-l
Appendix E Interpolating State Growing Seasons E-l
Appendix F Growing Seasons for Major Crops by State F-l
Appendix G Summary Statistics for Crop Yield Concentration Responses based on W126
Metric G-l
G. 1 Crop Yield Concentration Responses based on Ten Percent Adjusted W126 Metric G-l
G. 1.1 Median Concentration Responses for Crops G-2
G.I.2 Summary Statistics of Concentration Responses for Crops G-6
G. 1.3 Yield Response Maps for Selected Field Crops based on Ten Percent Adjusted W126
Metric G-12
G.2 Crop Yield Concentration Responses based on Non-Adjusted W126 Metric G-21
G.2.1 Median Concentration Responses for Crops G-21
G.2.2 Summary Statistics of Concentration Responses for Crops G-25
G.2.3 Yield Response Maps for Selected Field Crops based on Non-Adjusted W126 Metric.
G-33
Appendix H Summary Statistics for Tree Seedling Biomass Responses based on Ten Percent
Adjusted W126 Metric H-l
H.I Tree Seedling Biomass Responses based on Ten Percent Adjusted W126 Metric H-l
H.I.I Median Concentration Responses for Crops H-2
H. 1.2 Summary Statistics of Concentration Responses for Tree Seedlings H-8
H.2.3 Seedling Biomass Response Maps for Selected Tree Species: H-10
H.2 Tree Seedling Biomass Responses based on Non-Adjusted W126 Metric H-ll
H.2.1 Median Concentration Responses for Crops H-ll
H.2.2 Summary Statistics of Concentration Responses for Tree Seedlings H-l7
H.2.3 Seedling Biomass Response Maps for Selected Tree Species: H-20
Abt Associates Inc. Contents ii
-------
Appendix I AGSIM© Model Specifications 1-1
Model Specification 1-1
Supply Components 1-1
Market Clearing Identities 1-6
Historical Observation Period 1-7
Alternative Specifications Considered 1-7
Baseline 1-7
Regional Effects Sub-Model 1-8
AGSIM© Output 1-9
Uncertainty 1-9
Appendix J TREGRO simulations of red maple and yellow poplar trees under scenarios of
reduced ozone exposure at two locations in the southern Appalachian Mountains J-l
Procedure J-l
Results J-4
Uncertainty J-6
Appendix K Report from the Preliminary Evaluation of O3 Interpolation Approaches using
53 monitored dropout sites K-l
K.I. Introduction K-l
K.2 Methods K-l
K.2.1 Monitor Data K-2
K.2.2 Model Data K-3
K.2.3 Interpolation Approaches K-3
K.2.4 Choice of Dropout Monitor Sites K-6
K.2.5 Evaluation Criteria K-7
K.3. Results K-9
K.3.1 Eastern Dropout Results K-9
K. 3.2 Results of Western Monitors K-13
K.3.3 Monitor-Level Prediction Data K-14
Appendix L Comparison of Hour and Metric Interpolation Approaches L-l
Example of Over-prediction in Metric-based SUM06 Interpolation L-l
An Example of Under-prediction in Hour-based SUM06 Interpolation L-3
References R-l
Abt Associates Inc. Contents iii
-------
List of Tables
Table 3-1: Interpolation Approaches Considered 3-4
Table 3-2: Results from Previous Investigation: Eastern US (41 locations) 3-4
Table 3-3: Results from Previous Investigation: Western US (12 locations) 3-5
Table 3-4: Results from Current Investigation: Eastern US: 8-hour Maximum (approx 800 locations) 3-6
Table 3-5: Results from Current Investigation: Eastern US: SUM06 (approx 800 locations) 3-7
Table 3-6: Results from Current Investigation: Eastern US: W126 (approx 800 locations) 3-8
Table 3-7: Results from Current Investigation: Western US: 8-hour Maximum (approx. 300 locations) 3-9
Table 3-8: Results from Current Investigation: Western US: SUM06 (approx. 300 locations) 3-10
Table 3-9: Results from Current Investigation: Western US: W126 (approx. 300 locations) 3-11
Table 4-1 Rollback Scenarios Described in Terms of Mean, Median, Maximum, and Minimum 8-Hour Maximum
Values 4-3
Table 4-2 Rollback Scenarios Described in Terms of Mean, Median, Maximum, and Minimum SUM06 Values... 4-3
Table 4-3 Rollback Scenarios Described in Terms of Mean, Median, Maximum, and Minimum W126 Values 4-3
Table 5-1: Composite SUM06 Relative Yield Loss Functions for Major Crops, Beans, Lettuce, and Potatoes 5-6
Table 5-2: Composite W126 Relative Yield Loss Functions for Major Crops, Beans, Lettuce, and Potatoes 5-6
Table 5-3: Composite 12-hour-average and 7-hour-average Relative Yield Loss Functions for Fruits and Other
Vegetables 5-7
Table 5-4: Median SUM06 Relative Yield Loss Functions for Tree Seedlings 5-7
Table 5-5: Median W126 Relative Yield Loss Functions for Tree Seedlings 5-7
Table 5-6: Growing Ranges Reported in 2002 Census of Agriculture vs. 2001 Crop County Data for Major Crops. 5-
12
Table 5-7: Median Yield Gain from Baseline for Cotton (reduced W126) 5-14
Table 5-8: Maximum Yield Gain from Baseline for Cotton (reduced W126) 5-14
Table 5-9: Minimum Yield Gain from Baseline for Cotton (reduced W126) 5-14
Table 5-10: Median Yield Gain from Baseline for Soybean (reduced W126) 5-14
Table 5-11: Maximum Yield Gain from Baseline for Soybean (reduced W126) 5-15
Table 5-12: Minimum Yield Gain from Baseline for Soybean (reduced W126) 5-15
Table 5-10: Median Yield Gain from Baseline for Aspen (reduced W126) 5-32
Table 5-11: Median Yield Gain from Baseline for Black Cherry (reduced W126) 5-32
Table 5-12: Minimum Yield Gain from Baseline for Ponderosa Pine (reduced W126) 5-32
Table 6-1: Total Undiscounted Economic Surplus Effect, 2001/02 through 2014/15 (with lOpct adjustment) 6-2
Table 6-2: Total Undiscounted Economic Surplus Effect of Soybean Yield Response, 2001/02 through 2014/15
(with lOpct adjustment) 6-3
Table 6-3: Total Undiscounted Economic Surplus Effect, 2001/02 through 2014/15 (with lOpct adjustment) 6-4
Table 6-4: Total Undiscounted Economic Surplus Effect of Soybean Yield Response, 2001/02 through 2014/15
(without lOpct adjustment) 6-5
Table 6-5: Comparison between Adjusted vs. Non-Adjusted Metric on Undiscounted Economic Surplus Effects,
2001/02 through 2014/15 6-6
Table 6-6: Comparison Between SUM06 vs. W126 Metric on Undiscounted Economic Surplus Effects, 2001/02
through 2014/15 6-6
Table 7-1 Predicted percent increases in total tree growth over a 3-year period under the 4 ozone (O3) reduction
scenarios 7-1
Table A-l: Comparison of Interpolation Methods for California. (8-hour Maximum) A-2
Table A-2: Comparison of Interpolation Methods for California. (SUM06) A-3
Table A-3: Comparison of Interpolation Methods for California. (W126) A-4
Table A-4: Comparison of Interpolation Methods for the Western US Excluding California (8-hour Maximum).. A-5
Table A-5: Comparison of Interpolation Methods for the Western US Excluding California (SUM06) A-6
Table A-6: Comparison of Interpolation Methods for the Western US Excluding California (W126) A-7
Table G-l: Yield Losses for Crops, Fruits and Vegetables -2001 Baseline (reduced W126) G-6
Table G-2: Yield Gains from Baseline for Crops, Fruits and Vegetables - 84 ppb 8-Hour Rollback (reduced W126)
G-6
Abt Associates Inc. Contents iv
-------
Table G-3: Yield Gains from Baseline for Crops, Fruits and Vegetables - 70 ppb 8-Hour Rollback (reduced W126)
G-7
Table G-4: Yield Gains from Baseline for Crops, Fruits and Vegetables - 25 ppm-hr SUM06 Rollback (reduced
W126) G-7
Table G-5: Yield Gains from Baseline for Crops, Fruits and Vegetables - 15 ppm-hr SUM06 Rollback (reduced
W126) G-8
Table G-6: Yield Gains from Baseline for Crops, Fruits and Vegetables - 21 ppm-hr W126 Rollback (reduced
W126) G-9
Table G-7: Yield Gains from Baseline for Crops, Fruits and Vegetables - 13 ppm-hr W126 Rollback (reduced
W126) G-9
Table G-8: Absolute Crop Yield Losses-84 ppb 8-Hour Rollback (reduced W126) G-10
Table G-9: Absolute Crop Yield Losses-70 ppb 8-Hour Rollback (reduced W126) G-10
Table G-10: Absolute Crop Yield Losses - 25 ppm-hr SUM06 Rollback (reduced W126) G-ll
Table G-ll: Absolute Crop Yield Losses - 15 ppm-hr SUM06 Rollback (reduced W126) G-12
Table G-12: Yield Losses for Crops, Fruits and Vegetables - 2001 Baseline (unreduced W126) G-25
Table G-13: Yield Gains from Baseline for Crops, Fruits and Vegetables - 84 ppb Rollback (unreduced W126) G-27
Table G-14: Yield Gains from Baseline for Crops, Fruits and Vegetables - 70 ppb Rollback (unreduced W126) G-28
Table G-15: Yield Gains from Baseline for Crops, Fruits and Vegetables - 25 ppm-hr Rollback (unreduced W126)
G-29
Table G-16: Yield Gains from Baseline for Crops, Fruits and Vegetables - 15 ppm-hr W126 Rollback (unreduced
W126) G-30
Table G-17: Yield Gains from Baseline for Crops, Fruits and Vegetables - 21 ppm-hr W126 Rollback (unreduced
W126) G-31
Table G-18: Yield Gains from Baseline for Crops, Fruits and Vegetables - 13 ppm-hr W126 Rollback (unreduced
W126) G-32
Table H-l: Absolute Tree SeedlingBiomass Loss - 2001 Baseline (reduced W126) H-8
Table H-2: Median Tree Seedling Biomass Gain-84 ppb Rollback (reduced W126) H-9
Table H-3: Median Tree Seedling Biomass Gain-70 ppb Rollback (reduced W126) H-9
Table H-4: Median Tree Seedling Biomass Gain-25 ppm-hr Rollback (reduced W126) H-9
Table H-5: Median Tree Seedling Biomass Gain- 15 ppm-hr Rollback (reduced W126) H-10
Table H-6: Median Tree Seedling Biomass Gain-21 ppm-hr Rollback (reduced W126) H-10
Table H-7: Median Tree Seedling Biomass Gain- 13 ppm-hr Rollback (reduced W126) H-10
Table H-8: Tree Seedling Biomass Loss - 2001 Baseline (unreduced W126) H-17
Table H-9: Median Tree Seedling Biomass Gain from Baseline - 84 ppb Rollback (unreduced W126) H-18
Table H-10: Median Tree Seedling Biomass Gain from Baseline - 70 ppb Rollback (unreduced W126) H-18
Table H-l 1: Median Tree Seedling Biomass Gain from Baseline - 25 ppm-hr Rollback (unreduced W126) H-18
Table H-12: Median Tree Seedling Biomass Gain from Baseline - 15 ppm-hr Rollback (unreduced W126) H-19
Table H-13: Median Tree Seedling Biomass Loss - 21 ppm-hr W126 Rollback (unreduced W126) H-19
Table H-14: Median Tree Seedling Biomass Gain from Baseline - 13 ppm-hr W126 Rollback (unreduced W126) H-
19
Table J-l Predicted percent increases in total tree growth over a 3-year period under the 4 ozone (O3) reduction
scenarios J-4
Table K-l Summary of O3 Prediction Accuracy by Region and Metric K-9
Table K-2 Evaluation Statistics for Eastern Dropout Monitors K-10
Table K-3 Evaluation Statistics for Adjusted Neighbor Values At Eastern Dropout Monitors, by Distance from
Neighbors K-ll
Table K-4 Evaluation Statistics for Eastern Dropout Monitors, by Low/High O3 Levels K-12
Table K-5 Evaluation Statistics for Western Dropout Monitors K-13
Table K-6 Evaluation Statistics for Adjusted Neighbor Values At Western Dropout Monitors, by Distance from
Neighbors K-13
Table K-7 Evaluation Statistics for Western Dropout Monitors, by Low/High O3 Levels K-13
Table K-8 Model Predictions K-15
Table K-9 Hour-VNA Predictrions K-16
Table K-10 Hour-Month-Decile Predictions K-17
Table K-ll Hour-Month-Hour Predictions K-18
Table K-12 Hour-Season-Decile Predictions K-19
Table K-13 Hour-Season-Hour Predictions K-20
Abt Associates Inc. Contents
-------
List of Figures
Figure 1-1: Weighting Function Used to Calculate W126 Exposure Index (SUM06 weighting shown in dotted line)
1-3
Figure 5-1: Shares of Value of Production for Selected Crops, Fruits, and Vegetables (2001, total value is $53
billion). Source: NASS 5-2
Figure 5-2: Estimation of Yield and Biomass Losses-Data Sources and Data Flow 5-4
Figure 5-3: Relative W126 Concentration-Response Curves for Major Crops (percent yield loss/ppb) 5-8
Figure 5-4: Relative W126 Concentration-Response Curves for Selected Tree Species (percent biomass loss/ppb)5-9
Figure 5-5: Median Yield Gain from Baseline for Cotton (reduced W126) 5-13
Figure 5-6: Median Yield Gain from Baseline for Soybean (reduced W126) 5-13
Figure 5-5: Median Biomass Gain from Baseline for Aspen (reduced W126) 5-30
Figure 5-5: Median Biomass Gain from Baseline for BlackCherry (reduced W126) 5-31
Figure 5-5: Median Biomass Gain from Baseline for Ponderosa Pine (reduced W126) 5-31
Figure 7-1 Tree growth response of red maple and yellow poplar in forests of Shenandoah National Park, Virginia
and Cranberry, North Carolina to ozone (O3) reduction scenarios 7-2
Figure F-l: Typical Harvest Seasons for Sorghum by State F-l
Figure F-2: Typical Harvest Seasons for Cotton by State F-l
Figure F-3: Typical Harvest Seasons for Soybean by State F-2
Figure F-4: Typical Harvest Seasons for Winter Wheat by State F-2
Figure G-l: Yield Losses for Crops - 2001 Baseline (W126, Median C-R) G-2
Figure G-2: Yield Gains from Baseline for Crops - 84 ppb Rollback (W126, Median C-R) G-3
Figure G-3: Yield Gains fromBaseline for Crops - 70 ppb Rollback (W126, Median C-R) G-3
Figure G-4: Yield Gains from Baseline for Crops - 25 ppm-hr Rollback (W126, Median C-R) G-4
Figure G-5: Yield Gains fromBaseline for Crops - 15 ppm-hr Rollback (W126, Median C-R) G-4
Figure G-6: Yield Losses for Crops-21 ppm-hr W126 Rollback (W126, Median C-R) G-5
Figure G-7: Yield Losses for Crops - 13 ppm-hr W126 Rollback (reduced W126, Median C-R) G-5
Figure G-8: Yield Losses for Crops - 2001 Baseline (unreduced W126, Median C-R) G-21
Figure G-9: Yield Gains from Baseline - 84 ppb Rollback (unreduced W126, Median C-R) G-22
Figure G-10: Yield Gains fromBaseline - 70 ppb Rollback (unreduced W126, Median C-R) G-22
Figure G-ll: Yield Gains fromBaseline - 25 ppm-hr Rollback (unreduced W126, Median C-R) G-23
Figure G-12: Yield Gains from Baseline - 15 ppm-hr Rollback (unreduced W126, Median C-R) G-24
Figure G-13: Yield Gains from Baseline - 21 ppm-hr Rollback (unreduced W126, Median C-R) G-24
Figure G-14: Yield Gains from Baseline - 13 ppm-hr Rollback (unreduced W126, Median C-R) G-25
Figure H-l: Median Tree Seedling Biomass Loss -2001 Baseline (reduced W126) H-2
Figure H-2: Median Tree Seedling Biomass Gain from Baseline - 84 ppb Rollback (reduced W126) H-3
Figure H-3: Median Tree Seedling Biomass Gain from Baseline - 70 ppb Rollback (reduced W126) H-4
Figure H-4: Median Tree Seedling Biomass Gain from Baseline - 25 ppm-hr Rollback (reduced W126) H-5
Figure H-5: Median Tree Seedling Biomass Gain from Baseline - 15 ppm-hr Rollback (reduced W126) H-6
Figure H-6: Median Tree Seedling Biomass Gain from Baseline - 21 ppm-hr Rollback (reduced W126) H-7
Figure H-7: Median Tree Seedling Biomass Gain from Baseline - 13 ppm-hr Rollback (reduced W126) H-8
Figure H-8: Median Tree Seedling Biomass Loss -2001 Baseline (unreduced W126) H-ll
Figure H-9: Median Tree Seedling Biomass Gain from Baseline - 84 ppb Rollback (unreduced W126) H-12
Figure H-10: Median Tree Seedling Biomass Gain from Baseline - 70 ppb Rollback (unreduced W126) H-13
Figure H-ll: Median Tree Seedling Biomass Gain from Baseline -25 ppm-hr Rollback (unreduced W126) H-14
Figure H-12: Median Tree Seedling Biomass Gain from Baseline - 15 ppm-hr Rollback (unreduced W126) H-15
Figure H-13: Median Tree Seedling Biomass Gain from Baseline-21 ppm-hr Rollback (unreduced W126) H-16
Figure H-14: Median Tree Seedling Biomass Gain from Baseline - 13 ppm-hr Rollback (unreduced W126) H-17
Figure J-l Tree growth response of red maple and yellow poplar in forests of Shenandoah National Park, Virginia
and Cranberry, North Carolina to ozone (O3) reduction scenarios J-5
Figure K-l Location of "Dropout" Monitor Sites (Triangle = West; Pentagon = East) K-6
Figure K-2 Location of "Dropout" Monitor Sites and Other AQS and CASTNet Monitor Sites K-7
Abt Associates Inc. Contents vi
-------
List of Maps
Map 5-1: As-is Yield loss for Cotton (Gossypium hirsutum) 5-16
Map 5-2: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 4th Highest 8-hour
Maximum to 84ppb 5-17
Map 5-3: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 4th Highest 8-hour
Maximum to 70 ppb 5-18
Map 5-4: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 3-month SUM06 to 25
ppm-hour 5-19
Map 5-5: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 3-month SUM06 to 15
ppm-hour 5-20
Map 5-6: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback W126 to 21 ppm-hr
(W126) 5-21
Map 5-7: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback W126 to 13 ppm-hour
(W126) 5-22
Map 5-8: As-is Yield loss for Soybeans (Glycine max) 5-23
Map 5-9: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 4th Highest 8-hour Maximum
to 84 ppb 5-24
Map 5-10: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 4th Highest 8-hour
Maximum to 70 ppb 5-25
Map 5-11: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 3-month SUM06 to 25 ppm-
hour 5-26
Map 5-12: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 3-month SUM06 to 15 ppm-
hour 5-27
Map 5-13: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback W126 to 21 ppm-hour
(W126) 5-28
Map 5-14: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback W126 to 13 ppm-hour
(W126) 5-29
Map 5-15: As-is Biomass Loss for Aspen 5-33
Map 5-16: Biomass Gain from Baseline for Aspen. Quadratic Rollback 4th Highest 8-hour Maximum to 84 ppb... 5-
34
Map 5-17: Biomass Gain from Baseline for Aspen. Quadratic Rollback 4th Highest 8-hour Maximum to 70 ppb... 5-
35
Map 5-18: Biomass Gain from Baseline for Aspen. Quadratic Rollback 3-month SUM06 to 25 ppm-hour 5-36
Map 5-19: Biomass Gain from Baseline for Aspen. Quadratic Rollback 3-month SUM06 to 15 ppm-hour 5-37
Map 5-19: Biomass Gain from Baseline for Aspen. Quadratic Rollback W126 to 21 ppm-hour 5-38
Map 5-19: Biomass Gain from Baseline for Aspen. Quadratic Rollback W126 to 13 ppm-hour 5-39
Map 5-20: As-is Biomass Loss for Black Cherry (Prunus serotina) 5-40
Map 5-21: Biomass Gain from Baseline for Black Cherry (Prunus serotina). Quadratic Rollback 4th Highest 8-hour
Maximum to 84ppb 5-41
Map 5-22: Biomass Gain from Baseline for Black Cherry (Prunus serotina). Quadratic Rollback 4th Highest 8-hour
Maximum to 70 ppb 5-42
Map 5-23: Biomass Gain from Baseline for Black Cherry (Prunus serotina). Quadratic Rollback 3-month SUM06 to
25 ppm-hour 5-43
Map 5-24: Biomass Gain from Baseline for Black Cherry (Prunus serotina). Quadratic Rollback 3-month SUM06 to
15 ppm-hour 5-44
Map 5-24: Biomass Gain from Baseline for Black Cherry (Prunus serotina). Quadratic Rollback 3-month W126 to
21 ppm-hour 5-45
Map 5-24: Biomass Gain from Baseline for Black Cherry (Prunus serotina). Quadratic Rollback 3-month W126 to
13 ppm-hour 5-46
Map 5-25: As is Biomass Loss for Ponderosa Pine (Pinus ponderosa) seedlings. April - October SUM06 5-47
Map 5-26: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 4th Highest 8-
hour Maximum to 84 ppb 5-48
Map 5-27: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 4th Highest 8-
hour Maximum to 70 ppb 5-49
Abt Associates Inc. Contents vii
-------
Map 5-28: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 3-month SUM06
to 25 ppm-hour 5-50
Map 5-29: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 3-month SUM06
to 15 ppm-hour 5-51
Map 5-29: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback W126 to 21
ppm-hour 5-52
Map 5-29: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback W126 to 13
ppm-hour 5-53
MapD-1: Ozone levels as-is D-2
Map D-2: Ozone levels reduced until the maximum 8 hour average at each location is 84ppb (ppm-h) D-3
Map D-3: Ozone levels reduced until the maximum 8 hour average at each location is 70ppb (ppm-h) D-4
Map D-4: Ozone levels reduced until the maximum 3-month SUM06 at each location is less than 25 ppm-h (parts
per million - hour) D-5
Map D-5: Ozone levels reduced until the maximum 3-month SUM06 at each location is less than 15 ppm-h (parts
per million - hour) D-6
MapE-1: U.S. Continental Climatic Classification E-l
Map E-2: U.S. States Climatic Classification E-l
Map G-l: As-is Yield Loss for Corn (Zea mays) (Reduced W126) G-13
Map G-2: As-is Yield loss for Wheat (Triticum aestivum) (Reduced W126) G-14
Map G-3: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 4th Highest 8-hour
Maximum to 84 ppb (Reduced W126) G-15
Map G-4: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 4th Highest 8-hour
Maximum to 70 ppb (Reduced W126) G-16
Map G-5: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 3-month SUM06 to 25
ppm-hour (Reduced W126) G-17
Map G-6: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 3-month SUM06 to 15
ppm-hour (Reduced W126) G-18
Map G-7: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback W126 to 21 ppm-hour
(Reduced W126) G-19
Map G-8: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback W126 to 13 ppm-hour
(Reduced W126) G-20
Map G-9: As-is Yield Loss for Corn (Zea mays) (Unreduced W126) G-33
Map G-10: As-is Yield loss for Cotton (Gossypium hirsutum) (Unreduced W126) G-34
Map G-l 1: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 4th Highest 8-hour
Maximum to 84 ppb (Unreduced W126) G-35
Map G-12: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 4th Highest 8-hour
Maximum to 70 ppb (Unreduced W126) G-36
Map G-13: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 3-month SUM06 to 25
ppm-hour (Unreduced W126) G-37
Map G-14: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback 3-month SUM06 to 15
ppm-hour (Unreduced W126) G-38
Map G-15: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback W126 to 21 ppm-hr
(Unreduced W126) G-39
Map G-16: Yield Gain from Baseline for Cotton (Gossypium hirsutum). Quadratic Rollback W126 to 13 ppm-hour
(Unreduced W126) G-40
Map G-17: As-is Yield loss for Soybeans (Glycine max) (Unreduced W126) G-41
Map G-18: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 4th Highest 8-hour
Maximum to 84 ppb (Unreduced W126) G-42
Map G-19: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 4th Highest 8-hour
Maximum to 70 ppb (Unreduced W126) G-43
Map G-20: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 3-month SUM06 to 25 ppm-
hour (Unreduced W126) G-44
Map G-21: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback 3-month SUM06 to 15 ppm-
hour (Unreduced W126) G-45
Map G-22: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback W126 to 21 ppm-hour
(Unreduced W126) G-46
Abt Associates Inc. Contents viii
-------
Map G-23: Yield Gain from Baseline for Soybeans (Glycine max). Quadratic Rollback W126 to 13 ppm-hour
(Unreduced W126) G-47
Map G-24: As-is Yield loss for Wheat (Triticum aestivum) (Unreduced W126) G-48
Map G-25: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 4th Highest 8-hour
Maximum to 84 ppb (Unreduced W126) G-49
Map G-26: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 4th Highest 8-hour
Maximum to 70 ppb (Unreduced W126) G-50
Map G-27: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 3-month SUM06 to 25
ppm-hour (Unreduced W126) G-51
Map G-28: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback 3-month SUM06 to 15
ppm-hour (Unreduced W126) G-52
Map G-29: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback W126 to 21 ppm-hour
(Unreduced W126) G-53
Map G-30: Yield Gain from Baseline for Wheat (Triticum aestivum). Quadratic Rollback W126 to 13 ppm-hour
(Unreduced W126) G-54
Map H-l: As is Biomass Loss for Ponderosa Pine (Pinus ponderosa) seedlings. April - October (unreduced W126).
H-20
Map H-2: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 4th Highest 8-
hour Maximum to 84 ppb (unreduced W126) H-21
Map H-3: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 4th Highest 8-
hour Maximum to 70 ppb (unreduced W126) H-22
Map H-4: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 3-month SUM06
to 25 ppm-hour (unreduced W126) H-23
Map H-5: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback 3-month SUM06
to 15 ppm-hour (unreduced W126) H-24
Map H-6: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback W126 to 21 ppm-
hour (unreduced W126) H-25
Map H-7: Biomass Gain from Baseline for Ponderosa Pine (Pinus ponderosa). Quadratic Rollback W126 to 13 ppm-
hour (unreduced W126) H-26
Map H-8: As-is Biomass Loss for Black Cherry (unreduced W126) H-27
Map H-9: Biomass Gain from Baseline for Black Cherry. Quadratic Rollback 4th Highest 8-hour Maximum to 84
ppb (unreduced W126) H-28
Map H-10: Biomass Gain from Baseline for Black Cherry. Quadratic Rollback 4th Highest 8-hour Maximum to 70
ppb (unreduced W126) H-29
Map H-l 1: Biomass Gain from Baseline for Black Cherry. Quadratic Rollback 3-month SUM06 to 25 ppm-hour
(unreduced W126) H-30
Map H-12: Biomass Gain from Baseline for Black Cherry. Quadratic Rollback 3-month SUM06 to 15 ppm-hour
(unreduced W126) H-31
MapH-13: Biomass Gain from Baseline for Black Cherry. Quadratic Rollback W126 to 21 ppm-hour (unreduced
W126) H-32
Map H-14: Biomass Gain from Baseline for Black Cherry. Quadratic Rollback W126 to 13 ppm-hour (unreduced
W126) H-33
Map H-15: As-is Biomass Loss for Aspen (unreduced W126) H-34
Map H-16: Biomass Gain from Baseline for Aspen. Quadratic Rollback 4th Highest 8-hour Maximum to 84 ppb
(unreduced W126) H-35
Map H-17: Biomass Gain from Baseline for Aspen. Quadratic Rollback 4th Highest 8-hour Maximum to 70 ppb
(unreduced W126) H-36
Map H-18: Biomass Gain from Baseline for Aspen. Quadratic Rollback 3-month SUM06 to 25 ppm-hour
(unreduced W126) H-37
Map H-19: Biomass Gain from Baseline for Aspen. Quadratic Rollback 3-month SUM06 to 15 ppm-hour
(unreduced W126) H-38
Map H-20: Biomass Gain from Baseline for Aspen. Quadratic Rollback W126 to 21 ppm-hour (unreduced W126).
H-39
Map H-21: Biomass Gain from Baseline for Aspen. Quadratic Rollback W126 to 13 ppm-hour (unreduced W126).
H-40
Abt Associates Inc. Contents ix
-------
Abt Associates Inc. Contents
-------
Introduction
During the last review of the secondary ozone (O3) NAAQS, as part of the development of the 1996 O3
Staff Paper (SP), EPA conducted analyses that assessed national O3 air quality, vegetation exposures and
risk, and impacts to economic benefits. At the time of the last review, large rural sections of the country
had little or no monitor coverage, including important growing regions for agricultural crops and forested
ecosystems. Since O3 monitor coverage in agricultural and rural/remote sites has changed little since the
last review, EPA must again rely on generated O3 air quality information in non-monitored areas to
provide national O3 exposure coverage. Given a number of recent air quality related developments, EPA
has decided to use a different method to generate a national exposure surface in this review.
In this report we present analyses of national O3 air quality, vegetation exposures and risk, and impact to
economic benefits that incorporates improved methods for estimating O3 at unmonitored locations. We
present quantitative evaluations of these new methods and an application of several such methods to
improve upon the results of the 1996 analysis. Ultimately, our purpose is to evaluate the economic
benefits associated with several alternative O3 standards currently under consideration.
The organization of the report is as follows:
Section 1 defines the O3 metrics used in this report
Section 2 describes the data used to produce this analysis
Section 3 describes the methods considered for estimating O3 at unmonitored locations; presents an
evaluation of these methods; and describes the generation of the Potential O3 Exposure Surface
(POES) under a method chosen from the options considered.
Section 4 presents the methods and results of the "rollback" procedure, which estimates hypothetical air
quality under several air quality standards that are currently being considered. It presents
descriptive statistics for all air quality scenarios (alternative scenarios as well as current air
quality as given by the POES).
Section 5 presents the methods and results of crop yield and tree seedling biomass loss estimates under
each air quality scenario.
Section 6 presents the methods and results of the evaluation of impacts to economic benefits.
Section 7 discusses the impacts on mature tree growth of just meeting various alternative O3 standards.
l. Definition of O3 Metrics
To quantify the overall O3 levels for a given time period, we used three O3 metrics, namely maximum 8-
hour average, 12-hour SUM06, and 12-hour W126. These can be calculated on the daily-level (e.g. "daily
maximum 8-hour average") as well as on the yearly-level (e.g. "annual 4th highest daily maximum 8-hour
average").
We define each of these metrics below. Since O3 monitor data often contains missing values, criteria are
given to determine if sufficient data exists to generate a valid metric. In certain cases, a valid metric can
be generated, but its value must be adjusted to compensate for missing values.
Abt Associates Inc Chapter 1-1
-------
1.1 Daily SUM06 and Annual Maximum 3-Month SUM06
The daily SUM06 metric is the sum of all O3 values greater than or equal to 0.06 parts per million (ppm)
observed from 8am-8pm. In order for a day to have a valid SUM06 value, 75 percent of the hours from
8am-8pm must be valid. To adjust for missing hourly O3 values, we scale SUM06 by the ratio of (number
of possible hourly O3 values) / (number of valid hourly O3 values)0.
The yearly SUM06 metric is the "annual maximum 3-month SUM06." To compute it, we calculate the
sum of all daily SUM06 values over all possible 3-month periods. To adjust for missing days, we scale
each monthly SUM06 value by the ratio of (number of days in the month) / (number of valid days)d. The
greatest of these 3-month SUM06 values is the annual maximum 3-month SUM06. In order for a 3-
month period to have a valid "3-month SUM06" value, each month in the 3-month period must have at
least 75 percent valid days. In order for a year to have a valid yearly SUM06 value (annual maximum 3-
month SUM06), it must have at least one 3-month period with a valid 3-month SUM06 value.
1.2 Daily Maximum 8-hour Average and Annual 4th Highest Daily Maximum 8-
hour Average
The daily maximum 8-hour average is calculated from rolling 8-hour averages of hourly O3 data, where a
valid 8-hour average must have 75 percent of a potential of eight hours in any given 8-hour period (i.e., at
least six hours out of eight)6. The daily maximum 8-hour average is the greatest of the day's 8-hour
averages. For a daily maximum 8-hour average to be considered valid, the day must have at least 75
percent of the potential 8-hour averages (i.e., 18 out of a potential of 24)f.
The yearly metric associated with the 8-hour maximum is the annual 4th highest daily maximum 8-hour
average. This is defined to be the 4th highest value amongst all of the valid daily 8-hour maximums
throughout the year. The value is truncated at the ppb level. Thus if the 4th highest value is 84.378 ppb,
the official value of the annual metric is actually 84 ppb.
1.3 Daily W126 and Annual Maximum 3-month 12-hour W126
The daily W126 metric is a weighted sum of all O3 values observed from 8am-8pm. More formally, daily
W126 is defined as:
i
-------
where Ci = hourly ozone concentration at hour /', and wc =
1
1 + 4403e
-126C,
The following figure shows the relationship between O3 concentration and weighting under W126, and
the equivalent weighting scheme for SUM06. Note that while SUM06 uses an all-or-nothing threshold,
W126 gradually increases the weight of O3 values as they grow in magnitude:
Weighting Function Used to Calculate W126 Exposure
Index
0 20 40 60 80 100 120 140 160 180 200
Ozone Concentration (ppb)
Figure 1-1: Weighting Function Used to Calculate W126 Exposure Index (SUM06 weighting shown
in dotted line)8
In order for a day to have a valid W126 value, 75 percent of the hours from 8am-8pm must be valid.
Daily W126 values are scaled to account for missing observations in the same fashion as daily SUM06
values.
Annual Maximum 3-month W126 is defined similarly to Annual Maximum 3-month SUM06. Namely it
is the sum of daily W126 values from the 3-month period which yields the highest such sum. Validity
criteria and scaling procedures for missing values are the same as those given for SUM06.
1.4 Terminology
To simplify the discussions that follow, we will use the generic term for a given metric (e.g. SUM06) to
refer to the annual statistic (e.g. Annual maximum 3-month 12-hour SUM06). When referring to daily
statistics, we will always preface the generic name with the word "daily" (e.g. daily SUM06, etc.).
g Lefohn A. S., J.A. Laurence, and R.J. Kohut. 1988. A comparison of indices that describe the relationship
between exposure to O3 and reduction in the yield of agricultural crops. Atmospheric Environment 22:1229-1240.
Abt Associates Inc
Chapter
1-3
-------
-------
2. Input Data
2.1 Monitor Data
The monitor data used in this analysis was taken from the Air Quality System (AQS) and Clean Air
Status and Trends Network (CASTNet) for the year 2001. AQS O3 data was taken from the file
RD_501_44201_2001.zip, and information on the monitors was taken from the file
AMP500_1994_FEB05.zip. Both are available at
http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm .
CASTNet O3 data was taken from http://ww.epa.gov/camdisOl/prepack/ozone_2001 .zip , and information
on CASTNet monitors can be found at http://cfpub.epa.gov/gdm/index.cfm .
Our initial monitor data set was comprised of hourly readings from 1194 monitors distributed across the
US.
Completeness Criteria
Data from a given monitor was used only if the monitor was deemed "complete," i.e., if it had valid O3
values for at least 50 percent11 of the hours during its region's O3 season. Note that O3 seasons vary by
geography and range from year-round (in California) to periods as short as June-September (Montana). In
all states except Texas, O3 regions follow state boundaries. Texas is unique in that it contains parts of 2
different O3 regions, each with its own O3 season. To simplify matters, we have applied the shorter of
these seasons to the entirety of Texas. Out of an initial set of 1194 monitors, 1192 qualify as complete.
At the recommendation of the EPA Work Assignment Contracting Officer's Representative (WACOR),
an additional 77 downtown urban monitors were not included in the analysis to minimize the impact of
inner-city O3 depletion caused by NOx scavenging.1. Several additional exclusions were made for
computational purposesj.
h For example, if the O3 season were May-September, then a valid monitor would have to have at least
1,836 hourly observations out of a potential total of 3762 (= 153 days x 24 hours). Out of 1,194 monitors,
all but two of them have at least 50 percent valid readings during their O3 season. We considered raising
this threshold to 75 percent. This would eliminate an additional 79 monitors, leaving about 93 percent of
the original monitors remaining. In the end, we chose to keep the threshold at 50% to maximize the
number of useable monitors.
1 The following urban monitors were eliminated (given by monitor id = statecode, countycode, siteid,
POC): 550790041, 550790026, 515100009, 510130020, 484391002, 483550025, 482011037, 482011035, 482011034,
482010075, 482010070, 481410055, 481410044, 481410037,481130069, 470370011,420450002,420030010,410052002,
390610040, 360810124, 360810098, 360810097, 360610063, 360610010, 360050083,350010023, 340170006,340130016,
320310016, 320032002, 320030021, 320030016, 295100072, 261630016, 250250042, 220330009, 220330003,201730010,
180970057, 180890022, 170311003, 170310072, 170310042, 170310032, 110010043,110010041,110010025,060950004,
060850004,060831008, 060750005, 060731007, 060670010, 060591003, 060410001, 060375001, 060374002, 060371301,
060371103,060371002, 060370113,060370030, 060290014, 060290010, 060170020, 060133001, 060131003, 060090001,
060010005, 051190007, 040190002, 040139997, 040134005, 040134003, 040133002, 010730023.
JThree monitors were excluded because they were co-located with other monitors (i.e., had identical latitude and
longitude values). The monitors kept were those with the highest POC codes. While this is not ideal, we believe the
effect in a pool of over 1000 monitors to be negligible. Additionally, changes had to be made to account for single
monitors whose data was used in both AQS and CASTNet databases (under different identifiers). For the sake of
evaluating O3 prediction approaches (see p 3-4), we eliminated AQS data from drawn from the same monitors that
supplied CASTNet data. (Leaving both "twin monitors" present would cause there to be nearly perfect predictions
Abt Associates Inc Chapter 2-1
-------
2.2 Model Data
We used two CMAQ modeling datasets, one with a resolution of 12km x 12km, the other with a
resolution of 36km x 36km. The 12-km CMAQ grid consists of 188 x 213 cells covering the Eastern U.S.
(bounded approximately on the west by the 99 line of longitude) excluding the northernmost parts of
Maine, Wisconsin, Minnesota, and South Dakota, and the southernmost parts of Florida and eastern
Texas. The 36-km CMAQ grid consists of 112 x 148 cells covering the entire continental US. Each
dataset gives hourly O3 values for each cellk.
made for those monitors, skewing the analysis). These monitors were still used in generating the POES (See p. 3-1).
While not ideal, their effect in generating the POES is not nearly as pernicious as their effect in the evaluation phase.
k All of the CMAQ data was provided by EPA in netCDF (Network Common Data Form) format. Steve Howard
from EPA provided a program to convert from netCDF to text.
Abt Associates Inc Chapter 2-2
-------
3. Generating a National Potential O3 Exposure Surface
(POES)
3.1 Composite CMAQ Grid
To generate a national Potential O3 Exposure Surface (POES) we needed a set of geographical locations
for which O3 data would be generated. Ideally, these locations would be regularly spaced, cover the
continental U.S., and be close enough to each other to provide a good spatial resolution. We chose to use
the regularly spaced grid structure of the CMAQ data as a basis for these locations. Specifically, we
generated O3 values for the center of each grid cell in the 12km x 12km grid (which covers only the
Eastern U.S.), and for those grid cells in the 36km x 36km grid whose centers fell within the boundaries
of the continental U.S., but did not fall within a 12km x 12km grid cell. In this fashion, we produced the
densest possible grid of CMAQ grid cell centers which provides non-redundant coverage of the
continental U.S.
3.2 Interpolation Approaches
A number of approaches for generating the POES were considered and evaluated. All were variations on
two techniques -Voronoi Neighbor Averaging (VNA) and Enhanced Voronoi Neighbor Averaging
(eVNA). The former is based only on monitor data, and the latter uses both monitoring and CMAQ
modeling data.
We examined 8 variants of eVNA, as well as 2 standard VNA approaches and several eVNA / VNA
blends. Based on their relative strengths in predicting known O3 values, we chose to use a VNA
interpolation approach in the East, and a blend of VNA and eVNA in the West. The remainder of this
section describes in detail all approaches considered. Section 3.3 presents a quantitative evaluation of
these approaches.
3.2.1 Voronoi Neighbor Averaging (VNA)
VNA uses distance-weighted averages of neighboring monitor data to arrive at predictions for a pre-
determined non-monitored site (in our case, the center of a CMAQ modeling grid cell). VNA identifies
neighboring monitors for each such site using a Voronoi Neighbor Algorithm (see Appendix C), and
takes an inverse-distance-weighted average of each neighbor's value for the data point in question (hourly
O3 value, daily metric, etc) to arrive at a prediction for that data point corresponding to the non-monitored
site in question.
3.2.2 Enhanced Voronoi Neighbor Averaging (eVNA)
The eVNA approach attempts to improve the accuracy of VNA predictions by taking into consideration
modeling predictions for the areas involved. To illustrate the rationale behind eVNA, we consider a
simple fictional example.
Suppose we wish to predict the O3 level at a hypothetical monitor at location X for a given hour. Location
X has two equidistant neighboring monitors, monitor A and monitor B. Monitor A reports 32 ppb, and
monitor B reports 20 ppb. A simple VNA approach would calculate the O3 at location X to be 26 ppb (the
average of 32 and 20, with equal weights given to the two equidistant neighbors).
Suppose, however, that CMAQ modeling data shows O3 levels at location X to be about twice that of O3
levels at either location A or location B. For example, suppose the average CMAQ O3 values for locations
Abt Associates Inc Chapter 3-1
-------
A and B are 15 ppb, whereas average CMAQ O3 values for location X are 30 ppb. If CMAQ accurately
captures the relationship between locations A, B, and X, then we would expect the O3 value to be twice as
high at X, compared to A and B. That is we would expect a value closer to 52 ppb - the weighted-
distance average of 2*32 ppb and 2*20 ppb. The eVNA technique formalizes this technique over large
and more complicated sets of data.
Unlike VNA, which averages the "raw" monitor predictions, eVNA first "adjusts" the individual monitor
predictions, multiplying by an adjustment factor that reflects the relationship between the neighbor's and
the hypothetical monitor location's O3 levels, as determined by modeling data. For example, if the
modeling data suggested that O3 levels at neighbor A were generally twice as high as at location X, and
that O3 levels at neighbor B were generally half as high as at location X, we would multiply neighbor A's
O3 value by 0.5 and multiply neighbor B's O3 value by 2 before proceeding to take a distance weighted
average over the two neighbors.
3.2.3 Four Types of Condition-Specific Adjustment Factors
The eVNA approach, as we have described it thus far, is imperfect in that it assumes the O3-level
relationship between two locations to be constant throughout the year. In fact, the relationship may vary
with the season, or with the time of day, or with numerous other factors. To take this into account, we
have added an additional layer of complexity. Rather than condensing a year's worth of model data into a
single relationship between the O3 levels of two locations (and thus a single adjustment factor for each
neighbor-"grid cell center" pair), we determine the relationship for a number of different conditions. This
allows us to tailor our adjustments to the conditions at hand; if we are adjusting a monitor value in
January, we can use an adjustment factor that specifically reflects the modeled relationship between the
locations at hand during the month of January. Similarly, if we are adjusting an O3 value that is
particularly high, we can use an adjustment factor that describes the general modeled relationship
between the locations in question when O3 levels are high.
One can imagine many such ways to divide the data into subsets that reflect the particular conditions of
the data in question. We have chosen four such divisions, herein referred to as "conditions", which we
outline below. Each condition represents a separate and distinct effort to generate an O3 surface; i.e. each
of these four conditions can each be applied separately to the data to yield a different set of O3 predictions.
Month-Decile
We first sort CMAQ modeled hourly values into 12 groups by month. In each month-group we split
evenly the ordered hourly values into ten rank-ordered deciles. This gives us 120 groups of hourly O3
values (for every CMAQ grid cell). We calculate the average of the hourly values in each gridcell-month-
decile combination, and use this average as the "representative value" of that CMAQ grid cell for that
month-decile. In order to adjust a neighboring monitor value to reflect the modeled relationship to the
unmonitored or dropout site (a "dropout site" is monitored location for which we compute predictions
based on neighboring monitors so as to compare predicted data to actual data), the appropriate month-
decile adjustment factor must be used. For example, to adjust a monitor value that falls into the 10th decile
of January monitor values, we multiply by the ratio of [the representative value of the dropout's gridcell
for the 10th decile of January] over [the representative value of the neighbor's gridcell for the 10th decile of
January].
adjusted monitor value =
representativeCMAQd gndceUimontMecUe
monitor _ valuemhhor * - eq. (2)
representatrveCMAQ r gndcellmonth,decile
Abt Associates Inc Chapter 3-2
-------
Season-Decile
We first sort CMAQ modeled hourly values into four groups by season (Jan-Mar, Apr-Jun, July-Sep, Oct-
Dec). In each season-group we split evenly the ordered hourly values into ten rank-ordered deciles. This
gives us 40 groups of hourly O3 values (for every CMAQ grid cell). We calculate the average of the
hourly values in each gridcell-season-decile combination, and use this average as the representative value
of that CMAQ gridcell for that season-decile. In order to adjust a neighboring monitor value to reflect the
modeled relationship to the unmonitored or dropout site, the appropriate season-decile adjustment factor
must be used. For example, to adjust a monitor value that falls into the 10th decile of the Jan-Mar monitor
values, we multiply by the ratio of [the representative value of the dropout's modeled O3 data for the 10th
decile of Jan-Mar] over [the representative value of the neighbor's modeled O3 data for the 10th decile of
Jan-March].
Month-Hour
We first sort CMAQ modeled hourly values into 12 groups by month. In each month-group we split
evenly the ordered hourly values into 24 groups by time of day. This gives us 288 groups of hourly O3
values (for every CMAQ grid cell). We calculate the average of the hourly values in each gridcell-month-
hour combination, and use this average as the representative value of that CMAQ grid cell for that month-
hour. In order to adjust a neighboring monitor value to reflect the modeled relationship to the
unmonitored or dropout site, the appropriate month-hour adjustment factor must be used. For example, to
adjust a monitor value from 9am in the month of January, we multiply by the ratio of [the representative
value of the dropout's gridcell for the 9am hour in January] over [the representative value of the
neighbor's gridcell for the 9am hour in January].
Season-Hour
We first sort CMAQ modeled hourly values into four groups by season. In each season-group we split
evenly the ordered hourly values into 24 groups by time of day. This gives us 96 groups of hourly O3
values (for every CMAQ grid cell). We calculate the average of the hourly values in each gridcell-season-
hour combination, and use this average as the representative value of that CMAQ grid cell for that month-
hour. In order to adjust a neighboring monitor value to reflect the modeled relationship to the
unmonitored or dropout site, the appropriate season-hour adjustment factor must be used. For example, to
adjust a monitor value from 9am in the Jan-Mar season, we multiply by the ratio of [the representative
value of the dropout's gridcell for the 9am hour in the Jan-Mar season] over [the representative value of
the neighbor's gridcell for the 9am hour in the Jan-Mar season].
3.2.4 Interpolating Hourly Data vs. Metrics
So far we have been speaking only of interpolating hourly O3 values from neighbor sites to a dropout
location. In principle, the exact same techniques can be used to interpolate daily (or even annual) metrics
from neighbors to dropout.
For example, suppose we had a day's worth of hourly O3 values for monitor A and monitor B. We wanted
to predict the daily SUM06 value for location X, situated at the midpoint between monitors A and B. We
have two options. We can use eVNA to generate hourly O3 predictions for location X, then calculate the
daily SUM06 from these hourly predictions. Alternately, we can calculate daily SUM06 at each of the
neighbor sites, and then interpolate daily SUM06 values using eVNA.
We examine both of these methods. For each of the four eVNA conditions outlined above, we generate a
set of predictions based on interpolating hourly O3 values, and a set of predictions based on interpolating
daily metrics.
To interpolate daily metrics, we class hourly data according to some condition (month-decile, month-
hour, season-decile, season-hour). As with hourly-techniques, we adjust neighboring monitor values at the
hourly level (scale by a ratio of representative CMAQ values). However, before taking a distance-
Abt Associates Inc Chapter 3-3
-------
weighted average over the set of neighbors, we compute daily metrics (SUM06 and 8-hour maximum
average) from the adjusted hourly neighbor data. These metrics are then distance-weight averaged to
produce daily metric predictions at the dropout site.
3.2.5 Summary of Approaches
The variations listed above make up the following 10 interpolation approaches for consideration:
Table 3-1: Interpolation Approaches Considered
Name
l.Hour-VNA
2. Metric- VNA
3. Hour-Month-Decile
4. Metric-Month-Decile
5. Hour-Month-Hour
6. Metric-Month-Hour
7. Hour-Season-Decile
8. Metric-Season-Decile
9. Hour-Season-Hour
10. Metric-Season-Hour
Traits
Technique:
VNA or eVNA
VNA
eVNA
Condition?
N/A
Month-Decile
Month-Hour
Season-Decile
Season-Hour
What gets
interpolated?
Hour
Metric
Hour
Metric
Hour
Metric
Hour
Metric
Hour
Metric
For the sake of comparison, we also examine the performance of the CMAQ modeling data.
3.3 Evaluation of Interpolation Approaches
A previous investigation tested the predictive power of these ten approaches by comparing predictions for
53 monitored U.S. sites (referred to as "dropouts"; data from these sites were not used in the generation of
predictions) to the actual O3 data from those sites. These data were split into an Eastern and Western
region approximately at the -99th line of longitude1. This split was made to capture the effects of the
West's sparse monitor coverage. The following results were generated:
Table 3-2: Results from Previous Investigation: Eastern U.S. (41 locations)
Adjustment
Method
model-predictions
VNA (no adjustment)
VNA (no adjustment)
month-decile
month-hour
season-decile
season-hour
month-decile
month-hour
season-decile
What gets
interpolated
Hour
Metric
Hour
Hour
Hour
Hour
Metric
Metric
Metric
SUM06
Bias
37.59
4.18
116.69
-4.67
10.31
-5.09
14.33
108.47
115.06
109.30
SUM06
Error
69.22
32.61
118.89
27.60
36.31
27.02
39.05
111.35
117.02
111.98
8-Hour
Bias
-6.72
-2.42
-1.37
-3.09
-0.64
-3.14
-0.70
-1.96
1.33
-1.86
8-Hour
Error
9.26
6.23
5.86
6.09
6.17
5.92
6.30
5.80
7.02
5.73
The Eastern region ends at the edge of the 12km x 12km CMAQ grid. The Western region begins at longitude -
99.503. Since the 12km x 12km CMAQ grid is not exactly parallel to lines of longitude, there is a small wedge
of overlap between these two regions. The entire continental US is covered either by the Western grid, by the
Eastern grid, or in a few cases by both.
Abt Associates Inc
Chapter
3-4
-------
season-hour
Metric
115.11
117.08
1.18
6.93
Table 3-3: Results from Previous Investigation: Western U.S. (12 locations)
Adjustment
Method
model-predictions
VNA (no adjustments)
VNA (no adjustments)
month-decile
month-hour
season-decile
season-hour
month-decile
month-hour
season-decile
season-hour
What gets
interpolated
Hour
Metric
Hour
Hour
Hour
Hour
Metric
Metric
Metric
Metric
SUM06
Bias
143.03
-10.91
203.76
-18.97
-17.11
-19.50
-15.53
163.44
161.56
154.73
160.90
SUM06
Error
149.98
83.23
203.76
73.49
73.40
71.62
71.64
163.44
164.58
155.54
162.65
8 -hour
Bias
22.23
3.00
4.77
1.95
4.50
1.69
3.83
3.92
7.87
3.64
7.44
8 -hour
Error
22.81
12.68
12.65
11.78
13.47
12.03
13.10
13.03
13.98
12.49
13.86
In these tables, bias and error refer to normalized mean bias and normalized mean error, which are
defined as follows:
r A A- .^^^
normalized mean bias = average . d ,(100*
actualSUM06i
) eq. (3)
SUM 06 normalized mean error = averageieduts (100
^ \predictedSUM06, -actiialSUM06J
actualSUM06,
)eq.(4)
and likewise for the 8-hour statistic. The greater the error, the less accurate the approach is on average.
The bias indicates whether there is a tendency to overpredict or underpredict and if so by how much. A
negative bias indicates underprediction and a positive bias indicates overprediction.
The full memorandum describing this previous work was delivered to U.S. EPA by Abt Associates on
April 12th 2006 and can be found in Appendix K. Based on these results, we chose to examine the hour-
month-decile and hour-month-hour approaches in more detail.
As in the previous investigation, we divided the country into an Eastern region and a Western region. This
time, however, we examined the performance of these approaches at predicting the values of all of the O3
monitors within the region in question. We also analyzed performance in terms of the W126 metric,
which had not been examined in the previous investigation.
In addition to evaluating the hour-month-decile and hour-month-hour approaches, we evaluated several
VNA - eVNA blends. These approaches adjust monitor values according to the eVNA technique for
distant neighbors but leave nearby neighbor values unadjusted. The technique is based on the observation
that VNA outperforms eVNA at close range, but eVNA outperforms VNA at longer ranges (see Appendix
K for full details of the investigation).
Abt Associates Inc
Chapter
3-5
-------
The following results were generated"
Table 3-4: Results from Current Investigation: Eastern US: 8-hour Maximum (approx 800
locations)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
790
Mean
-7.22
8.75
-8.0%
10.2%
-2.53
5.05
-2.7%
6.0%
-2.88
4.91
-3.1%
5.8%
-2.82
4.88
-3.0%
5.8%
-1.58
5.46
-1.5%
6.6%
-2.79
4.89
-3.0%
5.8%
-2.57
4.88
-2.7%
5.8%
-2.91
4.95
-3.1%
5.9%
Median
-7.19
7.79
-8.9%
9.6%
-2.71
3.96
-3.2%
4.7%
-2.90
4.04
-3.5%
4.8%
-2.88
4.00
-3.5%
4.8%
-1.78
3.93
-2.1%
4.8%
-2.85
4.04
-3.5%
4.8%
-2.73
4.01
-3.3%
4.8%
-3.00
4.02
-3.6%
4.8%
Min
-37.61
0.08
-34.0%
0.1%
-25.80
0.01
-27.5%
0.0%
-26.82
0.03
-24.8%
0.0%
-26.82
0.01
-24.8%
0.0%
-29.34
0.00
-34.7%
0.0%
-26.72
0.02
-24.7%
0.0%
-26.72
0.00
-25.9%
0.0%
-18.32
0.04
-20.2%
0.0%
Max
19.09
37.61
36.9%
36.9%
32.77
32.77
52.6%
52.6%
28.09
28.09
47.8%
47.8%
28.09
28.09
49.5%
49.5%
58.13
58.13
69.3%
69.3%
28.09
28.09
47.3%
47.3%
28.09
28.09
47.7%
47.7%
28.09
28.09
47.3%
47.3%
Range
45.83
37.54
66.4%
36.8%
54.96
32.75
77.1%
52.6%
46.41
28.05
67.9%
47.7%
46.41
28.05
69.6%
49.4%
81.22
58.13
94.7%
69.3%
46.41
28.05
67.5%
47.3%
46.41
28.09
67.9%
47.7%
46.41
28.05
67.4%
47.2%
STD
7.74
5.96
9.0%
6.4%
6.19
4.39
7.6%
5.3%
5.62
3.97
6.8%
4.7%
5.67
4.03
6.9%
4.8%
7.54
5.44
9.2%
6.6%
5.66
4.00
6.9%
4.8%
5.84
4.11
7.1%
4.9%
5.71
4.07
6.9%
4.8%
1 These results are slightly different than in previous drafts of this report. This is due to the removal of "twin
monitors" - see footnote j on page 2-1.
Abt Associates Inc
Chapter
3-6
-------
Table 3-5: Results from Current Investigation: Eastern US: SUM06 (approx 800 locations)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
Mean
-1.70
5.66
15.7%
45.4%
-1.46
4.61
1.2%
28.6%
-1.85
4.14
-1.3%
25.2%
-1.82
4.21
-1.1%
25.7%
-0.75
4.84
5.9%
31.2%
-1.76
4.12
-0.7%
25.1%
-1.58
4.23
0.6%
26.2%
-1.83
4.09
0.0%
25.8%
Median
-1.99
4.83
-9.6%
23.4%
-1.40
3.68
-7.4%
19.0%
-1.66
3.21
-8.6%
17.7%
-1.64
3.42
-8.8%
17.6%
-0.86
3.66
-4.1%
18.7%
-1.58
3.23
-8.4%
17.3%
-1.50
3.45
-6.9%
17.2%
-1.69
3.21
-8.4%
16.9%
Min
-26.70
0.00
-90.5%
0.0%
-24.00
0.00
-92.2%
0.0%
-22.58
0.00
-92.2%
0.0%
-22.58
0.00
-92.2%
0.0%
-22.22
0.04
-93.2%
0.2%
-20.91
0.00
-92.0%
0.0%
-20.91
0.00
-92.0%
0.0%
-18.60
0.00
-73.7%
0.0%
Max
26.18
26.70
4396.0%
4396.0%
23.56
24.00
1199.7%
1199.7%
23.05
23.05
1252.8%
1252.8%
23.05
23.05
1267.9%
1267.9%
25.53
25.53
1234.3%
1234.3%
23.05
23.05
1247.2%
1247.2%
23.05
23.05
1257.8%
1257.8%
23.05
23.05
1243.5%
1243.5%
Range
48.90
26.70
4486.5%
4396.0%
45.50
23.99
1277.0%
1199.7%
40.27
23.05
1325.2%
1252.8%
41.36
23.05
1343.1%
1267.8%
47.59
25.49
1312.1%
1234.1%
40.66
23.05
1319.5%
1247.2%
41.73
23.05
1334.5%
1257.7%
41.65
23.05
1317.2%
1243.5%
STD
6.99
4.44
193.8%
189.1%
5.91
3.97
63.0%
56.2%
5.09
3.49
59.7%
54.1%
5.20
3.55
60.8%
55.2%
6.42
4.28
68.6%
61.4%
5.09
3.47
59.9%
54.4%
5.28
3.53
62.5%
56.7%
5.05
3.48
66.9%
61.7%
Abt Associates Inc
Chapter
3-7
-------
Table 3-6: Results from Current Investigation: Eastern US: W126 (approx 800 locations)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
786
Mean
-1.25
4.07
4.6%
31.8%
-0.84
3.40
1.3%
25.3%
-1.19
3.00
-1.4%
21.8%
-1.14
3.07
-1.1%
22.3%
-0.32
3.61
5.2%
27.5%
-1.14
3.00
-1.0%
21.8%
-0.99
3.08
0.1%
22.7%
-1.21
2.99
-1.1%
21.9%
Median
-1.28
3.43
-8.2%
21.1%
-0.87
2.68
-5.6%
18.3%
-1.13
2.34
-7.0%
16.3%
-1.04
2.48
-6.6%
16.3%
-0.45
2.69
-2.5%
18.4%
-1.05
2.32
-6.6%
16.0%
-0.90
2.44
-5.4%
16.0%
-1.05
2.30
-6.8%
16.1%
Min
-17.11
0.01
-81.7%
0.0%
-18.09
0.00
-78.1%
0.0%
-14.58
0.01
-65.7%
0.0%
-14.58
0.01
-72.2%
0.1%
-18.16
0.00
-83.3%
0.0%
-14.58
0.00
-65.7%
0.0%
-14.58
0.00
-75.7%
0.0%
-14.58
0.01
-62.9%
0.1%
Max
19.83
19.83
646.0%
646.0%
16.69
18.09
521.9%
521.9%
18.33
18.33
547.0%
547.0%
18.33
18.33
558.1%
558.1%
19.94
19.94
541.8%
541.8%
18.33
18.33
543.8%
543.8%
18.33
18.33
548.5%
548.5%
18.33
18.33
544.0%
544.0%
Range
36.05
19.81
727.7%
646.0%
34.78
18.09
600.0%
521.9%
32.91
18.32
609.9%
546.9%
32.91
18.32
622.0%
558.1%
38.09
19.93
620.2%
541.8%
32.91
18.32
606.7%
543.7%
32.91
18.32
613.9%
548.4%
32.91
18.32
607.0%
544.0%
STD
5.03
3.21
62.5%
54.0%
4.35
2.84
43.3%
35.1%
3.75
2.54
38.7%
32.1%
3.82
2.55
39.5%
32.6%
4.81
3.19
47.5%
39.1%
3.76
2.54
39.0%
32.3%
3.89
2.56
40.2%
33.2%
3.75
2.56
40.6%
34.2%
Abt Associates Inc
Chapter
3-8
-------
Table 3-7: Results from Current Investigation: Western US: 8-hour Maximum (approx. 300
locations)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
287
Mean
4.63
11.27
9.2%
15.8%
-2.89
6.38
-3.0%
8.4%
-3.19
6.33
-3.1%
8.3%
-3.16
6.29
-3.1%
8.2%
-1.98
6.34
-1.8%
8.4%
-3.08
6.30
-3.0%
8.2%
-2.84
6.26
-2.7%
8.2%
-3.16
6.37
-3.0%
8.3%
Median
5.94
9.94
8.0%
13.5%
-2.59
4.81
-3.5%
7.1%
-2.66
4.99
-3.7%
7.0%
-2.83
5.08
-3.9%
6.9%
-2.14
4.85
-2.8%
6.7%
-2.63
4.85
-3.5%
6.7%
-2.75
5.05
-3.6%
6.9%
-2.63
5.17
-3.6%
7.1%
Min
-45.13
0.05
-37.6%
0.0%
-41.10
0.03
-34.2%
0.0%
-25.34
0.01
-32.5%
0.0%
-24.81
0.04
-32.5%
0.1%
-40.01
0.00
-33.3%
0.0%
-24.64
0.02
-29.8%
0.0%
-24.60
0.00
-29.8%
0.0%
-26.55
0.02
-27.6%
0.0%
Max
36.29
45.13
85.0%
85.0%
21.76
41.10
53.8%
53.8%
19.65
25.34
51.5%
51.5%
20.05
24.81
53.8%
53.8%
24.83
40.01
53.4%
53.4%
19.33
24.64
51.9%
51.9%
19.89
24.60
53.4%
53.4%
22.65
26.55
60.8%
60.8%
Range
81.41
45.08
122.6%
85.0%
62.86
41.07
88.1%
53.8%
44.99
25.33
84.0%
51.5%
44.86
24.77
86.3%
53.7%
64.83
40.00
86.7%
53.4%
43.96
24.62
81.7%
51.9%
44.48
24.60
83.2%
53.4%
49.20
26.53
88.4%
60.8%
STD
13.48
8.72
18.7%
13.6%
7.90
5.48
10.6%
7.1%
7.62
5.31
10.4%
7.0%
7.51
5.17
10.2%
6.8%
8.11
5.43
10.9%
7.1%
7.59
5.24
10.2%
6.8%
7.54
5.08
10.2%
6.6%
7.67
5.31
10.5%
7.0%
Abt Associates Inc
Chapter
3-9
-------
Table 3-8: Results from Current Investigation: Western US: SUM06 (approx. 300 locations)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count
286
286
279
279
286
286
279
279
286
286
279
279
286
286
279
279
286
286
279
279
286
286
279
279
286
286
279
279
286
286
279
279
Mean
3.77
10.90
298.6%
317.1%
-1.93
6.76
15.6%
63.4%
-2.96
6.84
13.8%
61.2%
-2.69
6.68
14.0%
60.7%
-1.77
6.76
18.3%
65.0%
-2.90
6.82
15.2%
62.4%
-2.62
6.68
16.0%
62.1%
-2.93
6.82
18.0%
63.7%
Median
3.98
9.01
32.1%
43.3%
-1.12
4.50
-10.9%
36.6%
-1.53
4.80
-14.8%
31.7%
-1.37
4.56
-13.2%
31.5%
-0.89
4.37
-8.9%
33.4%
-1.42
4.94
-14.5%
32.8%
-1.23
4.55
-12.2%
29.9%
-1.47
5.06
-13.9%
32.8%
Min
-44.93
0.00
-100.0%
0.0%
-32.15
0.00
-100.0%
0.2%
-33.75
0.00
-100.0%
0.1%
-31.81
0.00
-100.0%
0.1%
-30.79
0.00
-100.0%
0.2%
-33.81
0.00
-100.0%
0.0%
-32.27
0.00
-100.0%
0.1%
-34.39
0.00
-100.0%
0.1%
Max
41.02
44.93
29019.0%
29019.0%
44.64
44.64
3246.8%
3246.8%
41.26
41.26
2531.7%
2531.7%
44.64
44.64
2978.2%
2978.2%
40.58
40.58
3211.1%
3211.1%
40.15
40.15
2752.4%
2752.4%
39.65
39.65
3376.8%
3376.8%
41.06
41.06
2619.1%
2619.1%
Range
85.95
44.93
29119.0%
29019.0%
76.80
44.64
3346.8%
3246.6%
75.01
41.26
2631.7%
2531.7%
76.45
44.64
3078.2%
2978.1%
71.37
40.58
3311.1%
3210.9%
73.97
40.15
2852.4%
2752.4%
71.91
39.65
3476.8%
3376.8%
75.45
41.06
2719.1%
2619.1%
STD
13.81
9.28
1852.5%
1849.4%
9.76
7.30
210.8%
201.7%
9.62
7.38
178.9%
168.7%
9.54
7.32
196.6%
187.5%
9.72
7.21
211.3%
201.9%
9.50
7.22
191.8%
182.0%
9.39
7.10
219.1%
210.7%
9.53
7.27
189.6%
179.5%
Abt Associates Inc
Chapter
3-10
-------
Table 3-9: Results from Current Investigation: Western US: W126 (approx. 300 locations)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
286
Mean
2.53
8.18
112.9%
128.1%
-1.08
4.68
15.8%
48.6%
-1.76
4.77
13.9%
47.3%
-1.59
4.70
15.2%
48.3%
-0.84
4.75
16.8%
48.6%
-1.72
4.79
14.4%
47.7%
-1.50
4.75
14.9%
47.7%
-1.77
4.84
18.8%
52.5%
Median
3.50
6.67
38.3%
49.3%
-0.64
2.86
-5.0%
26.0%
-0.68
3.14
-8.3%
25.7%
-0.72
2.90
-8.5%
25.6%
-0.47
2.91
-3.9%
24.8%
-0.68
3.23
-7.1%
25.7%
-0.71
3.09
-7.7%
25.4%
-0.79
3.07
-7.8%
27.7%
Min
-41.98
0.03
-68.4%
0.2%
-26.39
0.02
-89.3%
0.2%
-29.36
0.01
-89.3%
0.0%
-28.32
0.01
-89.3%
0.0%
-27.12
0.00
-90.2%
0.1%
-29.59
0.00
-90.2%
0.0%
-28.59
0.00
-90.2%
0.0%
-29.80
0.00
-76.2%
0.0%
Max
26.57
41.98
5993.6%
5993.6%
27.84
27.84
2810.1%
2810.1%
28.41
29.36
2401.1%
2401.1%
28.80
28.80
2810.1%
2810.1%
28.57
28.57
2641.7%
2641.7%
28.07
29.59
2548.5%
2548.5%
28.22
28.59
2641.7%
2641.7%
28.58
29.80
3636.8%
3636.8%
Range
68.54
41.95
6062.0%
5993.3%
54.23
27.82
2899.4%
2809.9%
57.77
29.34
2490.4%
2401.0%
57.12
28.79
2899.4%
2810.1%
55.69
28.57
2731.9%
2641.6%
57.66
29.58
2638.7%
2548.5%
56.82
28.59
2731.9%
2641.7%
58.37
29.80
3713.0%
3636.7%
STD
10.66
7.28
389.2%
384.5%
6.95
5.25
176.2%
170.1%
6.95
5.34
154.1%
147.3%
6.88
5.26
176.3%
170.2%
7.07
5.29
167.0%
160.7%
6.93
5.30
161.8%
155.3%
6.86
5.18
166.6%
160.3%
7.01
5.38
223.0%
217.6%
Here bias, normalized bias, and normalized error are defined as follows: For monitor i with actual value a;
and predicted value Q:
Bias; = c;- ^
Normalized biasj = biasj / a;
Normalized error; = | biaSj | / a;
Bias, normalized bias, and normalized error are computed for each monitor, and the count, mean, median,
max, min, range, and STD of these three values over the relevant set of monitors (East or West) is shown
above. (Note: In the previous investigation, p 3-4 - 3-5, only mean normalized bias and mean normalized
error were shown)
Appendix A presents more detailed results for the West. Based on our results we offer the following
observations:
Compared to other approaches examined, the CMAQ model data is a poor predictor of O3 levels.
Abt Associates Inc
Chapter
3-11
-------
Overall, VNA and eVNA perform similarly, though in certain circumstances one will outperform
the other.
Blends of VNA and eVNA techniques can produce better results than pure applications of either
VNA or eVNA, especially in the Western U.S. (where monitor coverage is sparser).
3.4 Generation of the National Potential O$ Exposure Surface
Based on the empirical strengths of these techniques, as well as logistical and methodological concerns,
we chose the following approach to generate the POES:
For monitors in the Eastern U.S., the VNA interpolation was used. Though several VNA-eVNA
blends matched VNA's performance, none offered significantly more predictive power.
Ultimately, the largest factor separating VNA from VNA-eVNA blends was VNA's simplicity,
and the fact that it CMAQ data wasn't required to perform the interpolation, thus leaving open the
possibility of adding other monitor years.
For monitors in the Western U.S., we used a blend of eVNA and VNA techniques: For neighbors
less than 50 km from the site of interpolation, VNA techniques were used. For neighbors of
distance greater than 50 km, the Hour-month-hour eVNA approach was used. In the West, blends
proved more successful than either pure VNA or eVNA techniques. The 50km threshold was
chosen for consistency with previous studies that used a 50km threshold".
Using the chosen approach in each region, we interpolated O3 values for each gridcell center in the
"composite CMAQ grid" (This does not necessarily mean that we used data from the CMAQ model -
indeed in the East we did not. This simply means that we used the same coordinate system that the
CMAQ model uses - see section 3.1 for more detail). In the West, our predictions were made entirely on
the 36km grid, (the only grid which covers the Western U.S.). The East was primarily covered by the
12km CMAQ grid, but relied on the 36 km grid point for the parts of Maine, Minnesota, Florida and
Texas that were not covered by the 12km grid. The resulting surface gives hourly O3 values for 44432
regularly spaced locations throughout the U.S.
A description of the resulting O3 data will be postponed until the following section, where it can be
discussed with a similar description of four alternate O3 surfaces that correspond to four alternative
standards currently being considered.
n For example: U.S. EPA. 1999. The Benefits and Costs of the Clean Air Act: 1990 to 2010: EPA Report to
Congress. U.S. EPA, Office of Air and Radiation, Office of Policy. Washington, DC. EPA 410-R-99-001.
Abt Associates Inc Chapter 3-12
-------
4. Rollback
To estimate the crop yield and tree seedling biomass changes due to meeting a hypothetical standard, we
must first generate a hypothetical 2001 O3 surface that replicates O3 levels just meeting the standard. This
procedure is referred to as a "rollback".
4.1 Scenarios
Hypothetical O3 surfaces were generated for the following six hypothetical scenarios:
Ozone levels which just meet an 84 ppb 4th highest 8-hour maximum standard (in this case, we
performed a rollback to 84.999 ppb, since the 4th highest 8-hour maximum standard truncates
decimals)
Ozone levels which just meet a 70 ppb 4th highest 8-hour maximum standard (actually 70.999 for
the reasons above)
Ozone levels which just meet a 25ppm-h 3-month 12-hour SUM06 standard
Ozone levels which just meet a 15ppm-h 3-month 12-hour SUM06 standard.
Ozone levels which just meet a 21ppm-h 3-month 12-hour W126 standard.
Ozone levels which just meet a 13ppm-h 3-month 12-hour W126 standard.
We shall refer to these respective scenarios as "the 84 ppb rollback", "the 70 ppb rollback", "the 25ppm-h
rollback", the "15ppm-h rollback", the "21ppm-h rollback", and the "13ppm-h rollback". Together with
the "as is" scenario (i.e. the POES generated in Section 3), these comprise the five scenarios for which we
evaluate crop yield and tree seedling biomass loss, and the associated economic benefits.
4.2 Rollback Methodologies
The quadratic rollback reduces all hourly O3 values at a given location {cl \
-------
where:
c; is the O3 value at hour /' and
S is the standard to be met.
In other words, we want to choose the v and b that will cause the new 8-hour-maximum period to have an
average value equal to the standard. There generally exists more than one choice of v and b that will
satisfy this condition. We follow the approach used previously by EPA to select a v and b which satisfy
the above equation; namely, v=l if and only if such a choice yields a strictly monotonic rollback function.
Otherwise, v is chosen such that the domain of the function (hourly O3 values) is identical to the
maximum domain over which the rollback is strictly monotonic.
For a full account of the computations underlying this rollback, see the relevant Memo in Appendix B.
4.2.2 SUM06 Rollback
The SUM06 rollback is achieved through iterative 8-hour maximum rollbacks: We choose an "8-hour
maximum target" and perform a rollback to meet this "target". We evaluate the SUM06 value of the
resulting data; if it is below the standard we wish to be meeting, we increase the "8-hour maximum
target"; if the SUM06 is above the standard we wish to be meeting, we decrease the "8-hour maximum
target." We continue performing rollbacks and adjusting target values until the SUM06 value has
sufficiently (see below) approached the standard we wish to be meeting.
As the 8-hour maximum target is incrementally increased, SUM06 will occasionally exhibit discrete
jumps of magnitude. This occurs when an hourly value that was previously slightly less than .06 ppm
(and thus contributed nothing to the SUM06 total) was increased to .06 ppm or greater (and thus
contributes its full value). Because of these irregular jumps, one cannot always find an "8-hour maximum
target" which yields a SUM06 exactly equal to the standard.
In performing the SUM06 rollbacks, we terminated the iterative process once the SUM06 value fell below
and within .06 ppm-h of the standard (we will refer to this as "the approach condition"). For the purposes
of this study, we considered the effect of a .06ppm-h error in SUM06 to be negligible. Additionally,
iteration was terminated if the approach condition was not met after 25 iterations. The vast majority of
locations met the approach condition before reaching 25 iterations0.
4.2.3 W126 Rollback
The W126 rollback is performed in a similar fashion to the SUM06 rollback. We perform successive
rollbacks to meet 8-hour maximum targets, varying the target such that the W126 value of the ensuing O3
surface approaches the W126 standard we wish to meet.
The W126 metric is continuous with respect to variation in the target 8-hour maximum value - unlike the
SUM06 metric, there are no discrete jumps in W126 value and so there is no theoretical limit to how
closely one can approach the standard through continued iteration. For practical purposes, we continued
iteration until a surface was generated whose W126 value fell within .005ppm-h of the standard.
0 In the 25ppm-h rollback, only 3 out of over 40,000 locations did not meet the approach condition.
All results still fell within the range of 24.88-25.03 ppm-hr. In the 15ppm-h rollback, only 4 out of over
40,000 locations did not meet the approach condition. All results fell within the range of 14.90-15.06
ppm-hr.
Abt Associates Inc Chapter 4-2
-------
4.3 Aggregate Results
For each rollback scenario we compute maximum 3-month SUM06 values, 4th highest maximum 8-hour
average values, and maximum 3-month W126 values. The tables below give mean, median, maximum,
and minimum O3 levels in terms of these three O3 metrics.
Table 4-1 Rollback Scenarios Described in Terms of Mean, Median, Maximum, and Minimum 8-Hour
Maximum Values
8 -hour
as is
8-hour
rollback 84
Mean
Median
Max
Min
73.01635703
73.62485235
118.3070998
39.83235804
72.63478249
73.
84.
39.
.62485235
.99999924
.83235804
8-hour
rollback 70
68.
70.
70.
39.
.20098977
.99999904
.99999924
.83235804
SUM06
rollback 25
72
73
.78260827
.54231589
105.1970365
39
.83235804
SUM06
rollback 15
70.7770449
71.57599665
100.6138143
39.83235804
W126
rollback 21
72.91678111
73.61090821
105.1970365
39.83235804
W126
rollback 13
71.40617229
72.55692302
101.2226691
39.83235804
Table 4-2 Rollback Scenarios Described in Terms of Mean, Median, Maximum, and Minimum SUM06
Values
SUM06
as is
8-hour
rollback 84
Mean
Median
Max
Min
12.
11.
.50749036
.83841584
76.03335358
0
12.
11.
.26495298
.75921953
59.0338992
0
8-hour
rollback 70
8.719191633
8.539042804
35.79732406
0
SUM06
rollback 25
12
11
25
.24571169
.83841584
.02971601
0
SUM06
rollback 15
10.23400932
11.83841584
15.052334
0
W126
rollback 21
12.40493551
11.83841584
33.97458015
0
W126
rollback 13
10.8833816
11.
19.
.80423751
.54925481
0
Table 4-3 Rollback Scenarios Described in Terms of Mean, Median, Maximum, and Minimum W126
Values
W126
Mean
Median
Max
Min
as is
10.5704409
9.93151599
61.8103728
0.28970418
8-hour
rollback 84
10.3593442
9.88411229
40.9327805
0.28970418
8-hour
rollback 70
7.80108448
7.56656303
22.5694254
0.28970418
SUM06
rollback 25
10.3685444
9.93151599
21.2697139
0.28970418
SUM06
rollback 15
8.97140437
9.93151599
14.9021518
0.28970418
W126
rollback 21
10.4861036
9.93151599
20.9999942
0.28970418
W126
rollback 13
9.39789481
9.93151599
12.9999982
0.28970418
4.4 Air Quality Maps For Alternative Ozone Standards
In Appendix D, we present O3 air quality maps under the six rollback scenarios described above, and
under "as is" conditions as described by the POES. Ozone levels are shows in terms of the Maximum 3-
month 12-hour W126 metric. The maps show the entire continental U.S. as a whole, even though the data
for the East and Western U.S. were generated separately according to slightly different interpolation
methods.
Abt Associates Inc
Chapter
4-3
-------
-------
5. Crop and Tree Exposure
A direct consequence of elevated O3 levels is a reduction in agricultural output, which translates into
higher commodity prices, and a loss in economic welfare. To assess the benefits of attaining alternate air
quality standards, we need to quantify the relationship between air quality and crop yield, and then
incorporate the resulting yield effects into an agricultural economic model. This section documents the
methodology and data used to quantify the impact of reaching the alternate air quality standards on crop
yield and tree seedling biomass. In the next section we will present the resulting economic benefits.
Fifteen commodities, including eight field crops and seven fruits and vegetables, as well as ten tree
species were retained in the analysis'3. Based on USDA National Agricultural Statistics Service (NASS)
estimates, these fifteen commodities accounted for a yearly market value of $53 billion in 2001.
Field crops (8 crops) Fruits and vegetables (1 crops') Tree Species (10 species')
Cotton • Cantaloupes • Quaking Aspen
Field Corn • Grapes • Black Cherry
Grain Sorghum • Kidney Bean • Douglas Fir
Peanuts • Lettuce • Eastern White Pine
Potatoes • Onions • Ponderosa Pine
Rice • Tomatoes Processing • Red Alder
Soybean • Valencia Oranges • Red Maple
Winter Wheat • Sugar Maple
Tulip Poplar
Virginia Pine
The relative shares of market value for the selected commodities are illustrated in Figure 5-1 below.
To estimate crop yield loss under a hypothetical O3 standard, we must first generate a hypothetical 2001
O3 surface that replicates O3 levels just meeting the standard. This procedure is referred to as a "rollback".
The various air quality scenarios considered are described in Chapter 4 above.
The baseline is the year 2001 O3 surface. The 2001 surface and each rollback scenario were reported
using both monthly SUM06 and monthly W126 metrics for field crops (including potatoes, lettuce, and
kidney beans) and tree seedlings. A comparison of results between SUM06 and W126 is presented
subsequently. For other fruits and vegetables, we used monthly 7-hour-average and monthly 12-hour-
average metrics to calculate yield changes, as SUM06 and W126 functions were not available.
p Barley and Lemon were removed from the analysis for lack of consistent SUM06 and W126 concentration-
response functions.
Abt Associates Inc Chapter 5-1
-------
2%
3%
4%
6%
23%
• Corn
El Soybeans
Largest Share
H Wheat Winter
0 Cotton
U Grapes
S Lettuce
EH Oranges
El Peanuts
ffl Sorghum
DRice
D Onions j
M Tomatoes i
I
D Cantaloups j
D Dry Beans Smallest Share
Figure 5-1: Shares of Value of Production for Selected Crops, Fruits, and Vegetables (2001, total
value is $53 billion). Source: NASS"
The steps involved in generating yield and biomass changes are summarized below.
Step 1. Collect O3 concentration-response (C-R) functions for all crops, fruits, vegetables, and tree
seedlings and adjust C-R parameters to the appropriate air quality metric (SUM06, W126,
12-hour-average, and 7-hour-average depending on the C-R function). C-R functions and
parameters were derived from previous studies by Olszyk and Thompson (1988), Lee and
Hogsett (1996), and Abt Associates (1995).
Step 2. Identify areas of the country where each crop is cultivated. This requires getting growing
range boundaries for all crops, fruits, and vegetable. We relied on NASS 2001 Crop County
Data for all major field crops, and on the 2002 Census of Agriculture for field crops, fruits
and vegetables.
Step 3. Identify areas of the country where specific tree species grow. This requires getting
growing range boundaries for all tree species. We used the USGS tree species range maps
(Little, 1978) available in ArcGIS Shapefile format.
q The value of production for All Dry Edible Beans was used for kidney beans since NASS does not provide a
breakdown by bean species. Other edible bean species include navy bean, pinto bean, and black bean.
Abt Associates Inc
Chapter
5-2
-------
Step 4. Obtain usual harvest dates for all crops, fruits and vegetables to link crop growing season
and O3 exposure. The USDA reports usual planting and harvesting seasons for major crops
at the state level. Missing data was sourced from published crop profiles from the
Integrated Pest Management (IPM) centers.
Step 5. When appropriate, adjust all hourly O3 values down by 10% to account for the height
differential between monitoring stations and crop foliage, and compute monthly O3 metrics.
Yield changes and economic benefits were obtained for both reduced and unreduced
metrics. A comparison between the two approaches is available in Appendix L.
Step 6. Generate seasonal O3 indices for all commodities. Concentration-response functions are
calibrated for specific experimental exposure durations, so it was necessary to compute O3
indices over the corresponding period. Seasonal SUM06, W126, 7-hour-average, and 12-
hour-average O3 indices were derived from the monthly metrics estimated in Step 5 for all
air quality scenarios. For example, the concentration-response function for winter wheat is
calibrated for a 58-day growing season, so we identified typical harvest dates for winter
wheat by state, and calculated O3 exposure for the 58 days previous to harvest time.
Step 7. Generate relative crop yield and tree seedling biomass loss at the CMAQ grid level. The
seasonal O3 indices generated above were plugged into the C-R functions to generate
relative yield and biomass losses over each crop and tree growing range.
Step 8. Aggregate yield and biomass losses from gridcell to county to production regions (as
defined by the USDA Economic Research Service). To serve as input into the AGSIM©
model, yield losses at the gridcell level had to be combined into nine production regions
according to USDA Economic Research Service (ERS) classification. We used planted
acreages reported in NASS 2001 County Crop Data to generate weighted-average yield
changes across ERS region. We also relied on the U.S. Census Cartographic Boundaries for
an administrative map of U.S. counties.
The schematic below describes the process of combining all primary data sources and monthly O3
estimates to generate yield and biomass changes. Tables are presented with an indicative list of variables.
Abt Associates Inc Chapter 5-3
-------
USDA Usual
Planting and
Harvesting
Dates
IPM
Center
Crop
Profiles
ESRI Annual
Temperature
Zones
(2005)
Step 4
Monthly SUM06
gridcell_id
in_month
as_is
rollback_a4
rollbaek_70
rollback_Sum06_25
rollback Sum06 15
Monthly 12hr
USGS Tree
Species
Ranges
(1978)
USDA
Agricultural
Census
(2002)
USDA
County
Crop Data
(2001)
US Census
Bureau
Cartographic
Boundary
Step 2
Step 3
gridcelljd
month
as_is
rollback_34
rollback_70
rollback_Surn06_25
rollback Sum06 15
Seasonal Ozone Index
gridcelljd
cropjd
fn_type
ozindex
days
Month Iy7hr
gridcalljd
month
asjs
rollback 84
rollback_70
rollback_Sum06_
rollback_Sum06_
25
15
Tree Range I Gridcell
cropjd
gridcelljd
Crop Range I Gridcell
cropjd
gridcelljd
countyJd
statejd
ERS_region
acres_pl anted
acresJiarvBsted
# of farms
Step 6
Monthly W16
gridcelljd
month
asjs
rollback_84
rollback_7C
rollback_sumQ6_25
rollback sum06 15
Step?
t
Yield Loss / ERS Region
ERS region
cropjd
fn_type
yield
sum_weighl
Figure 5-2: Estimation of Yield and Biomass Losses - Data Sources and Data Flow
Abt Associates Inc
Chapter
5-4
-------
5.1 Ozone Concentration-Response Functions
Ozone concentration-response functions estimate the relationship between elevated O3 exposure and plant
yields. The data necessary to estimate these functions can come from tests in open fields, open-top
chambers, or econometric methods. The present crop assessment was built upon National Crop Loss
Assessment Network (NCLAN) O3 concentration-response functions. These functions come from tests in
open-top chambers, in which O3 is injected into the chamber through an inlet to replicate various O3
exposures. Typically, the experimental test data is fit to a Weibull function. Its general form is as follows
(Lesser etal., 1990):
X
ozone/
eq. (7)
Where Y is the estimated mean yield, ozone is the O3 exposure index, A the theoretical yield at zero O3
concentration, y the scale parameter for O3 exposure that reflects the dose at which the expected response
is reduced to 0.37^4, and Ithe shape parameter affecting the change in the predicted rate of loss. Because
the response of crop yield to O3 exposure in the NCLAN study varies by cultivar and experimental
location, we used estimated median C-R functions. Minimum and maximum C-R functions are also
presented in Table 5-land Table 5-2.
Relative yield losses (RYL) are defined as:
eq.(8)
base
Where Ybase is the estimated mean yield at the reference exposure index ("clean air" in the charcoal-
filtered air treatment). We rely on the original sources of the exposure-response functions for the
reference levels for different indices. The reference level is 27 ppb for 7-hour-average, 25 ppb for 12-
Hout-average, and 0 ppm/h for SUM06 and W126 (Olszyk, 1988).
Functions for field crops and tree seedlings were adjusted to a seasonal 12-Hour SUM06 and W126 O3
index, 12-hour-average index for fruits, and 7-hour-average index for rice and cantaloupes. All crop and
fruit/vegetable functions are calibrated for an experimental O3 exposure duration, typically corresponding
to the duration of each crop growing season.
A summary of all relative yield loss (RYL) functions and O3 indices is presented in tables below. Figure
5-3 and Figure 5-4 below provide a comparison of W126 median C-R function schedules for all major
crops and trees.
Abt Associates Inc Chapter 5-5
-------
Table 5-1: Composite SUM06 Relative Yield Loss Functions for Major Crops, Beans, Lettuce, and
Potatoes
Crop
Cotton
Field Corn
Winter Wheat
Soybeans
Potatoes
Grain Sorghum
Peanuts
Lettuce
Kidney Beans
RYL
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)x]
l-exp[-(ozone/y)x]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)x]
l-exp[-(ozone/y)x]
l-exp[-(ozone/y)^]
Index
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
Response
Max
Median
Min
Max
Median
Min
Max
Median
Min
Max
Median
Min
Min
Median
Max
~
y (ppm)
78
105.9
116.8
92.4
96.9
94.2
27.2
53.3
72.1
131.4
109.7
299.7
79.3
86.2
93.8
177.8
99.8
54.9
42.9
1
1
1
2
3
4
1
2
2
1
1
1
1
1
1
2
2
5
2
X
.311
.655
.523
.816
.194
.307
.000
.766
.353
.000
.567
.547
.654
.274
.000
.329
.219
.512
.537
days
119
114
119
83
83
83
58
58
58
104
93
104
62
66
70
85
112
53
57
Source: Lee and Hogsett (1996) table 10.2 for crops. Abt Associates Inc (1995) exhibit 11 for fruits and
vegetables. Note: Peanuts, Grain Sorghum, Lettuce, and Kidney Beans, only have one C-R function and
therefore do not have a Median, Max, and Min.
Table 5-2: Composite W126 Relative Yield Loss Functions for Major Crops, Beans, Lettuce, and
Potatoes
Crop
Cotton
Field Corn
Winter Wheat
Soybeans
Potatoes
Grain Sorghum
Peanuts
Lettuce
Kidney Beans
RYL
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
l-exp[-(ozone/y)^]
Source: Lee and Hogsett (1996) table 10.2 for crops
vegetables adjusted between SUM06 and W126.
Index
W126
W126
W126
W126
W126
W126
W126
W126
W126
. Abt Associates
Response
Max
Min
Median
Max
Min
Median
Median
Max
Min
Max
Min
Median
Median
Max
Min
~
y (ppm)
74.7
113.5
94.4
92.9
94.5
98.3
53.7
25.0
76.1
130.3
470.2
110.0
99.5
96.3
113.8
205.9
97.4
54.6
44.2
Inc (1995) exhibit 11
X
1.0700
1.4100
1.5720
2.5940
4.1900
2.9730
2.3910
1.0000
2.1000
1
1
1
.0000
.1283
.3670
1.2420
1.0000
1.2990
1
1
4
2
for
.9630
.9050
.9210
.3530
fruits
days
119
119
114
83
83
83
58
58
58
104
104
93
66
70
62
85
112
53
57
and
Abt Associates Inc Chapter 5-6
-------
Table 5-3: Composite 12-hour-average and 7-hour-average Relative Yield Loss Functions for Fruits
and Other Vegetables
Crop
Grapes
Tomatoes
Processing
Onions
V. Oranges
Cantaloupes
Rice
RYL
l-[y -(X*ozone)] ,
l-[y -(X*ozone)] ,
l-[y -(X*ozone)] ,
l-[y -(X*ozone)] ,
l-[y -(X*ozone)]
l-exp[-(ozone/y)x]
/ [y-(X*base!2)]
/ [y-(X*base!2)]
/ [y-(X*base!2)]
/ [y-(X*base!2)]
/ [y-(X*base7)]
/exp[-(base7/yf]
Index
12-Hour
12-Hour
12-Hour
12-Hour
7-Hour
7-Hour
Response
Min
Median
Max
Max
Median
Min
~
y (ppm)
1.1210
357.2540
9315.0000
8590.0000
9055.0000
6315.0000
5034.0000
53.7000
35.8000
0.2016
1
6.
2300.
64700
41277
32367
21070
10941
261.
280.
2.
.630
.000
.000
.000
.000
.000
.000
.100
.800
.474
days
152
166
138
76
76
76
105
214
77
69
Source: Abt Associates Inc (1995) exhibit 11.
Note: Onions, Rice, Oranges, and Cantaloupes only have one C-R function
Max and Min. base7 = 27 ppb and base 12 = 25 ppb which are equal to the
charcoal-filtered treatments.
and therefore do not have a
O3 concentrations in the
Table 5-4: Median SUM06 Relative Yield Loss Functions for Tree Seedlings
Tree Species
Aspen
Black Cherry
Douglas Fir
Eastern White Pine
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
RYL
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)x]
-exp[-(ozone/y)x]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
Index
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
SUM06
y (ppm) A,
131.92 .134
40.15 .084
124.82 6.702
57.69 .875
202.33 .008
183.46 .317
269.64 .700
34.91 6.395
40.77 2.609
1815.43 1.000
Source: Lee and Hogsett (1996) table 14. Individual exposure-response curves are reported using the
12hr-SUM06 index adjusted to a 92-day exposure duration.
Table 5-5: Median W126 Relative Yield Loss Functions for Tree Seedlings
Tree Species
RYL
Index
y (ppm)
Aspen
Black Cherry
Douglas Fir
Eastern White Pine
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)^]
-exp[-(ozone/y)x]
-exp[-(ozone/y)x]
-exp[-(ozone/y)x]
-exp[-(ozone/y)x]
W126
W126
W126
W126
W126
W126
W126
W126
W126
W126
109.81
38.92
106.83
63.23
159.63
179.06
318.12
36.35
51.38
1714.64
1.2198
0.9921
5.9631
1.6582
1.1900
1.2377
1.3756
5.7785
2.0889
1.0000
Source: Adjusted parameters from Lee and Hogsett (1996) table 14
Abt Associates Inc
Chapter
5-7
-------
100%
Dry Beans
Sorghum
Corn
Cotton
Peanut
Soybean
Wheat
Lettuce
Potato
p
o.
W126 (ppb)
Figure 5-3: Relative W126 Concentration-Response Curves for Major Crops (percent yield
loss/ppb). Curves were created with median functions if more than one function was available
Abt Associates Inc
Chapter
5-8
-------
100%
95% --
90% --
85% --
80% --
2 75% -
ffl
"5
a 70% -
65% -
60% -
55% -
50%
Aspen
—*— Black Cherry
Douglas Fir
Eastern White Pine
Pond. Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
<\ (x b
rv
W126 (ppb)
Figure 5-4: Relative W126 Concentration-Response Curves for Selected Tree Species (percent
biomass loss/ppb). Curves were created with median functions if more than one function was
available
5.2 Derivation of Seasonal Ozone Indices for Crop and Tree Species
Ozone indices were generated at the CMAQ grid level (12 km x 12 km resolution for the Eastern U.S. and
36 km x 36 km resolution for the Western U.S. or 44,432 gridcells). All SUM06, W126, 12-Hour and 7-
Hour O3 values were derived from monthly average O3 estimates based on hourly values adjusted down
by 10% to reflect a height differential between monitoring stations and crop foliage. For comparison we
also derived monthly O3 estimates without a 10% adjustment. The results of those analyses can be found
in Appendix G2.2 and H2.2. All results presented in Chapter 5 used estimated O3 exposure metrics with a
10% reduction of hourly values. Throughout the document we will refer to results with 10% reduction of
hourly values as "reduced" and results without 10% reduction of hourly values as "unreduced".
5.2.1 Ten Percent Adjustment
In response to direction provided by the EPA WAM, we applied a 10% reduction to the hourly O3 values
in each of the rollback scenarios whenever those data were used to predict crop yield loss. This was done
to compensate for the height difference between crops and O3 monitors (which may be placed
significantly higher than crop level). It should be noted that this 10% reduction has a magnified effect on
the values of metrics such as SUM06 and W126.
Abt Associates Inc
Chapter
5-9
-------
In the case of SUM06, O3 values between 60-66 ppb which previously contributed their full values to
SUM06 do not contribute at all to SUM06, because after a 10% reduction, they fall below the 60 ppb
threshold; O3 values greater than 66 do contribute to SUM06, but their contribution is reduced by 10%.
In the case of W126, not only are hourly O3 values reduced, but the weight assigned to them in the W126
sum is also reduced. In the "As is" scenario, we observe a 53% overall reduction in SUM06 levels and a
42% overall reduction in W 126 values1.
5.2.2 Characterization of Crop and Tree Growing Seasons
To generate seasonal SUM06 and W126 O3 indices used in the crop C-R functions, we chose to have the
last day of the experimental duration coincide with a mid-point harvest day. Since harvest dates tend to
vary across geographical regions, we relied on USDA tables (USD A, 1997) and crop profiles published
by the Integrated Pest Management Centers (IPM Centers, various).
When data were available in the 1997 USDA "Usual Planting and Harvesting Dates" then an early and
late harvest dates were recorded as well as the mid-point day of the most active harvesting season. When
data were collected from the IPM Centers, we used the mid-point of each state harvesting season as
anchor.
Harvest dates are not reported at the county level, so state-level approximations were used instead. When
harvest dates could not be found, we extrapolated a mid-point harvest date based on states situated in the
same climatic zone. States were grouped into four climatic zones based on yearly high temperatures as
reported in ESRI Annual World Temperature Zones (2005). The climatic classification used is presented
in Map E-l and
Map E-2 in Appendix E. Experimental durations for crops typically range between 2 and 3 months, so it
is reasonable to assume that a difference of a few days in the anchor harvest date has no significant effect
on yield change estimates. Growing seasons are shown in Figure F-l through Figure F-4.
In the case of field crops we used a scaled sum of monthly W126 (and SUM06) indices over the reference
duration starting from a mid-point harvest day and moving back in time:
crop . exp . duration
SUM06crop= £ SUMOe^ eq.
Where dt is the number of O3 exposure days in month /', Dt is the total number of days in month /', and
SUM06, is the estimated 12-Hour SUM06 O3 index in month /'.
Lettuce and potatoes are grown throughout the year, so we used the highest monthly rolling SUM06
(W126) adjusted for each crop exposure duration (53 days for lettuce and 62/66/70 days for potatoes):
Dec
crop
=MAX
m=Jan
crop, exp .duration
V SUM06
eq. (10)
r These percentages compare the sum of all SUM06 values (there are 44,432 such values for each location
in the POES) in the unreduced POES with the sum of all SUM06 values in the 10% reduced POES
(similarly for the W126 metric).
Abt Associates Inc Chapter 5-10
-------
For rice and cantaloupe, we used a weighted average of monthly 7-hour-average indices over the
reference duration starting from a mid-point harvest day and moving back in time:
crop, exp .duration
7/7F
crop
'crop. exp. duration
LA
eq. (11)
Onions and tomatoes are also grown throughout the year, so we used the highest monthly rolling 12-hour-
average weighted average adjusted for each crop exposure duration (105 days for onions and 76 days for
tomatoes):
\2hr^ = MAX
m=Jan
crop
crop. exp. duration
v
eq. (12)
For fruit trees, we assumed an April-October growing season and derived a 12-hour-average weighted
average index over this period:
Oct
\2hr =1=Apr
crop
eq. (13)
i=Apr
For tree seedlings we used the highest 3-month 12-Hour SUM06 (W126) index for the months of April
through October:
Oct
SUM06tree =
i-Apr
eq. (14)
5.3 Characterization of Crop and Tree Species Growing Ranges
Growing ranges for crops, fruits, and vegetables were derived from the 2002 Census of Agriculture and
from NASS 2001 County Crop Data. The 2002 Agricultural Census is based on farm surveys and we
assumed that any county having at least one farm growing a particular crop would be part of the growing
range for that crop. NASS County Crop Data on the other hand reports actual acreage planted and
harvested on a yearly basis. In other words, data from the 2002 Agricultural Census can be regarded as a
potential growing range, while NASS data is an actual growing range for a particular year. Table 5-6
shows the differences between the two data sources. In all cases, the 2002 Census data translates into
much larger growing ranges, and we therefore used the 2002 Census to convey average yield responses in
the maps and summary tables.
Since O3 levels and yield responses are computed at the CMAQ grid level, we then relied on spatial
interpolation techniques to relate the CMAQ grid to each crop growing range. A similar technique was
used to relate the USGS tree species maps to the CMAQ grid.
Abt Associates Inc
Chapter
5-11
-------
Table 5-6: Growing Ranges Reported in 2002 Census of Agriculture vs. 2001 Crop County Data for
Major Crops
Crop
Barley
Corn
Cotton
Dry Beans
Peanut
Potatoes
Rice
Sorghum
Soybean
Winter Wheat
'02 Census
Counties
1247
2587
662
580
406
1575
150
1316
2080
2516
of Agriculture
States
42
48
17
40
16
49
10
39
40
48
NASS '01
Crop County Data
Counties States
446
1994
181
171
191
129
101
597
1619
1666
22
41
17
11
9
13
6
18
31
41
Not Covered
Counties
801
593
481
409
215
1446
49
719
461
850
in '01
States
20
7
0
29
7
36
4
21
9
7
5.4 Estimation of Crop Yield and Tree Seedling Biomass Loss
Yield and biomass responses were estimated for four air quality scenarios using the 2001 surface as
baseline, or "as-is" scenario. For major field crops and trees, we computed yield changes based on both
monthly SUM06 and monthly W126 metrics. In the case of fruits and vegetables, O3 levels were
expressed in terms of the 12-hour-average and 7-hour-average metrics.
Results are summarized below as yield and biomass gains from the 2001 baseline. The box plots show
low and high quartiles, and minimum and maximum responses based on the monthly W126 metric Note
that the results presented here are based on simple county-level averages. No weights have been applied,
contrary to the regional statistics used in AGSIM©. The summary tables show median, mean, minimum,
and maximum responses based on the W126 metric. See Appendices G and I for additional tables and
maps based on the SUM06 metric.
5.4.1 Yield Impact on Cotton and Soybean
The box plots (Figures 5-5 and 5-6) and tables (5-7 to 5-11) below present unweighted yield responses
over the continental U.S. at the county level. The crop ranges used to compute national estimates are the
ones reported in the 2002 Census of Agriculture. The first box plot shows percent yield loss at the
baseline exposure level. The subsequent plots show percent yield gain from the baseline situation. The
shaded boxes represent the lowest and highest quartiles; the low and high bars show minimum and
maximum values. Exact numbers are presented in Table 5-7 through Table 5-12. Note that the minimums
and maximums reported on the maps will differ from the summary results presented in the tables because
the maps are based on individual gridcell results whereas the results in the tables and box plots have been
averaged across counties.
Results for all field crops, fruits and vegetables are shown in Appendix G. The results for exposures
without the 10% reduction can be found in Appendix G2.2
Abt Associates Inc
Chapter
5-12
-------
7% - -
a
'S
60
3% - -
- 7.7%
6.9%
5.90/
baseline
84 ppb 70 ppb
Figure 5-5: Median Yield Gain from Baseline for Cotton (reduced W126)
25 ppm-hr 15ppm-hr 21 ppm-hr 13ppm-hr
SUM06 W126
4% - -
'I 3°/
60
3.40/
2.60/
1.7%
baseline
84 ppb
70 ppb
25 ppm-hr 15 ppm-hr 21 ppm-hr 13 ppm-hr
SUM06 W126
Figure 5-6: Median Yield Gain from Baseline for Soybean (reduced W126)
Abt Associates Inc
Chapter 5-13
-------
The following tables show mean, maximum, minimum, median, and standard deviation in yield responses
for soybean and cotton for the six alternate air quality scenarios presented in Section 4. These are based
on straight average yield responses at the county-level (no production weights were applied). Median,
maximum and minimum C-R responses are included.
Table 5-7: Median Yield Gain from Baseline for Cotton (reduced W126)
Scenario
Baseline
84ppb
70ppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Table 5-8: Maximum
Scenario
Baseline
84ppb
70ppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Table 5-9: Minimum
Scenario
Baseline
84ppb
VOppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Max
7.71%
2.92%
6.37%
6.44%
7.01%
5.92%
6.94%
Yield Gain
Max
28.19%
7.39%
18.50%
18.79%
21.81%
16.44%
21.44%
Yield Gain
Max
12.49%
4.31%
9.78%
9.90%
10.96%
8.97%
10.84%
Min
0.04%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
from Baseline for
Min
0.86%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
from Baseline for
Min
0.11%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.78%
0.03%
0.32%
0.07%
0.26%
0.05%
0.19%
Cotton (reduced
Mean
5.63%
0.10%
1.65%
0.26%
1.34%
0.15%
0.95%
Median
0.65%
0.00%
0.21%
0.00%
0.05%
0.00%
0.00%
W126)
Median
5.14%
0.00%
1.20%
0.00%
0.32%
0.00%
0.00%
STD
0.69%
0.23%
0.55%
0.45%
0.59%
0.39%
0.56%
STD
2.92%
0.62%
1.99%
1.38%
2.21%
1.12%
2.00%
Cotton (reduced W126)
Mean
1.38%
0.05%
0.54%
0.12%
0.47%
0.07%
0.35%
Median
1.14%
0.00%
0.33%
0.00%
0.09%
0.00%
0.00%
STD
1.09%
0.33%
0.84%
0.68%
0.92%
0.59%
0.87%
Table 5-10: Median Yield Gain from Baseline for Soybean (reduced W126)
Scenario
Baseline
84ppb
70ppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Max
3.40%
1.73%
2.59%
1.73%
2.58%
1.25%
2.39%
Min
0.01%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.52%
0.03%
0.24%
0.03%
0.18%
0.01%
0.14%
Median
0.40%
0.00%
0.11%
0.00%
0.04%
0.00%
0.01%
STD
0.47%
0.15%
0.37%
0.13%
0.32%
0.09%
0.29%
Abt Associates Inc Chapter 5-14
-------
Table 5-11; Maximum Yield Gain from Baseline for Soybean (reduced W126)
Scenario
Baseline
84ppb
70ppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Max
8.21%
2.90%
5.61%
2.78%
5.03%
1.98%
4.56%
Min
0.26%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
2.41%
0.07%
0.78%
0.08%
0.61%
0.03%
0.45%
Median
2.24%
0.00%
0.48%
0.00%
0.19%
0.00%
0.03%
STD
1.38%
0.28%
0.90%
0.29%
0.82%
0.16%
0.70%
Table 5-12: Minimum Yield Gain from Baseline for Soybean (reduced W126)
Scenario
Baseline
84ppb
70ppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Max
1.46%
0.55%
1.07%
0.54%
0.97%
0.38%
0.89%
Min
0.03%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.37%
0.01%
0.13%
0.01%
0.11%
0.01%
0.08%
Median
0.33%
0.00%
0.08%
0.00%
0.03%
0.00%
0.01%
STD
0.24%
0.05%
0.16%
0.05%
0.15%
0.03%
0.13%
5.4.2 Yield Response Maps for Cotton and Soybean
Contrary to the above tables, the maps below were generated at the gridcell level. No weighting was
applied. The minimum and maximum statistics in the legends correspond to yield responses for individual
gridcells, and thus they tend to display more extreme yield responses than the statistic reported in the
previous tables (additional crop maps can be found in Appendix G). Once again, these maps report results
from 12-hr W126 exposures that have been calculated from 10% reduced hourly values. To compare
with results from unreduced exposures, see Appendix G2.2.
Abt Associates Inc Chapter 5-15
-------
O
"I
O
o
rt-
rt-
o
"H
5'
3
o.
e
r>
^ Yield (% loss)
^
o\ Z) <= 1 (m10 (max= 16,3)
No production value reported
O
o
rt-
p^-
o
SS
-------
» K.
oo ••
§ £
.5 O.
Is1
3 3
o »
oo §?
*> g
II
»•!
s.?
SI
o. s
o
3
'
O
a.
n
o
JU
p^-
Yield (% gain)
I 1<=1 (min = 0)
I 11 < YieW -==2
12 < YieW <=4
4 < YiekJ <-6
>6 (max = 9.6)
No production value reported
-------
- i.
oo ••
§ £
.5 O.
Is1
3 3
o »
, »
^ VI
O fB
II
»•!
s.?
SI
o. s
*?
>— o
3
'
O
o.
ss
JU
r+-
Yield (% gain)
I I <=1 (min = 0)
LD 1 < Yield <=2
I ] 2 < YieW -==4
^H 4 < Yield <=B
^•610 (max= 14.7)
No production value reported
-------
O (71
II
II
n 5'
o- 5
O
o
p^-
p^-
o
o
o
3
cr
i'
c"
O
o.
n
o
3
o
Yield (% gain)
I I <=1 (min = 0)
LD 1 < Yield -==2
I ] 2 < YieW -==4
^H 4 < Yield <=6
^•610 (max= 15.3)
No production value reported
-------
O (71
II
II
n 5'
o- 5
O
o
p^-
p^-
o
o
o
3
cr
I'
c"
O
o.
n
o
(max= 15,8)
3 No production value reported
o
Yield (% gain)
I I <=1 (min = 0)
L~| 1 < Yield -==2
I ] 2 < YieW -==4
^H 4 < Ylekl <=B
-------
3 £
s- ••
Z, 55
M S
a
a.
O
o
p^-
«-f-
o
"H
£'
3
o Yield (% gain)
O I |<=1 (min = 0)
ID 1 < Yield <=2
_@ 2 < Yield <=4
^' Hi 4 < Yield <=6
o. HH6< Yield <=10
sf !•! >10 (max= 14.6)
^" | No production value reported
-------
3
i
0
3
a 10 (max= 15.8)
^" | No production value reported
-------
g
p
(71
do
C/5
O
fB
P
9
ft
3
8?
o.
e
o
ft
o.
/r
•5»
/r\
s
Yield (% loss)
I l<=1 (min = 0.1)
I 11 < Yield <=2
>2 (max = 3.4)
No production value reported
s3^
t. ^
C/5
O
8?
O
n
-------
C £
-l ^3
s
s-
O
•0
T3 O
9" 3
w »
n w
o. 2.
S 5-
n n
i
o
*r , i '
jj*Q*
% i ' 2
'i l~^ff°*r' ^r--, ', <•' •*-.] kV-f
. ' / ,- . -t, i ^ -•-• ^ • • L* '"•• tf
Y -i / V • --1 .,(t f ' /A i V
N V; !^^ i 4^?^^
^ • "^&f
>-• . -• r ' < -ijS-^
S ~n
*)
w
&yyi • i
• iu;> •-'.'
i&
: : *> i^
I,' ..' £1 i *«*.£
^<4 {___-
Pi/<^f;
';.:. il^ili{^
.Yield(%gain)
cro
rD
v>
P-K
9°
>1 (max = 1.73)
No production value reported
-------
.
s s-
I s
S. ^
o S.
-i s
0 =T
? I
a1 s
O. st
C:7"--
3 ~*w 'V-V, ^n
^ cKT ph ' ^
V v ; i-' ' - ••- ; HT-'A
- *- ' I —i V V /" - 7>*J
i ' ,>rv!-, f__- LV-'J I / L' •'> 'x ,V'S^"V,
ri. _1 -,;7<^-'"^; -/. I r .yV'iSP S
ft
3
ts
O
o.
88
c Yield (% gain)
| ]<=1 (min = 0)
£ 3 1 < Yield <=2
H>2 (max = 2.6)
V
K^^
. •/ y.
CTQ
No oroduotion valwe reported
-------
O (71
T3 JL
•o a
3 O
=r S
I*
_ o
?
1 SL
a 5'
C/5
O
85
ft
3
ts
•fi/
o
o.
»
o_
f Yield (% gain)
5^
w I | <=1 (min = 0)
3 L J >1 (max = 1 .73)
O
5- No production value reported
-------
O (71
T3 JL
•a o.
3 O
I?
1 SL
••.- f , 1 <---,.
v-r "^~\ \ -^''"-- .••
••- A ''-i . **•? \ I .-•' ,
i / ,i •, - LJ ' L' ^i
/ -• '•. \ ^'•-: •--:-., 3s*^r- /r- •
- ,...:•' -•}- *#* ^ ,.\ / ..^-i
• ..—'.' -- --_!_. ..-iJ .., f S %•.) .<| _ ., '• • .•*•'•> i'/'tiL
L-V/ < "':• f^i-;... :
p '- L j-s • f ^y 4
^rr-i V\l-.-^-<'^^
•,;
88
2
5'
3
5
65
O.
•!
«-f-
S'
B Yield (% gain)
o I I <=1 (min = 0)
SlZ!~h < Yield <=2
>2 (max = 2.6)
.:"
^-_
• i
'I /// /O i
i • ^E,
T^
'. • .-'f\
/&
I
No production value reported
-------
O. Q.
1 O
ft as
II
ON 3
~* 03
—
5"
o
vj
a1
n
a.
~ Yield (% gain)
» I I <=1 (min = 0)
^ I I >1 (max= 1.8)
r> No production value reported
-------
Q. Q.
o O
o
Cd
—
5'
o
vj
a1
n
98
a.
Yield (% gain)
I | <=1 (min = 0)
I 11 < Yield <=2
[ | >2 (max = 3.2)
No production value reported
-------
5.4.3 Biomass Impact on Tree Seedlings
The mean national response is a weighted average of gridcell-level results. Results in the western grid
were weighted by nine to account for the difference in gridcell size between the eastern and western grid
(1,296 km2 vs. 144 km2). Results for all tree species are included in Appendix H. All results presented in
this section use estimated O3 exposure metrics with a 10% reduction of hourly values. The results for
exposures without the 10% reduction can be found in Appendix H2.2
The following tables show count, mean, maximum, minimum, median, and standard deviation statistics
for Aspen, Black Cherry and Ponderosa Pine based on the six scenarios described in Chapter 4.
10% - -
'I
60
o
£
4%
7.9%
4.20/
Figure 5-7: Median Biomass Gain from Baseline for Aspen (reduced W126)
Abt Associates Inc
Chapter
5-30
-------
40.9%
16.80/
28,6%-
15.4°/
11.3"/
j/ SUM06
Figure 5-8: Median Biomass Gain from Baseline for BlackCherry (reduced W126)
'I
60
O
15% - -
10% - -
- -9.3%
15.7%
16.80/
17.7%
15.7%
SUM06 „<$•
(<*
Figure 5-9: Median Biomass Gain from Baseline for Ponderosa Pine (reduced W126)
Abt Associates Inc
Chapter 5-31
-------
Table 5-13: Median Yield Gain from Baseline for Aspen (reduced W126)
Scenario
Baseline
84ppb
70ppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Table 5-14: Median
Scenario
Baseline
84ppb
70ppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Max
11.5%
5.8%
9.3%
5.3%
8.5%
4.2%
7.9%
Yield Gain from
Max
40.9%
16.8%
28.6%
15.4%
25.0%
11.3%
23.2%
Min
0.1%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Mean
2.7%
0.1%
1.1%
0.1%
0.6%
0.0%
0.5%
Baseline for Black Cherry
Min
2.8%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Table 5-15: Minimum Yield Gain from Baseline for
Scenario
Baseline
84ppb
VOppb
SUM06 25 ppm-hr
SUM06 15 ppm-hr
W 126 21 ppm-hr
W 126 13 ppm-hr
Max
19.9%
9.3%
15.7%
16.8%
17.7%
15.7%
17.7%
Min
0.1%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Mean
17.2%
0.4%
5.4%
0.3%
3.4%
0.1%
2.4%
Median
2.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
(reduced W126)
Median
16.8%
0.0%
4.0%
0.0%
0.9%
0.0%
0.0%
STD
2.5%
0.6%
1.9%
0.5%
1.5%
0.3%
1.3%
STD
6.5%
1.6%
5.3%
1.4%
4.6%
0.8%
3.8%
Ponderosa Pine (reduced W126)
Mean
1.3%
0.1%
0.3%
0.2%
0.3%
0.2%
0.3%
Median
0.8%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
STD
5.5%
2.1%
4.1%
3.8%
4.5%
3.4%
4.4%
Abt Associates Inc
Chapter
5-32
-------
s
o'
fts
•a
§
I
1
o.
^
Biomass (% loss)
QZ1 <1 (min = 0.1)
I | 1 < Biomass <=2
Q^j 2 < Biomass <=4
^H 4 * Biomass <=6
^B 3 < Biomass <=10
•• >10 (max = 12)
No production value reported
VI
TS
»
n
VI
TS
O
C/l
C/5
re
re
a
5'
era
5
o'
5
re
OB
O
OB
re
o
•s
C/5
2.
JT
r?
a
H
re
re
C/5
re
S"
-------
p p
3
C
3
=
o
oo S
•^ 3
•a &
3 B.
y B
& if
8 I
o
"S
>
VI
O
O
o.
P
P
n
CTQ
ft
oo
s-
o
Biomass (% gain)
I |<1 (min = 0)
I I 1 < Biomass <=2
2 < Biomass <=4
| 4 < Biomass <=6
Hi >6 (max = 6.4)
_ No production value reported
-------
M "O
3tt\
i
C H-
3 ^
o ffi
o §
B-
o
"S
>
VI
O
O
e
P
o.
•!
«-f-
n'
o
CTQ
oo
o
Biomass (% gain)
I |<1 (min = 0)
I 11 < Biomass <=2
| 2 < Biomass <=4
14 < Biomass <=6
^H >6 (max = 9.7)
No production value reported
-------
c -a
/X 71
rs oo
O. ••
CO
ft O
a 3
2
5'
o
a
t/>
rs
o
>
t/>
rs
O
o.
o_
cr
3
o
05
I
o
(71
Biomass (% gain)
| | <1 (min = 0)
I | '\ < Biomass <=2
Q^ 2 < Biomass <=4
JH>4 (max = 5.7)
No production value reported
-------
S3
(Jl
o. ••
CO
o •••
ft O
a 3
O
03
S3
o
•1
>
O.
S3
o_
5
S3
3
o
05
i
o
h^
(71
Bio mass (% gain)
CZ1 <1 (min = 0)
I 11 < Biomass <=2
^] 2 < Biomass <=4
^H 4 < Biomass <=6
^H >6 (max 8.7)
No production value reported
-------
S.-S
g V
S.S
o
B.
o
td
88
O
P
O.
70
o
o Biomass (% gain)
^ [ZZl^l (min = 0)
£ LZ11 < Biomass <=2
^ IU 2 < Biomass <=4
o
N> ^H 4 < Biomass <=6
•o HH6< Biomass <=10
3 ••>10 (max = 14.9)
o
No production value reported
-------
S.-S
£ <*
8 k
O. H-
^B
K s-
.^ 3
O
s.
O
Cd
88
O
>
VI
•a
p
O.
n
o
O
Biomass (% gain)
I I <=1 (min = 0)
H) 1 < Biomass <=2
HI 2 < Biomass <=4
IB! 4 < Biomass <=6
^H 6 < Biomass <=10
•• >10 (max = 1S.2)
i No production value reported
-------
g
5"
(Jl
IN*
ffi
5'
3
S3
r
o
S
S"
O
O
fB
O.
e
rs
ft
O.
Bio mass {% loss)
I I <5 (min = 2.8)
I I 5 < Biornass <=1Q
HJ 10 < Biomass <=20
HI 20 < Biomass <=3Q
^B >30 (max =40.9)
No production value reported
-!
^
70
-------
si.
CTQ '
s- ^>
n <*>
* *
o
,_,
o. Cd
^
ft
O
o
§ Bio mas s {% gain)
I" I l<5 (min = 0)
5 <• Biomass <=10
>10 (max= 16.8)
No production value reported
-------
si.
CTQ '
-r l>>
o
o
o.
"S
s>
<-f.
o'
^ Biomass (% gain)
" <5 (min - 0)
5< Biomass <=10
10 < Biomass <=20
>20 (max= 28.6)
No production value reported
-------
I
C/5
O O
3
n n
O- 5T.
p
r>
n
o
o
e
o.
i
s>
<-f.
n'
Bio mass {% gain)
| <5 (min = 0)
| 5 < Biomass <=10
| >10 (max= 15.4)
No production value reported
-------
I
O
O O
3
n n
O- 5T.
p
r>
n
o
«-K
3'
88
o
e
o.
i
s>
<-f-
n'
Biomass (% gain)
5 (r"!'1 = 0)
| 5< Biomass <=10
I 10 < Biomass <=20
>20 (max = 25)
No production value reported
-------
I
«- o
o g
3 S.
50 03
*§•
H- „
KJ 03
s2 £T
O
cr
re
5
re
o
o
e
o.
i
88
<-f-
o'
Biomass (% gain)
I I <=5 (min = 0)
I | 5 < Biomass <=10
^H>10 (max= 12.7)
| No production value reported
-------
I
oo
«- o
o g
3 S.
s
o. n
$*
h^ i^
" ' O
O
cr
n
5
ft
O
o
e
o.
i
88
«-f-
o'
Biomass (% gain)
^] <=5 (min = 0)
I | 5 < Biomass <=10
HI10 * Biomass <=20
•I >20 (max = 24.7)
| No production value reported
-------
O (71
O. W>
O So'
O
9
O.
n
o
o
9
O.
n
o.
CTQ
i
O
o
rt-
o
!T
fB
Bio mass (% loss)
EZZ!<=1 (mm =0.1)
I 11 < Biomass <-2
^]2< Biomass <=4
j^H 4 < Biomass <=6
•• 6 < Biomass <=10
IHI ' i :• (max= 19.9)
No production value reported
O
o.
n
o
2
5"
fB
fB
O
VI
fB
88
-------
s«
CTQ ' .
% O
^ 63
00 !K
J_ O
§ O
fej ^
a. 5-
1 ^
ii
o" 63
§•§
II
^g
< P
N> hfl
sS 5'
rs
O
O.
ft
P
O.
o
o_
p
Bio mass (% gain)
CZ3 6 (max = 9.3)
No production value reported
-------
" (71
era '.
rf -
VI
oo SP.
J_ O
S o
3 S»
|i
o" 03
^i P
§•§
O ^^
o. ?
^g
< p
N> hfl
xS 5'
rs
O
O.
ft
P
O.
o
o
Biomass (% gain)
I l<=1 (min = 0>
HJ1 < Biomass <=2
^]2< Biomass <=4
BH 4 < Biomass <=6
|^H 6 < Biomass <=10
H|>10 (max= 15.7)
No production value reported
-------
o
S (71
O w
tt O
•U- o
I3
03
It
5
5'
O
O.
P
O.
n
o
Biomass (% gain)
I l<=1 {mm=0}
HJ1 < Biomass <=2
^] 2 < Biomass <=4
BH 4 < Biomass <=6
|^H 6 < Biomass <=10
^H >10 (max= 16.8)
No production value reported
-------
o
S (71
is-
O en
££
•O s
3
•U- o
? 3
S 03
^- ^
^ o
2 B
10 (max= 17.7)
No production value reported
-------
(71
II
O
9
5*
o
O
o.
5
5"
o
o.
n
o
VI
»
O.
n
o
Biomass (% gain)
I |<=1 (min=0)
I 11 < Biomass <=2
j^H 2 < Biomass <=4
^^| 4 < Biomass <=6
BB 6 < Biomass <=10
•I >IO (max =15.7)
Ho production value reported
-------
(71
II
O
9
5*
o
O
o.
5
5"
o
o.
n
o
VI
»
O.
n
o
Biomass (% gain)
Value
I |<=1 (min=0)
I 11 < Biomass <=2
j^H 2 < Biomass <=4
^^| 4 < Biomass <=6
BB 6 < Biomass <=10
•I >IO (max =17.7)
Ho production value reported
-------
-------
6. Economic Benefits
6.1 AGSIM©
AGSIM© is a large-scale econometric simulation model of regional crop and national livestock
production in the United States. The model ties together econometrically estimated demand and supply
equations and solves for the set of crop and livestock prices that simultaneously clears all markets in a
given year for given exogenous factors. The model is capable of analyzing the economic effects of
changes in farm programs and other relevant policies on 1) changes in regional per-acre crop yields, 2)
changes in production costs, 3) changes in target prices and set-aside rates, 4) changes in paid land
diversion by crop and region, and 5) acreage in the conservation reserve. The simulation model can
provide a full welfare evaluation of 1) domestic consumers' surplus, 2) farm income, 3) government
program payments, and foreign surplus. A complete description of the model is provided in Appendix I.
6.2 Methodology
AGSIM© was used to quantify the economic benefits resulting from meeting alternate O3 standards as
specified in Section 4. Crop yield changes generated in Section 5 were fed into the model, which then
solved for the new level of supply and demand. No other exogenous shock was applied. Yield changes
incorporated into AGSIM© were broken down by ERS region. The mean regional yield response is a
weighted average of county-level results. Weights were derived from NASS 2001 County Crop Data
(planted acreage) and when NASS data was not available for a particular crop, we used the 2002 Census
of Agriculture instead (number of harvested farms). All negative values were zeroed out before averaging.
Five scenarios were considered based on the five rollback scenarios presented in Section 4. For each
rollback, we measured the economic benefits incurred for median, minimum, and maximum yield
responses, for a total of fifteen model runs. In addition a soybean-only scenario was considered, keeping
all other crop yields constant.
A summary of simulation results is presented in the following section.
6.3 Results
Total benefits are summarized in Table 6-1 below. This table provides the overall change in producer and
consumer surplus for field crops plus fruits and vegetables and field crops by themselves. The results for
the soybean-only runs are separated in Table 6-2.
Leaving out farm payments, since they do not translate into increased welfare, total economic gains range
from $7 million under the 84 ppb rollback to $ 199 million under the 70 ppb rollback (based on the
median concentration-response levels). When fruits and vegetable yield changes are added, total gains
range from $75 to $383 million.
Table 6-3 and Table 6-4 present similar results for the scenarios based on the non-adjusted metric (i.e.
hourly O3 values were not reduced by 10 percent).
Table 6-5 compares results from the scenarios based on the 10%-adjusted metric vs. the non-adjusted
metric. In general, the results without the 10% reduction are 10 to 100% higher. The differences are
somewhat greater for the "As Is Soybeans Only MEDIAN" and the "Roll 70 Soybeans Only MIN"
scenarios.
Abt Associates Inc Chapter 6-1
-------
Table 6-6 compares the results for the 25 vs. 21 scenarios and the 15 versus the 13 scenarios. The
SUM06 scenarios (25 and 15) have greater benefits than the comparable W126 scenarios (21 and 13).
The effect of the 10 percent reduction is reasonably similar across the different scenarios, with a
somewhat more pronounced effect on the W126/13 scenario.
Table 6-1: Total Undiscounted Economic Surplus Effect, 2001/02 through 2014/15 (with 10%
adjustment)
Scenario
AS-IS
roll_84
roll_70
roll_SUM06 25
roll_SUM06 15
roll_W12621
roll_W126 13
C-R Level
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
Change in Total Economic Surplus ($ million)
Field crops and
F&V
Avg/year
$1,792
$3,250
$1,709
$80
$114
$75
$407
$749
$381
$189
$269
$182
$400
$710
$379
$144
$179
$137
$310
$440
$291
Field crops and
F&V
(less farm
payments)
Avg/year
$1,669
$2,365
$1,586
$75
$96
$70
$383
$564
$356
$181
$230
$172
$367
$532
$345
$10
$35
$14
$29
$136
$32
Field crops only
Avg/year
$326
$1,687
$328
$11
$40
$12
$71
$383
$71
$22
$89
$28
$90
$373
$93
$149
$204
$145
$326
$562
$308
Field crops only
(less farm
payments)
Avg/year
$203
$802
$205
$7
$22
$8
$46
$199
$46
$14
$50
$18
$56
$195
$58
$16
$60
$21
$45
$258
$49
Source: AGSIM© model simulation results.
Abt Associates Inc
Chapter
6-2
-------
Table 6-2: Total Undiscounted Economic Surplus Effect of Soybean Yield Response, 2001/02
through 2014/15 (with 10% adjustment)
Scenario
AS-IS
roll_84
roll_70
roll_SUM06 25
roll_SUM06 15
roll_W12621
roll_W126 13
C-R Level
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
Change in Total Economic
Avg/year
$74
$310
$46
$3
$6
$1
$28
$79
$28
$3
$7
$1
$23
$66
$11
$1
$2
$0
$12
$32
$6
Surplus ($ million)
(farm payments excl.)
Avg/year
$52
$216
$32
$2
$4
$1
$20
$55
$5
$2
$5
$1
$16
$46
$8
$1
$2
$0
$17
$47
$8
Source: AGSIM© model simulation results.
Abt Associates Inc
Chapter
6-3
-------
Table 6-3: Total Undiscounted Economic Surplus Effect, 2001/02 through 2014/15 (without 10%
adjustment)
Scenario
AS-IS
roll_84
roll_70
roll_SUM06 25
roll_SUM06 15
roll_W12621
roll_W126 13
C-R Level
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
Change in Total Economic Surplus ($ million)
Field crops and
F&V
Avg/year
$2,121
$4,448
$1,996
$93
$127
$86
$498
$948
$454
$219
$309
$211
$454
$866
$421
$174
$230
$168
$397
$699
$371
Field crops and
F&V
(less farm
payments)
Avg/year
$2,007
$3,095
$1,899
$87
$107
$80
$454
$686
$417
$206
$261
$196
$415
$628
$385
$164
$201
$157
$366
$529
$341
Field crops only
Avg/year
$349
$2,570
$317
$18
$44
$16
$127
$545
$111
$35
$110
$40
$111
$494
$104
$26
$70
$32
$86
$363
$84
Field crops only
(less farm
payments)
Avg/year
$236
$1,217
$220
$11
$25
$10
$83
$282
$74
$22
$62
$26
$72
$256
$68
$16
$41
$20
$55
$192
$55
Source: AGSIM© model simulation results.
Abt Associates Inc
Chapter
6-4
-------
Table 6-4: Total Undiscounted Economic Surplus Effect of Soybean Yield Response, 2001/02
through 2014/15 (without 10% adjustment)
Scenario
AS-IS
roll_84
roll_70
roll_SUM06 25
roll_SUM06 15
roll_W12621
roll_W126 13
C-R Level
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
Change in Total Economic
Avg/year
$114
$383
$61
$3
$5
$1
$38
$85
$16
$3
$6
$1
$30
$70
$13
$1
$2
$0
$22
$48
$9
Surplus ($ million)
(farm payments excl.)
Avg/year
$163
$553
$88
$5
$7
$1
$54
$122
$22
$5
$9
$2
$43
$101
$19
$2
$3
$1
$31
$69
$13
Source: AGSIM© model simulation results.
Abt Associates Inc
Chapter
6-5
-------
Table 6-5: Comparison between Adjusted vs. Non-Adjusted Metric on Undiscounted Economic
Surplus Effects, 2001/02 through 2014/15
Scenario
C-R Level
Ratio of Adjusted to Non-Adjusted Metric
Field crops and F& V
(less farm payments)
Avg/year
Field crops only
(less farm payments)
Avg/year
AS-IS
roll_84
roll_70
roll_SUM06 25
roll_SUM06 15
roll_W12621
roll_W126 13
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
20%
31%
20%
15%
11%
13%
19%
22%
17%
14%
13%
14%
13%
18%
12%
15%
12%
15%
18%
20%
17%
16%
52%
7%
50%
12%
32%
79%
42%
61%
55%
23%
43%
27%
31%
18%
62%
17%
47%
89%
41%
72%
Source: AGSIM© model simulation results.
Table 6-6: Comparison Between SUM06 vs. W126 Metric on Undiscounted Economic Surplus
Effects, 2001/02 through 2014/15
Scenario
C-R Level
Ratio of SUM06
Field crops and F& V
(less farm payments)
Avg/year
25 ppm-hr SUM06 vs.
21ppm-hrW126
15 ppm-hr SUM06 vs.
13 ppm-hr W 126
Source: AGSIM© model
all medians
7 max, 9 median
7 min, 9 median
all medians
7 max, 9 median
7 min, 9 median
simulation results.
10%
Adjusted
26%
30%
25%
13%
19%
13%
Non-
Adjusted
26%
29%
25%
18%
21%
18%
to W 126 Metric
Field crops only
(less farm payments)
Avg/year
10%
Adjusted
34%
52%
27%
29%
33%
25%
Non-
Adjusted
40%
45%
31%
91%
44%
82%
Abt Associates Inc
Chapter
6-6
-------
7. Tree Growth Simulation
In the 1996 O3 Staff Paper, analyses on trees were limited to the seedling growth stage. In order to go
beyond the seedling stage for the current review, we used a tree growth simulation model, TREGRO, as a
tool to evaluate the effect of just meeting alternate O3 standards on select O3-sensitive tree species.
The response of total tree growth of two species, red maple and yellow (or tulip) poplar was simulated in
two locations (Shenandoah National Park, VA, and Cranberry, NC) in the southern Appalachian
Mountains to the five scenarios of 63 reduction used previously in this report. These simulations were
done using the computer model, TREGRO. The results of this investigation are given below. A report
providing the details of the methodology and results from this examination can be found in Appendix J.
7.1 Summary of Results
The simulations produced a prediction of average annual total tree growth over the 3-year period for each
scenario. These results were compared to the base scenario, which consisted of a prediction of growth
under the hourly meteorology and 03 conditions for the period 1993-1995.
The predictions indicated substantial increases in 3-year total tree growth increments with reduction of 03
exposure, particularly under Scenario 3, a rollback to conform to the standard of the 1st highest maximum
8-hour average being no greater than 0.070 ppm. Yellow poplar had nearly a twenty percent increase in
growth in response to this scenario, an average annual increase of 6.5%.
Table 7-1 Predicted percent increases in total tree growth over a 3-year period under the 4 ozone
(O3) reduction scenarios.
Yellow Yellow
Red maple, Red maple, poplar,
Shenandoah Cranberry Shenandoah
Scenario 1 1.22% 0.08%
Scenario 2 1.02% 0.20%
Scenario 3 8.14% 6.92% 1.15%
Scenario 4 6.72% 4.15% 1.03%
Scenario_5 4.49% 2.99% 0.60%
Scenario 1 : Rollback current EPA standard 4th highest max.
Scenario 2: Rollback SUM06 25 ppm-hr
Scenario 3: Rollback 1st highest max. 8-hr avg. 0.070 ppm
Scenario 4: Rollback 4th highest max. 8-hr avg. 0.070 ppm
Scenario 5: Rollback SUM06 15ppm-hr
poplar,
Cranberry
19.61%
11.73%
8.26%
8-hr avg. 0.085 ppm
Abt Associates Inc Chapter 7-1
-------
20%
Scenario_l
IScenario_2
Scenario_3
Scenario_4
I Scenario 5
U
Figure 7-1 Tree growth response of red maple and yellow poplar in forests of Shenandoah National Park,
Virginia and Cranberry, North Carolina to ozone (O3) reduction scenarios.
Red maple was simulated to have a similar response to Scenario 3 in Shenandoah and in Cranberry.
However, it had nearly twice the increase to Scenario 4 at Shenandoah as it did at Cranberry. The
response to Scenario 5 was slightly less than Scenario 4 for red maple and for yellow poplar in both
Shenandoah and Cranberry. The response of yellow poplar at Cranberry to Scenario 5 was still very
large, with growth projected to increase more than 8% under this level of O3 reduction.
Yellow poplar had a very different response to 03 reduction at Shenandoah compared to Cranberry. The
temperatures at Cranberry are more in the middle of the range of temperatures over which yellow poplar
is found than are the cool temperatures of Shenandoah, making conditions at Cranberry more ideal for
growth. Higher growth rates may cause greater sensitivity to 63. Red maple has a much larger
geographical distribution, so that the temperature differences between Shenandoah and Cranberry are less
likely to affect the growth response. This phenomenon was reflected in the simulations.
Finally, Scenarios 1 and 2 produced very little growth response in either species. These scenarios
produced no change in the predicted 63 exposure at Cranberry, so they were not even simulated. At
Shenandoah, the change in 03 exposure to Scenarios 1 and 2 was very slight.
Methodological details, including a discussion of the uncertainty of the investigation, can be found in
Appendix J.
Abt Associates Inc
Chapter
7-2
-------
-------
Appendices
-------
Appendix A Detailed Results from Dropout Monitor
Investigation
Section 3 presents O3 interpolation performance results for the Eastern U.S. and the Western US. Because
the monitor coverage in California is much denser than the rest of the Western US, we examine here the
performance results for California alone, as well as for the West excluding California. For a detailed
explanation of the data and methods used to generate each metric, please refer section 1-3 of this report.
Choice of Interpolation Approaches
Two interpolation techniques were retained; 1) VNA (distance-weighted averages of neighboring monitor
values), and 2) eVNA (VNA with model-adjusted neighboring monitor values).
In the case of eVNA, we also compared two condition-specific adjustment methods; 1) Month-Decile
(hourly monitor values are split evenly into ten rank-ordered deciles for every month), and 2) Month-
Hour (hourly monitor values are split evenly into 24 groups by time of day for every month). We also
included data for predictions based only on the CMAQ-generated values for the Western U.S. at the 36
km x 36 km grid level.
Finally 4 mixed VNA-eVNA interpolations were compiled based on the distance between a monitor and
its neighbors. The first mixed approach uses VNA for neighbors under 50 km and HMD for neighbors
beyond 50 km (HMD_VNA_50). Similarly the second mixed approach uses VNA for neighbors under
100 km and HMD for neighbors beyond 100 km (HMD_VNA_100). The last 2 mixed approaches use a
similar blend of VNA and HMH.
The approaches are named accordingly:
VNA Distance-weighted averages with no scaling
HMD eVNA interpolation with Hour-Month-Decile scaling
HMH eVNA interpolation with Hour-Month-Hour scaling
CMAQ Model-generated values
• HMD_VNA_50 Mixed VNA and HMD interpolation at 50 km cutoff
• HMD_VNA_100 Mixed VNA and HMD interpolation at 100 km cutoff
• HMH_VNA_50 Mixed VNA and HMH interpolation at 50 km cutoff
• HMH_VNA_100 Mixed VNA and HMH interpolation at 100 km cutoff
Choice of Dropout Monitor Sites
The present iteration includes all "complete" AQS and CASTNET monitor sites in the Western U.S. (299
monitors). Some monitors identified as located in densely populated urban areas were left out. Additional
AQS monitors were excluded when it was discovered that they corresponded to CASTNet monitors and
contained identical data.
Results
Summary results showing a comparison between seven interpolation methods and the CMAQ model
estimations are presented in Tables A-l through A-6. Tables A1-A3 refer to the set of monitors within
California; tables A4-A6 refer to the Western U.S. excluding California. In both sets, one table is given
for each of the three metrics considered - (8-hour maximum, SUM06, and W126).
Abt Associates Inc Appendix A-l
-------
Table A-l: Com
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
parison of Interpolation
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Methods
for California. (8-hour
Count Mean Median Min
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
1.01
12.71
5.4%
16.5%
-3.59
7.75
-3.4%
9.4%
-4.22
7.65
-3.8%
9.2%
-4.13
7.59
-3.8%
9.2%
-2.87
7.64
-2.6%
9.4%
-4.15
7.69
-3.7%
9.3%
-3.86
7.65
-3.5%
9.3%
-4.18
7.80
-3.7%
9.5%
1.24
11.18
1.6%
13.8%
-3.76
6.42
-4.3%
7.7%
-3.04
6.60
-3.8%
7.9%
-3.40
6.60
-4.1%
7.6%
-2.71
5.93
-3.1%
7.6%
-3.30
6.36
-3.8%
7.9%
-3.72
6.15
-4.2%
7.5%
-3.26
6.64
-3.7%
7.9%
Maximum)
Max Range STD
-45.13
0.05
-37.6%
0.0%
-41.10
0.03
-34.2%
0.0%
-25.34
0.07
-25.7%
0.1%
-24.81
0.18
-25.7%
0.2%
-40.01
0.16
-33.3%
0.2%
-24.64
0.07
-25.6%
0.1%
-24.60
0.16
-25.6%
0.2%
-26.55
0.07
-27.6%
0.1%
36.29
45.13
81.41
45.08
85.0% 122.6%
85.0%
21.76
41.10
53.8%
53.8%
19.18
25.34
51.5%
51.5%
20.05
24.81
53.8%
53.8%
22.19
40.01
53.4%
53.4%
19.33
24.64
51.9%
51.9%
19.89
24.60
53.4%
53.4%
22.65
26.55
60.8%
60.8%
85.0%
62.86
41.07
88.1%
53.8%
44.52
25.26
77.2%
51.4%
44.86
24.63
79.6%
53.7%
62.20
39.85
86.7%
53.2%
43.96
24.56
77.5%
51.8%
44.48
24.44
79.0%
53.2%
49.20
26.48
88.4%
60.7%
16.12
9.97
21.4%
14.6%
9.41
6.42
11.7%
7.8%
8.84
6.13
11.2%
7.4%
8.75
6.00
11.1%
7.4%
9.47
6.29
12.0%
7.8%
8.89
6.09
11.2%
7.4%
8.85
5.89
11.3%
7.3%
9.07
6.24
11.8%
7.9%
Abt Associates Inc
Appendix
A-2
-------
Table A-2: Com
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
parison of Interpolation
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Methods
for California. (SUM06)
Count Mean Median Min
150
150
149
149
150
150
149
149
150
150
149
149
150
150
149
149
150
150
149
149
150
150
149
149
150
150
149
149
150
150
149
149
-0.04
11.95
201.8%
227.2%
-1.88
7.66
11.6%
55.8%
-3.17
7.82
11.9%
53.9%
-2.78
7.69
12.4%
54.0%
-1.80
7.73
12.8%
57.3%
-3.14
7.80
11.7%
53.5%
-2.76
7.70
11.0%
52.7%
-3.16
7.84
14.5%
56.3%
2.86
9.36
14.5%
44.6%
-1.18
4.82
-7.6%
37.6%
-1.77
5.10
-11.7%
32.8%
-1.54
4.92
-11.4%
28.6%
-1.10
4.88
-7.3%
36.8%
-1.61
5.14
-11.7%
32.8%
-1.69
4.90
-10.4%
28.7%
-1.90
5.14
-12.0%
32.8%
Max Range STD
-44.93
0.00
-73.0%
0.0%
-32.15
0.13
-96.7%
0.9%
-33.75
0.13
-92.6%
1.0%
-31.81
0.08
-96.7%
0.3%
-30.79
0.04
-95.1%
0.2%
-33.81
0.01
-92.6%
0.0%
-32.27
0.01
-96.7%
1.1%
-34.39
0.04
-92.6%
1.0%
34.62
44.93
6816.2%
6816.2%
44.64
44.64
342.7%
342.7%
41.26
41.26
430.3%
430.3%
44.64
44.64
430.3%
430.3%
40.58
40.58
422.6%
422.6%
40.15
40.15
430.3%
430.3%
39.65
39.65
430.3%
430.3%
41.06
41.06
499.2%
499.2%
79.55
44.92
6889.3%
6816.2%
76.80
44.52
439.4%
341.8%
75.01
41.13
522.9%
429.3%
76.45
44.56
526.9%
430.0%
71.37
40.54
517.7%
422.4%
73.97
40.14
522.9%
430.2%
71.91
39.63
526.9%
429.2%
75.45
41.02
591.8%
498.2%
15.38
9.69
662.3%
654.1%
11.22
8.42
87.4%
68.3%
11.16
8.56
90.8%
74.0%
11.11
8.49
91.8%
75.2%
11.19
8.28
90.5%
71.2%
11.00
8.37
90.1%
73.4%
10.89
8.18
88.9%
72.4%
11.08
8.45
99.1%
82.8%
Abt Associates Inc
Appendix
A-3
-------
Table A-3: Com
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
parison of Interpolation
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Methods
for California. (W126)
Count Mean Median Min
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
-0.31
9.60
124.1%
148.1%
-1.36
5.82
24.7%
58.9%
-2.30
6.00
21.8%
56.0%
-2.03
5.92
24.6%
58.4%
-1.14
5.91
24.6%
58.4%
-2.25
6.02
22.8%
57.0%
-1.92
5.96
23.5%
57.5%
-2.29
6.08
31.4%
65.9%
2.87
7.32
17.1%
49.1%
-0.66
3.54
-4.1%
28.5%
-0.92
3.55
-5.9%
28.4%
-0.92
3.39
-5.9%
28.3%
-0.55
3.29
-5.1%
30.0%
-0.85
3.60
-5.6%
28.4%
-0.97
3.50
-6.5%
28.9%
-1.06
3.70
-7.8%
27.7%
Max Range STD
-41.98
0.05
-65.0%
0.2%
-26.39
0.11
-73.4%
1.0%
-29.36
0.02
-72.7%
0.0%
-28.32
0.02
-71.1%
0.0%
-27.12
0.02
-73.4%
0.2%
-29.59
0.02
-74.0%
0.0%
-28.59
0.02
-72.6%
0.0%
-29.80
0.02
-72.0%
0.0%
26.57
41.98
5993.6%
5993.6%
27.84
27.84
2810.1%
2810.1%
28.41
29.36
2401.1%
2401.1%
28.80
28.80
2810.1%
2810.1%
28.57
28.57
2641.7%
2641.7%
28.07
29.59
2548.5%
2548.5%
28.22
28.59
2641.7%
2641.7%
28.58
29.80
3636.8%
3636.8%
68.54
41.93
6058.6%
5993.3%
54.23
27.73
2883.5%
2809.2%
57.77
29.34
2473.8%
2401.0%
57.12
28.79
2881.3%
2810.1%
55.69
28.55
2715.1%
2641.4%
57.66
29.57
2622.5%
2548.4%
56.82
28.58
2714.3%
2641.6%
58.37
29.78
3708.8%
3636.7%
12.70
8.33
510.4%
503.9%
8.54
6.39
235.5%
229.4%
8.55
6.51
203.7%
197.0%
8.49
6.42
235.6%
229.6%
8.64
6.40
222.4%
216.0%
8.53
6.45
215.1%
208.7%
8.44
6.27
222.1%
215.8%
8.67
6.58
302.2%
296.6%
Abt Associates Inc
Appendix
A-4
-------
Table A-4: Comparison of Interpolation Methods for the Western U.S. Excluding California (8-
hour Maximum)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count Mean
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
137
8.59
9.69
13.4%
15.0%
-2.12
4.87
-2.5%
7.2%
-2.07
4.89
-2.4%
7.3%
-2.09
4.87
-2.4%
7.2%
-1.00
4.92
-1.0%
7.3%
-1.91
4.78
-2.1%
7.1%
-1.74
4.73
-2.0%
7.0%
-2.04
4.80
-2.3%
7.1%
Median Min
8.95
9.37
13.0%
13.1%
-2.41
4.15
-3.2%
6.0%
-2.35
4.19
-3.6%
6.1%
-2.57
4.26
-3.8%
6.1%
-1.89
4.16
-2.6%
5.8%
-2.30
3.93
-3.3%
5.6%
-2.36
4.16
-3.3%
5.9%
-2.33
4.19
-3.3%
5.8%
Max
-11.09
0.10
-19.8%
0.1%
-18.17
0.04
-32.5%
0.1%
-18.17
0.01
-32.5%
0.0%
-18.17
0.04
-32.5%
0.1%
-16.66
0.00
-29.8%
0.0%
-16.66
0.02
-29.8%
0.0%
-16.66
0.00
-29.8%
0.0%
-13.05
0.02
-22.3%
0.0%
Range STD
32.86
32.86
63.5%
63.5%
19.65
19.65
43.4%
43.4%
19.65
19.65
43.4%
43.4%
19.65
19.65
43.4%
43.4%
24.83
24.83
35.1%
35.1%
15.87
16.66
35.1%
35.1%
15.87
16.66
35.1%
35.1%
13.56
13.56
26.2%
26.2%
43.95
32.76
83.4%
63.4%
37.82
19.61
76.0%
43.4%
37.82
19.64
76.0%
43.4%
37.82
19.61
76.0%
43.4%
41.49
24.82
64.9%
35.1%
32.53
16.64
64.9%
35.0%
32.53
16.66
64.9%
35.1%
26.61
13.54
48.4%
26.1%
8.12
6.76
14.0%
12.3%
5.72
3.67
9.1%
6.1%
5.80
3.75
9.4%
6.3%
5.66
3.57
9.1%
6.0%
6.15
3.82
9.4%
6.1%
5.63
3.52
8.9%
5.8%
5.56
3.40
8.7%
5.6%
5.53
3.42
8.7%
5.6%
Abt Associates Inc
Appendix
A-5
-------
Table A-5: Comparison of Interpolation Methods for the Western U.S. Excluding California
(SUM06)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count Mean
136
136
130
130
136
136
130
130
136
136
130
130
136
136
130
130
136
136
130
130
136
136
130
130
136
136
130
130
136
136
130
130
7.98
9.74
409.5%
420.1%
-2.00
5.78
20.2%
72.0%
-2.73
5.76
15.9%
69.6%
-2.59
5.57
15.7%
68.5%
-1.74
5.68
24.7%
73.8%
-2.64
5.74
19.2%
72.6%
-2.46
5.56
21.7%
73.0%
-2.68
5.70
22.0%
72.2%
Median Min Max Range STD
5.95
7.28
39.2%
40.3%
-1.10
3.97
-13.2%
35.0%
-1.38
4.46
-21.1%
31.7%
-1.16
4.31
-20.2%
31.7%
-0.66
3.72
-10.6%
30.1%
-1.37
4.48
-21.0%
32.9%
-1.09
4.22
-15.8%
31.4%
-1.28
4.63
-17.9%
32.9%
-14.52
0.00
-100.0%
0.6%
-26.67
0.00
-100.0%
0.2%
-27.49
0.00
-100.0%
0.1%
-26.86
0.00
-100.0%
0.1%
-25.27
0.00
-100.0%
0.3%
-27.75
0.00
-100.0%
0.1%
-26.87
0.00
-100.0%
0.1%
-28.19
0.00
-100.0%
0.1%
41.02
41.02
29019.0%
29019.0%
18.13
26.67
3246.8%
3246.8%
17.37
27.49
2531.7%
2531.7%
17.37
26.86
2978.2%
2978.2%
18.27
25.27
3211.1%
3211.1%
17.25
27.75
2752.4%
2752.4%
17.25
26.87
3376.8%
3376.8%
17.31
28.19
2619.1%
2619.1%
55.54
41.02
29119.0%
29018.4%
44.80
26.67
3346.8%
3246.6%
44.86
27.49
2631.7%
2531.7%
44.23
26.86
3078.2%
2978.1%
43.54
25.27
3311.1%
3210.8%
45.00
27.75
2852.4%
2752.4%
44.11
26.87
3476.8%
3376.8%
45.49
28.19
2719.1%
2619.1%
10.31
8.66
2615.2%
2613.5%
7.84
5.66
294.2%
286.0%
7.56
5.61
243.4%
233.8%
7.42
5.54
270.7%
262.4%
7.78
5.60
293.8%
285.5%
7.50
5.50
263.9%
254.4%
7.40
5.46
306.4%
298.4%
7.44
5.48
256.7%
247.3%
Abt Associates Inc
Appendix
A-6
-------
Table A-6: Comparison of Interpolation Methods for the Western U.S. Excluding California
(W126)
Interpolation
CMAQ
HMD
HMD VNA 100
HMD VNA 50
HMH
HMH VNA 100
HMH VNA 50
VNA
Measure
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Bias
Error
Norm. Bias
Norm. Error
Count Mean
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
136
5.65
6.63
100.6%
106.1%
-0.77
3.41
6.1%
37.2%
-1.17
3.43
5.2%
37.7%
-1.11
3.36
4.9%
37.1%
-0.51
3.48
8.2%
37.8%
-1.14
3.44
5.2%
37.4%
-1.03
3.40
5.4%
36.9%
-1.21
3.47
4.9%
37.6%
Median Min
4.60
5.42
49.3%
50.1%
-0.63
2.32
-5.3%
24.0%
-0.61
2.43
-9.3%
23.8%
-0.66
2.30
-9.3%
23.9%
-0.25
2.52
-3.3%
22.7%
-0.63
2.38
-8.1%
23.5%
-0.63
2.30
-8.1%
23.2%
-0.68
2.56
-6.4%
27.7%
Max Range STD
-6.79
0.03
-68.4%
0.9%
-13.51
0.02
-89.3%
0.2%
-14.30
0.01
-89.3%
0.1%
-13.97
0.01
-89.3%
0.1%
-12.90
0.00
-90.2%
0.1%
-14.34
0.00
-90.2%
0.0%
-13.88
0.00
-90.2%
0.0%
-14.80
0.00
-76.2%
0.1%
26.40
26.40
1120.0%
1120.0%
11.39
13.51
352.8%
352.8%
11.34
14.30
352.8%
352.8%
11.34
13.97
352.8%
352.8%
12.42
12.90
353.5%
353.5%
11.34
14.34
353.5%
353.5%
11.34
13.88
353.5%
353.5%
11.33
14.80
297.9%
297.9%
33.19
26.38
1188.5%
1119.1%
24.90
13.49
442.1%
352.6%
25.63
14.28
442.1%
352.7%
25.30
13.95
442.1%
352.7%
25.32
12.90
443.7%
353.4%
25.68
14.34
443.7%
353.5%
25.22
13.88
443.7%
353.5%
26.13
14.80
374.1%
297.9%
6.52
5.52
176.0%
172.8%
4.57
3.14
62.8%
51.0%
4.49
3.13
63.7%
51.6%
4.42
3.07
62.5%
50.5%
4.75
3.26
63.2%
51.3%
4.50
3.12
62.1%
49.9%
4.48
3.09
61.2%
49.1%
4.48
3.08
58.8%
45.5%
Abt Associates Inc
Appendix
A-7
-------
-------
Appendix B Memorandum Regarding the Quadratic
Rollback
The following is an excerpt from a memorandum entitled "A Comparison between Different Rollback
methodologies Applied to Ambient Ozone Concentrations". The included section describes in detail the
mathematics behind the 8-hour maximum rollback method. The first page of the memorandum was
included for reference purposes. The relevant section begins on the second page at Quadratic Rollback
and continues until Data.
Abt Associates Inc Appendix B-l
-------
UNITED STAT1S ENVIRONMENTAL PROTECTION AGINCY
Offlc* of Air Qyallty Pl«nr»tnfl ant* St«n*ir«l* JOAQPS)
Research Triangle Park, North Carolina 27711
MEMORANDUM
SUBJECT; A Comparison between Different Rollback Methodologies Applied
to Ambienl Cteone Concentrations
FROM; Michael Rizzo, EPA-OAQPS, Air Quality Data Analysis Group
TO: Ozone N AAQS Rev lew Docket (O AR-2QOS-0172)
DATE: November 7, 2005
For the prior ozone N AAQS review, one of the methods, referred to as the "Quadratic
Method," developed for adjusting ozone ambient air concentrations to simulate just
meeting alternative sJandsrds combined berth linear and quadratic elements to reduce
larger concentrations more than smaller ones.1 In (his regard, the Quadratic method
attempts to account for reductions in emissions without greatly affecting lower
concentrations near ambient background levels. Other rollback algorithms have either fit
the data to a particylar distribution such as the Weibull method or used a linear,
proportional rollback where all of the ambient measurements are reduced equally
regardle&s of their individual magnitudes.2
This memorandum will compare two of the above mentioned rollback methodologies:
quadratic and pereentile proportional. As the name implies, the quadratic method uses a
quadratic equation to reduce higher ozone concentrations mace ihjm smaller ones. The
amount of rollback depends on the magnitude of the reduction of the existing fourth
maximum to meet the standard. Sites which have data with high ozone concentrations
are subjected to a more substantial rollback than those which are at or below the National
Ambient Air Quality Standards, In contrast, the perccntiJe proportional rollback uses a
dual linear approach where ozone concentrations less than a. specified percentiie are not
rolled back while those greater thin the pcrceislile value are proportionally rolled back
haswl upon the difference between the measured fourth maximum and the calculated
value needed to attain the standard.
Ehiff, Marcus; Herat, Robert L.; Johnson. Ted BL; "QuaAatk Rollback: A T«Jinijqae to Modd
Aitibicnt Coneentotiwis Due u> Undefined Emission Controls"- fifet NV. 98-TA32.07 Air aai
Waste ManagemciM Animal Meeting, Saa Diego. California hme 14*18. 1998.
ra, Ted; A Clriikte 10 Seieeted Alfeofithms. Dis£fa¥n,lioiMt. &rad Databases Used ifl E.J
Models Pcvclonerf by the Office of Air Quality naimmy and Stttidanfa, EPA Grant No,
CRB27033. MayHHE. |]i!i?;i>sm,!»».s:Mfcl»^^
Abt Associates Inc Appendix B-2
-------
Methodology
Quadratic Rollback
The Quadratic Rollback method takes the form of the following equation
|_^ . — f .Lr .
J J J
where : C'j is the rolled back concentration
tj is the quadratic rollback factor unique to each measurement
Q is the original measured concentration
The quadratic rollback factor is defined as
J J
where: V and B are positive constants determined from an individual site's
measurements
In order to calculate the V and B constants, other parameters must be known first. These
are listed below:
where: Ij is the average concentration for time period \ which, in the case of the
current ozone standard, refers to an 8 hour time period so lt is an 8 hour
average
Cj is the original measured concentration
Nj is the number of hours for time period i
J, =max(Cy),€,
where: Jj is the maximum one hoar value for time period i which, in mis case, is
the maximum 1 hour value in the 8 hour time period
where: Qj is the average of the squared concentrations for time period i
J - max( J.)
where: J is the maximum of all time periods, i=l, 2, 3,..... etc. and is also the
maximum value over the entire length of time analyzed
-------
For each time period i, a transformation factor, 3(k is Computed psing the above.
parameters:
The appropriate X, which corresponds to the metric corresponding the standard thai the
data being rolled back to is compared to two limes Utt^ffuJucLof lie maximum one: horn
value (J) and a standard concentration level (S). fSFexample, if tie current ozone ^-- -^
-standard Is Being examined, The fourth 'Kfpffist X, from all of the time periods is used.
The standard concentration value can be the actual value to meet the NAAQS such as
0,084 ppm for the S hour ozone slandarxl. In this case, S refera to the target value of the
41'1 niaximnm 8 hour average for 2004 whose calculation is described below. A metric is
calculate! of the following form to determine the final wpaiion used for the rollback
calculation:
V =
X
The B coefficient is ealeulafcd for each concentration for lime period i as;
*,=
-------
If V is between 0 and 1. this is the mixed linear-quadratic case, The rollback equation
then becomes;
where;
Data
Im and Qn refer to the I and Q quantities for period i which refers to the
metric of the slandari
Data from eight sites within three major urban areas were used to calculate rolled back
ozone concentrations using the quadratic method. For each site, a high and low ozone
year was chosen based on historical data- The information detailing fte high and tow
years and the corresponding eight horn1 averaged 4* maxima are provided in Table I.
Table 1: List of Sites and Respective High/Low Ozone Years Utilized
Site ID
060371601
060372005
06065800!
060? 1 10(M
1 71630010
291831002
420170012
420450002
__Hjj|h ozone
Year
1994
1994
1994
1994
1995
1995
1995
1995
8 hour
4ih max
127
132
148
14S
84
112
111
108
Low ozone
Year
2002
2001
2000
2002
2001
2001
2003
2003
8 hour
4th max
74
90
106
105
78
85
87
80
City
Los Angeles
Los A,ngcjes_
Ltn Angeles
Los Angeles
Si 1 ouis
St. Louis
Philadelphia
Philadelphia
The ozone concentrations measured in the high year were rolled back to the low year
concentrations based on the differences in the 4 maxima.
The value to roll the fourth maximum 8 hour ozone values back to for each monitor-year
is denoted as S, The value of S for each monitor-year is determined by the amount of
rollback required to have the average fourth maximum name concentration over three
years attain the standard. To accomplish this, the design value for each site is multiplied
by a reduction factor calculated as:
Reduction Factor =
100
where: C,B is the attainmesit concentration for the ozone standard which is 0.084
ppm
Qh is ihc average of the fourth maxima at the design value monitor for a
particular area over a three year perkxl
Abt Associates Inc
Appendix
B-5
-------
-------
Appendix C Numerical Examples of VNA and eVNA
Below are numerical examples of VNA and eVNA. Note that the examples are given for the year 1995;
however, the same approach holds for any given year.
Numerical Example VNA
The first step in VNA is to identify the set of nearest monitors for each of the points of interest, such as
the centers of the grid-cells in a modeling domain. The figure below presents nine grid-cells and seven
monitors, with the focus on identifying the set of nearest neighbors to grid-cell "E."
#= Center Grid-Cell "E"
*
= Air Pollution Monitor
VNA identifies the nearest monitors, or "neighbors," by drawing a polygon, or "Voronoi" cell, around the
center of each grid-cell. These polygons have the special property that the boundaries are the same
distance from the two closest points.
# = Center Grid-Cell "E"
*
= Air Pollution Monitor
We then choose those monitors that share a boundary with the center of grid-cell "E." These are the
nearest-neighbors, which we use to estimate the air pollution level for this grid-cell.
Abt Associates Inc
Appendix
C-l
-------
# = Center Grid-Cell "E"
*
= Air Pollution Monitor
To estimate the air pollution level in each grid-cell, we calculate the annual and the binned daily metrics
for each of the neighboring monitors, and then calculate an inverse-distance weighted average of the
metrics. The further the monitor is from the grid-cell, the smaller the weight.
A
D
Monitor: *
1995 80 ppb
10 miles
G
*
B
Monitor:
1995 90 ppb ^
15 miles x
A
— -# - —
/ H
*
Monitor:
1995 100 ppb
20 miles
C
*
F
*
Monitor:
1995 60 ppb
15 miles
I
*
#= Center Grid-Cell "E"
*
= Air Pollution Monitor
The weight for the monitor 10 miles from the center of grid-cell E is calculated as follows:
J_
To
J_ J_ J_ _L
10 + 15 + 15 + 20
= 0.35 .
The weights for the other monitors would be calculated in a similar fashion. We then calculate an
inverse-distance weighted average for 1995 air pollution levels in grid-cell E as follows:
Forecast1995 = 0.35*80 ppb + 0.24*90 ppb+ 0.24*60 ppb + 0.18*100 ppb = 81.2 ppb .
Numerical Example eVNA
Abt Associates Inc
Appendix
C-2
-------
We also use VNA in combination with modeling data; we term this enhanced Voronoi Neighbor
Averaging (eVNA). For each of the neighbor monitors, we multiply the monitoring data with the ratio of
the base-year modeling data for the destination grid-cell to the base-year modeling data for grid-cell
containing the monitor.
Consider the example in the figure below. To forecast air pollution levels for 1995, we would multiply
the 1995 monitor value by the ratio of the 1995 model value in grid-cell E to the 1995 model value
^ ModelE 1995
Forecast 1995 = > Weight\ * Monitori * —
i=i Moaelil995
containing each of the neighbor monitors:
85^| ( 85^| ( 85
Forecastl995 = { 0.35*80*—J + [0.24*90*—J + [0.24*60*—J + ^0.18*10O* — ] = JO.Sppb
Abt Associates Inc Appendix C-3
-------
A
Model: B
1995 100ppb
Monitor:
1995 90 ppb ^
15 miles
Model: D
1995 95ppb
Monitor: *
1995 80ppb
10 miles
Model: /E
1995 85/ppb
/
-# —
Model: F
1995 80 ppb
Monitor:
1995 60 ppb
15 miles
Model: H
1995 120 ppb
*
Monitor:
1995 100 ppb
20 miles
# = Center Grid-Cell "E"
*
= Air Pollution Monitor
Abt Associates Inc
Appendix
C-4
-------
Appendix D Air Quality Maps
In this Appendix, we present O3 air quality maps under the four rollback scenarios described above, and
under "as is" conditions as described by the POES. O3 levels are shows in terms of the Maximum 3-
month 12-hour W126 metric. The maps show the entire continental U.S. as a whole, even though the data
for the East and Western U.S. were generated separately according to slightly different interpolation
methods.
Abt Associates Inc Appendix D-l
-------
o
g
W126 (ppm-h)
I I <^7 (min = 0.5}
I |731 (max - €2)
-------
o.
e
i>>
9
o
9
fB
fB
O.
n
3
g
3
e
3
oo
era
fB
&5
rt-
fB
O
n
o
VI
00
W126 (ppm-h)
I |<^7 (min
I |7
-------
o
g
O.
e
n
n
O.
n
3
g
3
e
3
oo
o.
CTQ
o
n
W126 (ppm-h)
I J <-7 (min • 0.5)
I l7-=W126<=13
---21 (max = 23)
-------
.
I?
2
O.
o.
U)
I
ff
05
W126 (ppm-h)
I J <-7 (min « 0.5)
I l7-=W126<=13
(max « 21. 3}
(71
-------
o.
I?
2
O.
U)
3
o
in
W126 (ppm-h)
I I <=7 (min = 0.5)
1>13 (max =15)
(71
-------
Appendix E Interpolating State Growing Seasons
Map E-l: U.S. Continental Climatic Classification
U.S. Continental Climatic Zones
Yearly High Temperature (Celsius)
I I 10
Source: ESRI Annual World Temperature Zones, 2005.
Map E-2: U.S. States Climatic Classification
Average Yearly High Temperature (Celsius)
6-10
| | 10-15
^H 15-20
^H >20
Source: Author Estimates. This classification was used to extrapolate typical state growing seasons when
primary source data was not available.
Abt Associates Inc
Appendix
E-l
-------
-------
Appendix F Growing Seasons for Major Crops by
State
11 28
10 II
AI. AR A7 CO DE OA IL KK KY LA MD MO MS NC NE MM OK Sf SD TN TX \'A
Figure F-l: Typical Harvest Seasons for Sorghum by State
Figure F-2: Typical Harvest Seasons for Cotton by State
Abt Associates Inc
Appendix F-l
-------
Figure F-3: Typical Harvest Seasons for Soybean by State
Figure F-4: Typical Harvest Seasons for Winter Wheat by State
Abt Associates Inc
Appendix F-2
-------
Appendix G Summary Statistics for Crop Yield
Concentration Responses based on W126 Metric
Appendix G summarizes relative crop yield gains under the six scenarios presented in Section 4:
Ozone levels which just meet an 84 ppb 4th highest 8-hour maximum standard (in this case, we
performed a rollback to 84.999 ppb, since the 4th highest 8-hour maximum standard truncates
decimals)
Ozone levels which just meet a 70 ppb 4th highest 8-hour maximum standard (actually 70.999 for
the reasons above)
Ozone levels which just meet a 25ppm-h 3-month 12-hour SUM06 standard
Ozone levels which just meet a 15ppm-h 3-month 12-hour SUM06 standard.
Ozone levels which just meet a 21ppm-h 3-month 12-hour W126 standard.
Ozone levels which just meet a 13ppm-h 3-month 12-hour W126 standard.
Appendix G. 1 presents results based on the reduced or "adjusted" O3 metric (i.e. hourly O3 values were
adjusted down by 10%). Appendix G.2 presents similar results based on the unreduced metric (unadjusted
hourly O3 values). Refer to Section 5 for a discussion of the data and method used to estimate O3
concentration responses on vegetation.
G.I Crop Yield Concentration Responses based on Ten Percent Adjusted W126
Metric
The box plots and tables in this section show yield responses over the continental U.S. at the county level.
Results were computed at the gridcell level, and subsequently averaged at the county level. The national
estimates presented below are simple averages across counties. The crop ranges used to compute national
estimates are the ones reported in the 2002 Census of Agriculture. The first box plot shows percent yield
loss at the baseline exposure level. The subsequent plots show percent yield gain from the baseline
situation. The shaded boxes represent the lowest and highest quartiles; the low and high bars show
minimum and maximum values. Exact numbers are presented in tables. Note that the minimums and
maximums reported on the maps will differ from the summary results presented in the tables because the
maps are based on individual gridcell results whereas the results in the tables and box plots are county
level averages.
Abt Associates Inc Appendix G-l
-------
G.I.I Median Concentration Responses for Crops
VI
_0
30%
25%
15% - -
10% - -
5% - - 3
0
18.1%
12.6%
1.0%
7.7%
0.2%
5.4%
17.0%
1.4%
4> s-SSr xO^
-------
Figure G-2: Yield Gains from Baseline for Crops - 84 ppb Rollback (reduced W126, Median C-R)
Figure G-3: Yield Gains from Baseline for Crops - 70 ppb Rollback (reduced W126, Median C-R)
Abt Associates Inc
Appendix G-3
-------
s
•a
60
T3
T3
1)
4% - - ^-
0.9%
4.0%
1.3
0°
A* ,^
P^ ^
Figure G-4: Yield Gains from Baseline for Crops - 25 ppm-hr Rollback (reduced W126, Median C-
12%
•1
60
T3 4% - -
10.7%
7.0%
2.60/
1.4%
xf ./ ^
• . *
-------
1 90^
i no/.
en
0
^3 oo/
a
1
9%
n% -
3.4
o/
5.S
io/
/O
0.0%
8.4
o/
0.3%
1
O.i
'O/
3 /O
?q
%
0.2%
5.E
o/
4.3% 4.2%
, 3.4%
2.0%
L2% 4,2%
0°
^ &
xx 4
Figure G-6: Yield Gains from Baseline for Crops - 21 ppm-hr W126 Rollback (reduced W126,
Median C-R)
en
en
2
§
T3
s
4% -
n% -
3.'
o/
8.3
r
i
0.0%
10.
r
4%
1
0.3%
O.S
10.
4%
0.2%
%
;,<-
^
%
n
5.t
2.4
O/
6.C
4.9%
1.3%
-i rn
i
.
Figure G-7: Yield Gains from Baseline for Crops - 13 ppm-hr W126 Rollback (reduced W126,
Median C-R)
Abt Associates Inc
Appendix G-5
-------
G.1.2 Summary Statistics of Concentration Responses for Crops
Similar results as in Chapter 5.4.1 are provided in Tables G-l through G-5 below. These yield change
estimates are based on the reduced W126 metric (i.e. hourly O3 levels have been reduced by 10% to
account for the elevation differential between monitor height and crop canopy). In Tables G-l through G-
5 rollback results are presented as differences or yield gains from baseline. Tables G-6 through G-l 1
present absolute losses instead. Results for Soybean and Cotton are presented in Chapter 5.4.1.
Tables show count, mean, maximum, minimum, median, and standard deviation in yield responses for
thirteen commodities and six alternate air quality scenarios. These are based on straight average yield
responses at the county-level (no production weights were applied).
Table G-l: Yield Losses for Crops, Fruits and Vegetables - 2001 Baseline (reduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Sorghum
Cantaloupe
Corn
Onion
Peanut
Valencia Orange
Tomato Processing
Winter Wheat
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
891
409
89
2,236
2,533
Max
3.78%
25.75%
23.50%
21.26%
0.00%
18.47%
12.62%
9.86%
18.11%
0.97%
23.49%
0.42%
0.17%
0.02%
8.07%
5.45%
16.95%
19.17%
13.78%
12.78%
29.20%
1.40%
1.11%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.33%
0.08%
0.05%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.57%
0.01%
0.72%
1.35%
0.97%
0.90%
0.44%
0.00%
0.00%
Mean
0.08%
11.47%
10.47%
9.47%
0.00%
5.31%
2.67%
1.93%
3.84%
0.04%
7.53%
0.01%
0.00%
0.00%
3.72%
0.41%
6.52%
9.30%
6.68%
6.20%
8.77%
0.10%
0.10%
Median
0.00%
11.90%
10.87%
9.83%
0.00%
5.28%
2.55%
1.81%
0.53%
0.03%
8.82%
0.00%
0.00%
0.00%
3.83%
0.27%
5.39%
9.53%
6.85%
6.35%
7.55%
0.04%
0.05%
STD
0.27%
3.27%
2.99%
2.70%
0.00%
2.62%
1.61%
1.21%
5.33%
0.05%
5.09%
0.03%
0.01%
0.00%
1.05%
0.47%
3.86%
2.19%
1.57%
1.46%
5.65%
0.14%
0.13%
Table G-2: Yield Gains from Baseline for Crops, Fruits and Vegetables - 84 ppb 8-Hour Rollback
(reduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
TV
587
1,883
678
1,585
300
Max
1.97%
3.25%
2.97%
2.68%
0.00%
4.87%
3.96%
3.23%
2.38%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.02%
0.12%
0.11%
0.10%
0.00%
0.20%
0.14%
0.11%
0.02%
Median
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
STD
0.14%
0.41%
0.37%
0.34%
0.00%
0.64%
0.46%
0.36%
0.16%
Abt Associates Inc
Ann
endix
G-6
-------
Grain Sorghum
Cantaloupe
Corn
Onion
Peanut
Valencia Orange
Tomato Processing
Winter Wheat
Table G-3: Yield Gains
(reduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Onion
Peanut
Valencia Orange
Tomato Processing
Winter Wheat
Median
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
from Baseline
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Table G-4: Yield Gains from Baseline
Rollback (reduced W126)
Crop
Kidney Beans
Grapes
Response
Median
Max
Median
Min
1,331
1,610
2,623
891
409
89
2,236
2,533
for Crops,
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
891
409
89
2,236
2,533
for Crops,
TV
587
1,883
0.46%
4.44%
0.23%
0.10%
0.01%
1.13%
2.30%
1.85%
2.69%
1.94%
1.80%
9.68%
0.82%
0.60%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Fruits and Vegetables
Max
3.52%
7.49%
6.84%
6.18%
0.00%
11.88%
9.25%
7.42%
6.72%
0.87%
8.72%
0.39%
0.16%
0.02%
2.34%
4.72%
4.93%
5.57%
4.00%
3.71%
18.62%
1.32%
1.02%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Fruits and Vegetables
Max
3.54%
7.52%
6.87%
6.21%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.17%
0.00%
0.00%
0.00%
0.06%
0.02%
0.13%
0.09%
0.06%
0.06%
0.23%
0.01%
0.01%
- 70 ppb
Mean
0.06%
1.52%
1.39%
1.25%
0.00%
1.87%
1.16%
0.88%
0.76%
0.02%
1.75%
0.01%
0.00%
0.00%
0.45%
0.21%
0.92%
1.11%
0.80%
0.74%
2.83%
0.07%
0.07%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.02%
0.57%
0.01%
0.01%
0.00%
0.17%
0.14%
0.35%
0.31%
0.23%
0.21%
0.98%
0.06%
0.05%
8-Hour Rollback
Median
0.00%
1.28%
1.16%
1.05%
0.00%
1.27%
0.73%
0.54%
0.10%
0.01%
1.32%
0.00%
0.00%
0.00%
0.29%
0.09%
0.52%
0.87%
0.63%
0.58%
1.34%
0.01%
0.01%
STD
0.25%
1.41%
1.29%
1.16%
0.00%
2.00%
1.30%
0.99%
1.15%
0.05%
1.88%
0.02%
0.01%
0.00%
0.50%
0.38%
1.14%
1.08%
0.77%
0.72%
3.52%
0.13%
0.11%
- 25 ppm-hr SUM06
Mean
0.03%
0.13%
0.12%
0.10%
Median
0.00%
0.00%
0.00%
0.00%
STD
0.24%
0.47%
0.43%
0.39%
Abt Associates Inc
Appendix
G-7
-------
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Onion
Peanut
Valencia Orange
Tomato Processing
Winter Wheat
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Table G-5: Yield Gains from Baseline
Rollback (reduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
678
1,585
300
1,331
1,610
2,623
891
409
89
2,236
2,533
for Crops,
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
891
409
2,090
89
2,236
2,533
0.00%
12.04%
9.36%
7.50%
6.87%
0.88%
8.60%
0.40%
0.16%
0.02%
2.36%
4.61%
4.95%
5.61%
4.03%
3.74%
18.48%
1.29%
0.99%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Fruits and Vegetables -
Max
3.69%
9.35%
8.53%
7.72%
0.00%
14.04%
10.59%
8.42%
8.32%
0.92%
10.71%
0.41%
0.16%
0.02%
2.93%
5.07%
5.03%
2.58%
0.97%
6.16%
6.97%
5.01%
4.65%
21.73%
1.36%
1.06%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.19%
0.13%
0.10%
0.06%
0.01%
0.16%
0.00%
0.00%
0.00%
0.05%
0.05%
0.38%
0.09%
0.06%
0.06%
0.23%
0.01%
0.01%
15 ppm-hr
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
SUM06
Mean Median
0.05%
1.01%
0.92%
0.83%
0.00%
1.26%
0.81%
0.62%
0.61%
0.02%
1.24%
0.01%
0.00%
0.00%
0.27%
0.18%
0.61%
0.18%
0.11%
0.72%
0.73%
0.52%
0.48%
2.02%
0.06%
0.05%
0.00%
0.43%
0.39%
0.36%
0.00%
0.24%
0.14%
0.11%
0.01%
0.00%
0.37%
0.00%
0.00%
0.00%
0.00%
0.01%
0.19%
0.04%
0.03%
0.00%
0.24%
0.17%
0.16%
0.27%
0.00%
0.00%
0.00%
0.76%
0.56%
0.44%
0.46%
0.04%
0.59%
0.02%
0.01%
0.00%
0.19%
0.29%
0.95%
0.34%
0.24%
0.22%
1.05%
0.07%
0.05%
STD
0.26%
.23%
.12%
.02%
0.00%
.76%
.18%
0.90%
1.29%
0.05%
1.58%
0.02%
0.01%
0.00%
0.43%
0.41%
0.82%
0.32%
0.15%
1.39%
0.93%
0.67%
0.62%
3.00%
0.12%
0.10%
Abt Associates Inc
Appendix
G-8
-------
Table G-6: Yield Gains from Baseline for Crops, Fruits and Vegetables - 21 ppm-hr W126
Rollback (reduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Sorghum
Cantaloupe
Corn
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
Table G-7: Yield Gains from Baseline
Rollback (reduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
891
409
2,090
89
2,236
2,533
for Crops,
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
Max
3.38%
6.34%
5.79%
5.24%
0.00%
10.50%
8.35%
6.73%
0.26%
0.82%
7.25%
0.38%
0.16%
0.02%
1.99%
4.29%
1.98%
1.25%
0.38%
4.18%
4.74%
3.40%
3.16%
16.08%
1.21%
0.92%
Fruits and
Max
3.67%
9.11%
8.32%
7.53%
0.00%
13.80%
10.45%
8.32%
0.35%
0.92%
10.43%
0.41%
0.16%
0.02%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Vegetables
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.02%
0.06%
0.05%
0.05%
0.00%
0.09%
0.07%
0.06%
0.00%
0.00%
0.07%
0.00%
0.00%
0.00%
0.03%
0.03%
0.03%
0.01%
0.01%
0.29%
0.04%
0.03%
0.03%
0.11%
0.01%
0.01%
Median
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
STD
0.22%
0.33%
0.30%
0.28%
0.00%
0.56%
0.43%
0.35%
0.02%
0.03%
0.43%
0.01%
0.01%
0.00%
0.14%
0.25%
0.16%
0.09%
0.03%
0.78%
0.24%
0.17%
0.16%
0.76%
0.06%
0.04%
- 13 ppm-hr W126
Mean
0.04%
0.73%
0.67%
0.61%
0.00%
0.96%
0.63%
0.49%
0.04%
0.02%
0.93%
0.01%
0.00%
0.00%
Median
0.00%
0.09%
0.08%
0.08%
0.00%
0.02%
0.01%
0.01%
0.00%
0.00%
0.05%
0.00%
0.00%
0.00%
STD
0.25%
1.04%
0.95%
0.86%
0.00%
1.53%
1.04%
0.81%
0.06%
0.05%
1.33%
0.02%
0.01%
0.00%
Abt Associates Inc
Appendix
G-9
-------
Onion
Median
891
2.86% 0.00% 0.20%
0.00% 0.37%
Peanut
Valencia Orange
Tomato Processing
Winter Wheat
Table G-8: Absolute
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Table G-9: Absolute
Crop
Kidney Beans
Grapes
Lettuce
Median
Median
Max
Median
Min
Max
Median
Min
Crop Yield Losses
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
Crop Yield Losses
Response
Median
Max
Median
Min
Median
409
89
2,236
2,533
- 84 ppb
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
- 70 ppb
TV
587
1,883
678
5.02%
6.00%
6.79%
4.88%
4.53%
21.30%
1.35%
1.05%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.14%
0.66%
0.52%
0.38%
0.35%
1.47%
0.05%
0.04%
0.00%
0.00%
0.02%
0.02%
0.02%
0.00%
0.00%
0.00%
0.38%
1.32%
0.78%
0.56%
0.52%
2.50%
0.11%
0.10%
8-Hour Rollback (reduced W126)
Max
1.87%
23.44%
21.40%
19.35%
0.00%
14.42%
9.11%
6.96%
17.79%
0.51%
20.82%
0.18%
0.06%
0.00%
21.84%
4.83%
8.44%
7.34%
3.15%
6.94%
2.61%
1.20%
15.43%
17.44%
12.53%
11.62%
23.00%
0.66%
0.58%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.33%
0.08%
0.05%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.86%
0.04%
0.11%
0.57%
0.01%
0.26%
0.01%
0.03%
0.72%
1.35%
0.97%
0.90%
0.44%
0.00%
0.00%
Mean
0.06%
11.35%
10.36%
9.37%
0.00%
5.11%
2.53%
1.82%
3.83%
0.03%
7.35%
0.01%
0.00%
0.00%
5.53%
0.74%
1.33%
3.66%
0.39%
2.34%
0.49%
0.35%
6.39%
9.21%
6.62%
6.14%
8.55%
0.09%
0.09%
Median
0.00%
11.73%
10.71%
9.69%
0.00%
5.17%
2.49%
1.77%
0.51%
0.03%
8.57%
0.00%
0.00%
0.00%
5.14%
0.65%
1.14%
3.78%
0.27%
2.23%
0.40%
0.33%
5.39%
9.45%
6.79%
6.30%
7.53%
0.04%
0.05%
STD
0.15%
3.21%
2.93%
2.65%
0.00%
2.33%
1.38%
1.03%
5.32%
0.04%
4.94%
0.01%
0.00%
0.00%
2.57%
0.53%
0.87%
1.00%
0.37%
1.28%
0.39%
0.22%
3.62%
2.12%
1.53%
1.41%
5.26%
0.10%
0.10%
8-Hour Rollback (reduced W126)
Max
0.27%
19.21%
17.53%
15.86%
0.00%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.02%
9.96%
9.09%
8.22%
0.00%
Median
0.00%
10.05%
9.17%
8.30%
0.00%
STD
0.04%
2.84%
2.59%
2.34%
0.00%
Abt Associates Inc
Appendix
G-10
-------
Potato
Rice
Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Table G-10: Absolute
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
Crop Yield Losses
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
6.59%
3.37%
2.44%
14.46%
0.10%
16.26%
0.05%
0.01%
0.00%
9.69%
1.45%
2.71%
5.97%
0.73%
4.10%
1.09%
0.65%
12.10%
14.19%
10.20%
9.46%
12.83%
0.14%
0.15%
- 25 ppm-hr SUM06
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
Max
0.65%
19.74%
18.02%
16.30%
0.00%
11.01%
6.45%
4.82%
17.04%
0.14%
16.02%
0.10%
0.03%
0.00%
11.44%
2.81%
3.39%
5.98%
1.25%
5.99%
2.09%
1.01%
0.33%
0.08%
0.05%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.86%
0.04%
0.11%
0.57%
0.01%
0.26%
0.01%
0.03%
0.72%
1.35%
0.97%
0.90%
0.44%
0.00%
0.00%
Rollback
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.33%
0.08%
0.05%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.86%
0.04%
0.11%
0.57%
0.01%
0.26%
0.01%
0.03%
3.44%
1.51%
1.05%
3.09%
0.02%
5.78%
0.00%
0.00%
0.00%
3.98%
0.46%
0.83%
3.27%
0.20%
1.63%
0.29%
0.23%
5.59%
8.19%
5.89%
5.46%
5.94%
0.03%
0.04%
3.64%
1.59%
1.11%
0.37%
0.01%
6.63%
0.00%
0.00%
0.00%
3.96%
0.40%
0.81%
3.24%
0.17%
1.58%
0.26%
0.22%
5.19%
8.35%
6.00%
5.57%
5.71%
0.02%
0.03%
1.10%
0.58%
0.42%
4.45%
0.01%
4.16%
0.00%
0.00%
0.00%
1.37%
0.25%
0.38%
0.86%
0.14%
0.77%
0.17%
0.12%
2.94%
1.79%
1.29%
1.20%
2.89%
0.03%
0.03%
(reduced W126)
Mean
0.05%
11.35%
10.36%
9.37%
0.00%
5.12%
2.54%
1.83%
3.78%
0.03%
7.37%
0.01%
0.00%
0.00%
5.36%
0.70%
1.26%
3.67%
0.36%
2.33%
0.49%
0.35%
Median
0.00%
11.82%
10.79%
9.76%
0.00%
5.27%
2.54%
1.81%
0.48%
0.03%
8.82%
0.00%
0.00%
0.00%
5.14%
0.64%
1.14%
3.83%
0.27%
2.24%
0.40%
0.33%
STD
0.09%
3.12%
2.85%
2.58%
0.00%
2.31%
1.36%
1.02%
5.30%
0.03%
4.89%
0.01%
0.00%
0.00%
2.16%
0.41%
0.67%
0.97%
0.27%
1.24%
0.40%
0.21%
Abt Associates Inc
Appendix
G-ll
-------
Valencia Orange
Median
89
12.30% 0.72%
6.14%
5.39% 3.18%
Tomato Processing
Winter Wheat
Table G-ll: Absolute
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Max
Median
Min
Max
Median
Min
Crop Yield Losses
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
2,236
2,533
14.21%
10.21%
9.47%
24.66%
0.79%
0.68%
- 15 ppm-hr SUM06
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
Max
0.37%
18.05%
16.48%
14.91%
0.00%
7.51%
3.96%
2.89%
13.48%
0.07%
14.24%
0.04%
0.01%
0.00%
6.89%
1.42%
1.70%
5.47%
0.60%
3.95%
1.55%
0.63%
11.14%
13.03%
9.36%
8.68%
14.60%
0.20%
0.20%
1.35%
0.97%
0.90%
0.44%
0.00%
0.00%
Rollback
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.33%
0.08%
0.05%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.86%
0.04%
0.11%
0.57%
0.01%
0.26%
0.01%
0.03%
0.72%
1.35%
0.97%
0.90%
0.44%
0.00%
0.00%
9.21%
6.62%
6.14%
8.54%
0.09%
0.09%
9.53%
6.85%
6.35%
7.49%
0.04%
0.05%
2.07%
1.49%
1.38%
5.28%
0.11%
0.10%
(reduced W126)
Mean
0.03%
10.46%
9.55%
8.64%
0.00%
4.05%
1.86%
1.31%
3.24%
0.02%
6.28%
0.00%
0.00%
0.00%
4.29%
0.52%
0.91%
3.45%
0.23%
1.80%
0.34%
0.26%
5.80%
8.57%
6.16%
5.71%
6.75%
0.04%
0.05%
Median
0.00%
10.79%
9.85%
8.91%
0.00%
4.41%
2.03%
1.43%
0.41%
0.02%
7.83%
0.00%
0.00%
0.00%
4.48%
0.46%
0.95%
3.58%
0.22%
1.83%
0.29%
0.26%
5.39%
8.90%
6.39%
5.93%
6.89%
0.03%
0.04%
STD
0.05%
2.68%
2.44%
2.21%
0.00%
1.36%
0.73%
0.53%
4.41%
0.01%
4.19%
0.00%
0.00%
0.00%
1.16%
0.29%
0.32%
0.82%
0.12%
0.77%
0.23%
0.12%
2.74%
1.67%
1.20%
1.11%
3.49%
0.04%
0.04%
G.1.3 Yield Response Maps for Selected Field Crops based on Ten Percent Adjusted W126
Metric
Maps for cotton and soybean based on the reduced O3 metric are presented in the body of the document
under Section 5.
Abt Associates Inc
Appendix
G-12
-------
o.
ft O
s 3
Ss O
| ||
s'Sr
£L ^
5T o
S> 3
^N
|g
ft ^
< ^
ga
»g
^ s
I*
cfS
i ; §i r' -±2
8 -'V s (
i — -i
; •' ?; . •; ^
y
•
Yield (% loss)
I 10 <= Yield <= 0.17
No production value reported
'. - .••^at^V_
r-t\\\'~j '., ','
/. /-a —- 2V is /?*-*&Ci <&?>(•:•?<*
, -\ '%.-, i >,,p• '
cp v^r fe
i^4 :&t***^s~^im
o
o
-a
o
1
VI
O
-------
P
TS
O
s
CL
o1
"S
3
I
e
3
CL
e
fB
CL
;\
j {
5 ^M^
;
4
:• .' ,;- -"•
/r\
c.
Yield (% loss)
^J <=1 (min = 0)
I |1 < Yield <=2
I |>2 (max = 3.2)
No producticn value reported
fB
a
8?
O
fB
-------
I
1
o.
I 9
00 w
.
3*5,
o a
oo P
*> 3
•C Ct
T3 S
a1 n
=
8.1
C
3
O
p
> M -«, , J- :-^-<^" Vv
I J *— f '!*>— ~. "1 .-?•••
h: - •
K^ -;
^••-- *
.>_5^ / ^
- u &
Yield (% gain)
I |<=1 (min = 0)
c Yi^ <=2
>2 (max = 2.6)
No production value reported
-------
B'
c.
I 9
l-I-
1 §•
3 I
•C ST.
"I
r> 3^
e
88
O
SS
o.
"S
88
Yield (% gain)
I |<«1 (min-0)
~] 1 < Yield <=2
!. Z!>2 (max = 3>
| No production value reported
"W
:V"N
4 ^1 I •
^
-------
o.
ft- (Jl
O ••
(71 ^ •
T3 -
•§ CL
3 O
s.
s
e
3
<
£
o
e
o.
i
P
<-f-
o'
*Tl
3
o
Yietd (% gain)
| | <=1 (mm = 0)
Q^ 1 < Yield <-2
^j >2 (max = 3)
No production value reported
-------
I
c.
• O-
3 O
-
II
oo
ja
P-K
'H
i
?v
e
88
rt-
^-
£
O
o.
i
^
o'
o_
»
3
o
Yield (% gain)
I |<=1 (min-0)
~] 1 < Yield <=2
M]>2 (max = 3.1)
| No production value reported
V
-------
>
^-
>
I
1
o.
3 V
n ,->
§•£
o a
n
e
3
O
p
o.
o
o
Yield (% gain)
| | <=1 (min =0)
| 11 < Yield <=2
| >2 (max = 2.9)
No production value reported
-------
B'
g
I
1
a.
•o £
lo
3- 00
o "_
?d a
fe r»i
§•£
•«
|
n
SB
o
e
88
fB
t«
O
o.
i
88
p^
o'
S3
Yield (% gain)
| |<=1 (min = 0)
| 11 < Yield <=2
| |>2 (max = 3.1)
No production value reported
M
O
-------
G.2 Crop Yield Concentration Responses based on Non-Adjusted W126 Metric
Crop yield responses for all crops based on unreduced O3 metrics are presented below.
G.2.1 Median Concentration Responses for Crops
u 10% - -
5% - -
-283%-
29.1%
19.7%
8.7%_
20.4%
1.20/0
0.0%
13.9%
as%
11.1%
9.6%
6.1%
16.4%
3.7%
CT v*
c
./
0*
Figure G-8: Yield Losses for Crops - 2001 Baseline (unreduced W126, Median C-R)
Abt Associates Inc
Appendix G-21
-------
1 9%
i no/
aoo/
60
2
U
u
^j
9%
n% -
3."
'O/
%
0.0%
4.*
1 / 0
0.2%
0."
7%
4.S
)%
4.1%
, 3.4%
2'5% 2-1% 2'2% 1.8%
1 3% -H 4
0.2%
Figure G-9: Yield Gains from Baseline - 84 ppb Rollback (unreduced W126, Median C-R)
12% - -
10% - -
•a
60
6% - -
4% - -
7.f
/O
7
.(.
' /
0
0.0%
1.
I
5C
—
/O
0.4%
1."
7%
9
•
JO,
0
O/
10.
1%
r
3%
1
r
%
i
8."
r
'O/
5.5%
4.1% 44%
1 3.3%
rh n
i ^ n 1 1 JL
K° . rP <<> ,&
' *//
x/. x
Figure G-10: Yield Gains from Baseline - 70 ppb Rollback (unreduced W126, Median C-R)
Abt Associates Inc
Appendix G-22
-------
14%
12°/
10% - -
1> 4% - -
12.8%
7.60/
10.5%
1"70/
. / /O
0.4%
8.50/
5.5%
2.6%
4.5%
Ov
^ ^
c*
c° o*
xei>0
A^
Figure G-ll: Yield Gains from Baseline - 25 ppm-hr Rollback (unreduced W126, Median C-R)
Abt Associates Inc
Appendix G-23
-------
12% - -
10% - -
60
6% - -
0%
9.5%
0.0%
11.9% 11.9%
1.9%
0.5%
0.5%
- -33%
4.2%
5.6%
3.5%
Figure G-12: Yield Gains from Baseline - 15 ppm-hr Rollback (unreduced W126, Median C-R)
MO/
1 9%
. S i o%
c3 1U/0 "
&Q
2
^ RO/
^ fi«/n
S
zl%
oo/
no/. -
7.C
/o
6.4
O/
0.0%
11.
0%
0.3%
l.(
5%
8.1
0.4%
9.:-
;o/
/o
2.2
o/
7.e
/o
4.6%
3.8%
3:0%
1.7%
c
C
Figure G-13: Yield Gains from Baseline - 21 ppm-hr Rollback (unreduced W126, Median C-R)
Abt Associates Inc
Appendix G-24
-------
•I
60
16%
14% - -
10% - -
6% - -
0%
14.8%
11.6%
0.5%
0.5%
1.7%
J2_9-P/
6.7%
5.4%
3.5%
Figure G-14: Yield Gains from Baseline - 13 ppm-hr Rollback (unreduced W126, Median C-R)
G.2.2 Summary Statistics of Concentration Responses for Crops
Similar results as in Chapter 5.4.1 are provided in Tables G-12 through G-18 below. These yield change
estimates are based on unreduced^!'126 metric (i.e. hourly O3 values have «ctf been reduced by 10%).
Rollback results are presented as differences or yield gains from baseline. Non-adjusted results for
Soybean and Cotton are presented in Chapter 5.4.1.
Table G-12: Yield Losses for Crops, Fruits and Vegetables - 2001 Baseline (unreduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Sorghum
Cantaloupe
Corn
Cotton
Abt Associates Inc
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
Max
8.67%
30.94%
28.25%
25.55%
0.00%
26.15%
19.71%
15.84%
1.24%
2.07%
29.09%
1.06%
0.48%
0.07%
40.17%
Min
0.00%
0.28%
0.26%
0.23%
0.00%
0.59%
0.17%
0.10%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
1.68%
Mean
0.24%
15.07%
13.76%
12.45%
0.00%
8.71%
4.95%
3.69%
0.56%
0.11%
10.87%
0.04%
0.01%
0.00%
10.12%
Apper
Median
0.02%
15.56%
14.20%
12.85%
0.00%
8.96%
4.96%
3.66%
0.63%
0.08%
12.79%
0.01%
0.00%
0.00%
9.54%
idix G-25
STD
0.66%
3.68%
3.36%
3.04%
0.00%
3.75%
2.59%
2.02%
0.35%
0.13%
6.42%
0.08%
0.03%
0.00%
4.50%
-------
Median 13.90% 0.12% 1.87% 1.64% 1.36%
Min 20.99% 0.26% 3.03% 2.64% 2.01%
Onion Median 891 9.60% 1.27% 4.77% 4.90% 1.17%
Peanut Median 409 11.13% 0.04% 1.17% 0.85% 1.11%
Soybean
Max
Median
Min
12.27%
2,090 6.11%
2.34%
0.50%
0.03%
0.06%
4.21%
1. 12%
0.69%
4.03%
0.92%
0.64%
2.23%
0.89%
0.41%
Valencia Orange Median 89 20.38% 2.34% 8.77% 7.53% 4.29%
Max 22.82% 3.01% 11.85% 12.11% 2.43%
Tomato Processing Median 2,236 16.40% 2.17% 8.52% 8.70% 1.75%
Min 15.21% 2.01% 7.90% 8.07% 1.62%
Max 41.57% 0.78% 14.35% 13.47% 8.07%
Winter Wheat Median 2,533 3.69% 0.00% 0.32% 0.16% 0.39%
Min 2.65% 0.00% 0.29% 0.17% 0.31%
Abt Associates Inc Appendix G-26
-------
Table G-13: Yield Gains
(unreduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
from Baseline
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
for Crops,
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
Fruits and
Max
3.68%
3.61%
3.30%
2.98%
0.00%
6.01%
4.83%
4.10%
0.15%
0.74%
4.93%
0.48%
0.23%
0.05%
7.83%
4.15%
5.69%
1.26%
3.40%
3.56%
2.52%
0.73%
2.06%
2.99%
2.15%
2.00%
10.17%
1.79%
1.16%
Vegetables
- 84 ppb
Min Mean
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.03%
0.14%
0.12%
0.11%
0.00%
0.24%
0.18%
0.15%
0.00%
0.01%
0.19%
0.01%
0.00%
0.00%
0.11%
0.05%
0.07%
0.06%
0.03%
0.09%
0.05%
0.02%
0.14%
0.10%
0.07%
0.07%
0.25%
0.03%
0.02%
Rollback
Median
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
STD
0.25%
0.46%
0.42%
0.38%
0.00%
0.75%
0.59%
0.48%
0.01%
0.04%
0.63%
0.04%
0.02%
0.00%
0.69%
0.33%
0.44%
0.19%
0.23%
0.36%
0.23%
0.07%
0.39%
0.35%
0.25%
0.23%
1.04%
0.13%
0.09%
Abt Associates Inc
Appendix
G-27
-------
Table G-14: Yield Gains
(unreduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
from Baseline
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
for Crops,
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
Fruits and
Max
7.52%
8.32%
7.60%
6.87%
0.00%
14.26%
12.53%
10.43%
0.42%
1.72%
9.69%
0.95%
0.44%
0.07%
21.99%
10.32%
14.48%
2.60%
8.73%
7.54%
4.12%
1.57%
5.48%
6.19%
4.45%
4.12%
22.31%
3.30%
2.27%
Vegetables
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
- 70 ppb
Mean
0.15%
1.69%
1.54%
1.39%
0.00%
2.51%
1.78%
1.39%
0.10%
0.05%
1.99%
0.03%
0.01%
0.00%
2.49%
0.66%
1.02%
0.50%
0.53%
1.18%
0.45%
0.22%
1.02%
1.23%
0.89%
0.82%
3.71%
0.20%
0.16%
Rollback
Median
0.00%
1.42%
1.29%
1.17%
0.00%
1.83%
1.24%
0.94%
0.09%
0.02%
1.49%
0.00%
0.00%
0.00%
1.95%
0.47%
0.68%
0.32%
0.28%
0.79%
0.25%
0.14%
0.57%
0.97%
0.70%
0.65%
2.06%
0.05%
0.05%
STD
0.56%
1.57%
1.43%
1.29%
0.00%
2.57%
1.90%
1.50%
0.08%
0.10%
2.07%
0.07%
0.03%
0.00%
2.67%
0.96%
1.37%
0.55%
0.80%
1.30%
0.63%
0.25%
1.26%
1.20%
0.86%
0.80%
4.31%
0.32%
0.24%
Abt Associates Inc
Appendix
G-28
-------
Table G-15: Yield Gains
(unreduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
from Baseline
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
for Crops,
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
Fruits and
Max
7.64%
8.36%
7.63%
6.90%
0.00%
14.48%
12.77%
10.63%
0.39%
1.75%
9.56%
0.96%
0.45%
0.07%
22.52%
10.52%
14.79%
2.62%
8.46%
3.47%
2.61%
0.73%
5.50%
6.24%
4.48%
4.16%
22.26%
3.26%
2.25%
Vegetables
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
- 25 ppm-hr
Rollback
Mean Median
0.06%
0.14%
0.13%
0.12%
0.00%
0.23%
0.18%
0.15%
0.01%
0.01%
0.18%
0.01%
0.00%
0.00%
0.34%
0.13%
0.19%
0.06%
0.11%
0.10%
0.05%
0.02%
0.42%
0.10%
0.07%
0.06%
0.28%
0.03%
0.02%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
STD
0.51%
0.53%
0.48%
0.43%
0.00%
0.91%
0.76%
0.62%
0.04%
0.08%
0.66%
0.04%
0.02%
0.00%
1.69%
0.73%
1.03%
0.21%
0.56%
0.37%
0.21%
0.08%
1.05%
0.37%
0.27%
0.25%
1.19%
0.16%
0.11%
Abt Associates Inc
Appendix
G-29
-------
Table G-16: Yield Gains from Baseline for Crops, Fruits and Vegetables - 15 ppm-hr W126
Rollback (unreduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
Max
8.20%
10.39%
9.48%
8.58%
0.00%
17.70%
15.10%
12.46%
0.50%
1.91%
11.90%
1.02%
0.47%
0.07%
27.52%
11.90%
17.09%
3.26%
9.79%
6.55%
4.18%
1.38%
6.84%
7.74%
5.57%
5.16%
27.43%
3.50%
2.46%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.12%
1.12%
1.03%
0.93%
0.00%
1.66%
1.23%
0.97%
0.08%
0.05%
1.39%
0.02%
0.01%
0.00%
2.01%
0.52%
0.87%
0.30%
0.45%
0.92%
0.33%
0.18%
0.80%
0.81%
0.58%
0.54%
2.60%
0.16%
0.13%
Median
0.00%
0.48%
0.44%
0.40%
0.00%
0.33%
0.24%
0.19%
0.01%
0.01%
0.46%
0.00%
0.00%
0.00%
0.51%
0.11%
0.19%
0.00%
0.02%
0.30%
0.09%
0.05%
0.00%
0.27%
0.19%
0.18%
0.41%
0.01%
0.01%
STD
0.59%
1.37%
1.25%
1.13%
0.00%
2.26%
1.72%
1.37%
0.10%
0.11%
1.75%
0.07%
0.03%
0.00%
3.05%
1.05%
1.55%
0.48%
0.89%
1.18%
0.54%
0.23%
1.54%
1.03%
0.74%
0.69%
3.63%
0.30%
0.22%
Abt Associates Inc
Appendix
G-30
-------
Table G-17: Yield Gains from Baseline for Crops, Fruits and Vegetables - 21 ppm-hr W126
Rollback (unreduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
Max
7.06%
7.05%
6.43%
5.82%
0.00%
12.22%
11.01%
9.22%
0.34%
1.60%
8.06%
0.90%
0.42%
0.07%
18.96%
9.35%
12.94%
2.21%
7.61%
2.34%
1.74%
0.49%
4.64%
5.26%
3.78%
3.51%
18.71%
3.02%
2.06%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.05%
0.06%
0.06%
0.05%
0.00%
0.11%
0.09%
0.08%
0.00%
0.01%
0.08%
0.00%
0.00%
0.00%
0.18%
0.08%
0.11%
0.03%
0.05%
0.03%
0.02%
0.01%
0.32%
0.04%
0.03%
0.03%
0.12%
0.02%
0.01%
Median
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
STD
0.46%
0.37%
0.34%
0.31%
0.00%
0.65%
0.57%
0.47%
0.03%
0.07%
0.47%
0.04%
0.02%
0.00%
1.31%
0.61%
0.85%
0.16%
0.45%
0.20%
0.13%
0.04%
0.87%
0.26%
0.19%
0.17%
0.84%
0.13%
0.09%
Abt Associates Inc
Appendix
G-31
-------
Table G-18: Yield Gains from Baseline for Crops, Fruits and Vegetables - 13 ppm-hr W126
Rollback (unreduced W126)
Crop
Kidney Beans
Grapes
Lettuce
Potato
Rice
Grain Sorghum
Cantaloupe
Corn
Cotton
Onion
Peanut
Soybean
Valencia Orange
Tomato Processing
Winter Wheat
Response
Median
Max
Median
Min
Median
Max
Median
Min
Median
Median
Median
Max
Median
Min
Max
Median
Min
Median
Median
Max
Median
Min
Median
Max
Median
Min
Max
Median
Min
TV
587
1,883
678
1,585
300
1,331
1,610
2,623
666
891
409
2,090
89
2,236
2,533
Max
8.15%
10.13%
9.24%
8.36%
0.00%
17.30%
14.83%
12.25%
0.49%
1.89%
11.59%
1.01%
0.46%
0.07%
26.90%
11.75%
16.82%
3.18%
9.65%
5.85%
3.80%
1.25%
6.67%
7.55%
5.43%
5.03%
26.74%
3.48%
2.43%
Min
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
Mean
0.11%
0.81%
0.74%
0.67%
0.00%
1.24%
0.93%
0.74%
0.05%
0.04%
1.03%
0.02%
0.01%
0.00%
1.38%
0.38%
0.63%
0.23%
0.33%
0.67%
0.25%
0.13%
0.73%
0.58%
0.42%
0.39%
1.84%
0.13%
0.10%
Median
0.00%
0.10%
0.09%
0.09%
0.00%
0.03%
0.02%
0.01%
0.00%
0.00%
0.09%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.05%
0.01%
0.01%
0.00%
0.03%
0.02%
0.02%
0.00%
0.00%
0.00%
STD
0.58%
1.16%
1.06%
0.96%
0.00%
1.93%
1.49%
1.19%
0.08%
0.10%
1.47%
0.06%
0.03%
0.00%
2.69%
0.98%
1.43%
0.41%
0.81%
0.99%
0.47%
0.20%
1.47%
0.87%
0.62%
0.58%
2.96%
0.27%
0.20%
Abt Associates Inc
Appendix
G-32
-------
>
o-
r>
I
o
I 9
71 VO
01
GO O L>
O C3 S
ft JO. fB
ft
•a
n
a.
IB
p
-
d- O.
ft j^J
f5 K>
O,
1
Yield (% gain)
] <=1 (min = 0)
No production value reported
O
o
•J
rf
01
TS
O
1
2-
a
O
9
re
O
•s
C/5
—
5"
re
a
s
2L
a
n
o
T3
CB
a-
o
s
o
9
a
^
h^
O\
re
•s
S"
-------
O
o
o.
o
h^
O
a
o"
O
o
«-f-
p^-
o
o
o
"H
5'
=r
•?'
Yield (% loss)
I 1 <=1 (min = 0)
I 11 < Yield <=2
g. | 2 < Yield <=4
^ ^B ^ < Yield <=6
M •§ 6< Yield <=10
T* ^H >10 (max = 25.6)
No production value reported
O.
o
fB
V)
O
9
VI
n
§
S3
01
-------
§55
-: «
^S
g £
§' »'
S ?
3 o
?3
oe 6d
4_ P
t«
•a n
2 O
o. o
"2.
S'
3
O
O.
o.
% Yield (% gain)
S- | | <=1 (min = 0)
S I 11 < Yield <=2
~"l< Yield <=4
4 < Yield <=6
6 < Yield <=10
>10 (max= 11.6)
No production value reported
-------
rD ^s
^ i
oo £
•3- ••
2 «!
3 2
ST3
-i «
e~, £2
VI
"O 2_
f 5'
rs
I?
IP
c S
r> o
a^
< o
O
O.
o.
2 Yield (% gain)
£ | | <=1 (min = 0)
3 | 11 < Yield <=2
~ 2 < Yield <=4
4 < Yield <=6
6 < Yield <=10
>10 (max = 20.3)
No production value reported
-------
g*
8s5
8" 5
(71 ^
5 S
B31
g"B
5?
V)
as-.
9 9
"t ^
S o
O. O
9
o
o
t«
VI
vj
"S.
5'
3
O
O.
o.
2 Yield (% gain)
£ | | <=1 (min = 0)
3 | 11 < Yield <=2
~ 2 < Yield <=4
4 < Yield <=6
6 < Yield <=10
>10 (max = 21.8)
No production value reported
-------
(71 Ja
s £
B31
g"B
5?
V)
as-.
9 9
"t ^
S o
O. O
9
o
o
o.
12.
£'
3
J3,
O
o.
»
o
Yield (% gain)
I | <=-\ (min = 0)
I 11 < Yield <=2
I 2 < Yield <=4
^B 4 < Yield <=6
••6< Yield <=10
H>10 (max = 23.2)
No production value reported
oo
-------
•a ^
3?
=r ,
n
$
O
o_
o"
O
o
e
3
a.
c.
Yield (% gain)
I | <=1 (min =0)
I 11 < Yield <=2
I 12 < Yield <=4
•• 4 < Yield <=6
^H 6 < Yield <=1Q
^H >10 (max= 20.3)
_! No production value reported
-------
a
o
0\ ST.
V S
»
s*
o
o_
o
o
o
12.
5'
3
j3,
O
o.
Yield {% gain)
I | <= (min = 0)
I 11 < Yield <=2
| 2 < Yield <=4
^M A < Yield <=6
o.
>10 (max = 23.1)
No production value reported
-------
o
h^
^1
05
O
3
ts
M
o.
Yield (% loss)
I | <=1 (min = 0)
I 11 < Yield <=2
^H 2 < Yield <=4
^H 4 < Yield <=6
•• >6 (max = 7.7)
No production value reported
I
§
8?
O
n
-s
-------
I.?
i-s
O
5 i
a- S
"^ ^
n —•
o.
i
o
a Yield (% gain)
M | | <=1 (min = 0)
| | 1 < Yield <=2
I | >2 (max = 3.2)
o
»
No production value reported
era
-------
I.?
i-s
O
"^ ^
n —•
n
•••
£ Yield (%
gain)
o.
o. [ "| <=1 (min = 0)
S- | J 1 < Yield «=2
HI 2 < Yield <=4
| >4 (max = 5.3)
n
ys
No production value reported
CTQ
-------
o o
(71 N<
"O n'
n
if
3 W
p . £5
o £.
2 (max = 3.6)
J No production value reported
-------
O Hi
(71 N<
"o S'
If
If
""£ MM
Qri &3
o £.
<* 5"
o- 3
V)
2
*c.
3
M
O
o.
65
P4-
o'
o.
Yield (% gain)
[ |<=1 (min = 0)
I 11 < Yield <=2
H 2 < Yield <=4
^Bj >4 (max = 5.4)
J No production value reported
-------
o.
as-
l
n
5*
C/5
O
n
3
O
o.
P
<-f-
S"
P
9r
$
Yield (% gain)
I | <=1 (min =0)
I 11 < Yield <=2
^B>2 (max = 2.7)
No production value reported
-------
o.
-! li.
IS
II-
n
n
C/5
O
VI
fB
3
O
Yield (% gain)
I | <=1 (min = 0)
I 11 < Yield <=2
I | 2 < Yield <=4
^H >4 (max = 5)
No production value reported
88
pr
^
-------
•a
B-
o
I
H
n
O.
o.
3
h-1
ON
c.
Yield (% loss)
^] <=1 (min = 0)
| 11 < Yield <= 2
| | 2 < Yield <=4
^| 4 < Yield <=6
HI >6 (max = 6.9)
No production value reported
8?
•8
o
VI
n
I
•a
00
-------
s
oo
4_
•a
c.
J3,
O
o.
»
o
Yield (% gain)
| | <=1 (min = 0)
I 11 < Yield <=2
| | 2 < Yield <=4
| >4 (max = 4.6)
No production value reported
-------
TO
ft ,«•.
if QJ
os ^
§55
i <*
o
3
i
03
T3 (^
TS ST.
o JJ
rs s»
o
e
3
o.
O
CL
»
o
Yield (% gain)
| | <=1 (min = 0)
| 11 < Yield <=2
| | 2 < Yield <=4
^| 4 < Yield <=6
| >6 (max = 6.4)
No production value reported
-------
(71 N<
"O n'
II
§ B-
" ?
03
EL
as-
»
88
I
O
CL
•«
^' Yield (% gain)
3 H <=1 (min =0)
o.
11 < Yield <=2
2 < Yield <=4
4 < Yield <=6
>6 (max = 6.4)
No production value reported
1
-------
O 00
h^
(71 Kj
I-
?!
n 60
£g
|
I
e
3
c.
O
CL
C5
O
Yield (% gain)
| | <=1 (min = 0)
| 11 < Yield <=2
| | 2 < Yield <=4
^| 4 < Yield <=6
| >6 (max = 6.7)
No production value reported
KJ
1
-------
B-
I
1
c.
I?
I
H
<-f-
n'
e
3
jj
^ Yield (% gain)
o [ Zl<=1 (min = °)
*•<
»_ | 11 < Yield <=2
3 ^| 2 < Yield <=4
2. ^| >4 (max = 5.9)
» [ No production value reported
..a*'
Oi
-------
3
±
o
I?
88
o.
88
«-f-
S'
o.
I
H
<-f-
n'
e
3
<' Yield (% gain)
3^ [ ^J <=1 (min = 0
O I 11 < Yield <=2
| 2 < Yield <=4
| 4 < Yield <=6
I >6 (max = 6.6)
» [ No production value reported
-------
Appendix H Summary Statistics for Tree Seedling
Biomass Responses based on Ten Percent Adjusted
W126 Metric
Appendix H summarizes relative tree seedling biomass gains under the six scenarios presented in Section
4:
Ozone levels which just meet an 84 ppb 4th highest 8-hour maximum standard (in this case, we
performed a rollback to 84.999 ppb, since the 4th highest 8-hour maximum standard truncates
decimals)
Ozone levels which just meet a 70 ppb 4th highest 8-hour maximum standard (actually 70.999 for
the reasons above)
Ozone levels which just meet a 25ppm-h 3-month 12-hour SUM06 standard
Ozone levels which just meet a 15ppm-h 3-month 12-hour SUM06 standard.
Ozone levels which just meet a 21ppm-h 3-month 12-hour W126 standard.
Ozone levels which just meet a 13ppm-h 3-month 12-hour W126 standard.
Appendix H. 1 present results based on the reduced or "adjusted" O3 metric (i.e. hourly O3 values were
adjusted down by 10%). Appendix H.2 present similar results based on the unreduced metric (unadjusted
hourly O3 values). Refer to Section 5 for a discussion of the data and method used to estimate O3
concentration responses on vegetation.
H.I Tree Seedling Biomass Responses based on Ten Percent Adjusted W126 Metric
This section includes summary box plots and tables showing concentration responses for all ten tree
species based on reduced hourly O3 values for the six O3 standards under consideration. Maps for aspen,
black cherry, and ponderosa pine are included in the body of the document.
Abt Associates Inc Appendix H-l
-------
H.1.1 Median Concentration Responses for Tree Seedlings
45% -
40% -
35% -
$ 30% -
O
£ 25% -
^ 20% -
15% -
10% -
5% -
0% -
40.9%
11.
5%
~
_
19.9%
0.0%
13.5% 13.6%
2.3% ' r^ — i
0.6% p^ 1-2%
— ' i ^=^ 1 !
Figure H-l: Median Tree Seedling Biomass Loss - 2001 Baseline (reduced W126)
Abt Associates Inc
Appendix H-2
-------
30%
S
•a
16.8%
O
5% - - -
9.30/0
9'9%
7.8%
1.3%
^
Figure H-2: Median Tree Seedling Biomass Gain from Baseline - 84 ppb Rollback (reduced W126)
Abt Associates Inc
Appendix H-3
-------
30%
25% - - -
S
•a
O
£
20% - - -
15.7%
12.7%
11.7%
10% - -
5% - - -
1.9%
3.0%
Figure H-3: Median Tree Seedling Biomass Gain from Baseline - 70 ppb Rollback (reduced W126)
Abt Associates Inc
Appendix H-4
-------
25% -
20% -
a
•a
«
s
2 15/0
5% -
n%
16.8%
15.4%
5QO/
.J /O
0.0%
-
2.8%
1.2%
0.0% °'5/°
<^ ^ /" ^ ^& ^& x ./" .y
Figure H-4: Median Tree Seedling Biomass Gain from Baseline - 25 ppm-hr Rollback (reduced
W126)
Abt Associates Inc
Appendix H-5
-------
30%
25% - -
•a
60
O
25.0%
15% - -
10% - - - -8:5%-
5% - -
n
17.7%
0.0%
12.1%
3.0%
1.8%
11.2%
4
, .
/
Figure H-5: Median Tree Seedling Biomass Gain from Baseline - 15 ppm-hr Rollback (reduced
W126)
Abt Associates Inc
Appendix H-6
-------
Z.J /O
20% -
•a
S? 15% -
8
o
f>
;| 10% -
1
5H-
11.3%
4.2%
15.7%
0.0%
7.5%
6.1%
2.6%
1-0% 0 40/
0.0% °'4/0
1
^
Figure H-6: Median Tree Seedling Biomass Gain from Baseline - 21 ppm-hr Rollback (reduced
W126)
Abt Associates Inc
Appendix H-7
-------
Z.J /O
20% -
'I
S? 15% -
%
a
o
£
;| 10% -
OJ
s
n°/
23.
79%
2%
17.7%
0.0%
11.7%
0.0%
1.7%
I
3.C
)%
10.7%
0.7%
&
t&
-------
Table H-2: Median Tree Seedling Biomass Gain - 84 ppb Rollback (reduced W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-3: Median
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-4: Median
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Tree Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Tree Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Max
5.8%
16.8%
0.0%
9.3%
0.0%
1.3%
2.8%
9.9%
0.6%
7.8%
Gain - 70
Max
9.3%
28.6%
0.0%
15.7%
0.0%
1.9%
3.0%
12.7%
0.9%
11.7%
Gain - 25
Max
5.3%
15.4%
0.0%
16.8%
0.0%
1.2%
2.8%
8.4%
0.5%
8.0%
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
ppb Rollback
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Mean
0.1%
0.4%
0.0%
0.1%
0.0%
0.0%
0.0%
0.2%
0.0%
0.3%
(reduced
Mean
1.1%
5.4%
0.0%
0.3%
0.0%
0.3%
0.1%
1.6%
0.3%
2.1%
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
W126)
Median
0.0%
4.0%
0.0%
0.0%
0.0%
0.2%
0.0%
1.1%
0.3%
1.7%
STD
0.6%
1.6%
0.0%
2.1%
0.0%
0.1%
0.1%
0.9%
0.1%
0.9%
STD
1.9%
5.3%
0.0%
4.1%
0.0%
0.3%
0.2%
1.8%
0.2%
2.1%
ppm-hr Rollback (reduced W126)
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Mean
0.1%
0.3%
0.0%
0.2%
0.0%
0.0%
0.0%
0.2%
0.0%
0.2%
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
STD
0.5%
1.4%
0.0%
3.8%
0.0%
0.1%
0.1%
0.8%
0.1%
0.8%
Abt Associates Inc
Appendix
H-9
-------
Table H-5: Median Tree Seedling Biomass Gain - 15 ppm-hr Rollback (reduced W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-6: Median Tree
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-7: Median Tree
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
H.2.3 Seedling Biomass Response Maps
Response maps for aspen,
Max
8.5%
25.0%
0.0%
17.7%
0.0%
1.8%
3.0%
12.1%
0.8%
11.2%
Gain - 21
Max
4.2%
11.3%
0.0%
15.7%
0.0%
1.0%
2.6%
7.5%
0.4%
6.1%
Gain - 13
Max
7.9%
23.2%
0.0%
17.7%
0.0%
1.7%
3.0%
11.7%
0.7%
10.7%
Min Mean
0.0% 0.6%
0.0% 3.4%
0.0% 0.0%
0.0% 0.3%
0.0% 0.0%
0.0% 0.2%
0.0% 0.1%
0.0% 1.2%
0.0% 0.2%
0.0% 1.3%
Median
0.0%
0.9%
0.0%
0.0%
0.0%
0.1%
0.0%
0.7%
0.2%
0.0%
STD
1.5%
4.6%
0.0%
4.5%
0.0%
0.3%
0.2%
1.7%
0.2%
1.9%
ppm-hr Rollback (reduced W126)
Min Mean
0.0% 0.0%
0.0% 0.1%
0.0% 0.0%
0.0% 0.2%
0.0% 0.0%
0.0% 0.0%
0.0% 0.0%
0.0% 0.1%
0.0% 0.0%
0.0% 0.1%
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
STD
0.3%
0.8%
0.0%
3.4%
0.0%
0.1%
0.1%
0.5%
0.1%
0.5%
ppm-hr Rollback (reduced W126)
Min Mean
0.0% 0.5%
0.0% 2.4%
0.0% 0.0%
0.0% 0.3%
0.0% 0.0%
0.0% 0.1%
0.0% 0.1%
0.0% 1.0%
0.0% 0.2%
0.0% 1.1%
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.3%
0.1%
0.0%
STD
1.3%
3.8%
0.0%
4.4%
0.0%
0.2%
0.2%
1.5%
0.1%
1.7%
for Selected Tree Species:
ponderosa pine and black cherry based on the reduced
O3 metric are
presented
in the body of the document under Section 5.
Abt Associates Inc Appendix H-10
-------
H.2 Tree Seedling Biomass Responses based on Non-Adjusted W126 Metric
This section include summary tables and maps showing concentration responses for tree seedlings based
on unreduced hourly O3 values.
H.2.1 Median Concentration Responses for Tree Seedlings
The following box plots and tables show count, mean, maximum, minimum, median, and standard
deviation statistics for the ten tree species and the six O3 standards under consideration. The concentration
responses were derived from unreduced hourly O3 values.
_o
VI
8
o
IS
T3
-------
30%
25% - - -
•1
%
I
20% - - -
15% ---
16.8%
u 10% - -
5% - - -
9.30/0
1.3%
9'9%
7.8%
Figure H-9: Median Tree Seedling Biomass Gain from Baseline - 84 ppb Rollback (unreduced
W126)
Abt Associates Inc
Appendix H-12
-------
30%
25% - - -
20% - - -
S
•a
15.7%
o
12.7%
11.7%
10% - -
5% - - -
1.9%
3.0%
Figure H-10: Median Tree Seedling Biomass Gain from Baseline - 70 ppb Rollback (unreduced
W126)
Abt Associates Inc
Appendix H-13
-------
25% -
20% -
'I
60
| 15% -
.0
5% -
16.8%
15.4%
5QO/
.J /O
0.0%
-
2.8%
1.2%
0.0% °'5/°
(*$ *£ .
-------
30%
25% - -
20% - -
S
•a
o
25.0%
10% - - - -8:5%-
5% - -
n
17.7%
0.0%
12.1%
3.0%
1.8%
11.2%
Figure H-12: Median Tree Seedling Biomass Gain from Baseline - 15 ppm-hr Rollback (unreduced
W126)
Abt Associates Inc
Appendix H-15
-------
60
I
£
20% -
15% -
10% -
n°/
11.0%
5.1%
19.1%
0.1%
20.4%
1.3%
0.0%
12.1%
9.4%
0.5%
&
Figure H-13: Median Tree Seedling Biomass Gain from Baseline - 21 ppm-hr Rollback (unreduced
W126)
Abt Associates Inc
Appendix H-16
-------
25% -
20% -
•| 15% "
60
c/3
c/3
o
'£ 10% -
n«/, -
10.4%
24.2%
22.7%
0.1%
24.2%
0.0%
2.4%
£_
20.
~\
9%
17.
0.9%
2%
&
Figure H-14: Median Tree Seedling Biomass Gain from Baseline - 13 ppm-hr Rollback (unreduced
W126)
H.2.2 Summary Statistics of Concentration Responses for Tree Seedlings
Table H-8: Tree Seedling Biomass Loss - 2001 Baseline (unreduced W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Max
17.6%
52.8%
0.1%
27.6%
1.2%
3.7%
24.5%
26.4%
1.7%
24.2%
Min
0.3%
5.2%
0.0%
0.2%
0.1%
0.1%
0.0%
0.1%
0.5%
0.4%
Mean
5.1%
27.3%
0.0%
2.7%
0.3%
1.2%
0.8%
6.4%
1.0%
6.9%
Median
4.1%
27.4%
0.0%
1.7%
0.2%
1.2%
0.2%
6.2%
1.0%
6.6%
STD
4.0%
8.6%
0.0%
8.7%
0.7%
0.6%
1.8%
4.4%
0.2%
4.6%
Abt Associates Inc
Appendix
H-17
-------
Table H-9: Median Tree Seedling Biomass Gain from Baseline - 84 ppb Rollback (unreduced
W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-10: Median
W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-ll: Median
W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Tree Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Tree Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Max
5.8%
16.8%
0.0%
10.0%
0.0%
1.9%
20.5%
16.7%
0.7%
11.5%
Gain from
Max
9.3%
31.2%
0.1%
19.1%
0.0%
2.9%
24.3%
23.6%
1.1%
18.4%
Gain from
Max
5.3%
15.8%
0.1%
20.9%
0.0%
1.8%
22.9%
16.0%
0.7%
12.7%
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Baseline -
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Baseline -
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Mean
0.1%
0.4%
0.0%
0.1%
0.0%
0.0%
0.2%
0.4%
0.0%
0.5%
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
STD
0.6%
1.7%
0.0%
2.3%
0.0%
0.2%
1.2%
1.6%
0.1%
1.4%
70 ppb Rollback (unreduced
Mean
1.1%
6.8%
0.0%
0.4%
0.0%
0.5%
0.8%
3.8%
0.4%
3.9%
25 ppm-hr
Mean
0.1%
0.3%
0.0%
0.2%
0.0%
0.0%
0.3%
0.4%
0.0%
0.3%
Median
0.0%
5.4%
0.0%
0.0%
0.0%
0.4%
0.2%
3.1%
0.4%
3.4%
STD
1.9%
6.3%
0.0%
5.3%
0.0%
0.5%
1.8%
3.8%
0.2%
3.7%
Rollback (unreduced
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
STD
0.5%
1.4%
0.0%
4.8%
0.0%
0.1%
1.3%
1.4%
0.1%
1.2%
Abt Associates Inc
Appendix
H-18
-------
Table H-12: Median Tree Seedling Biomass Gain from Baseline - 15 ppm-hr Rollback (unreduced
W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-13: Median
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
Table H-14: Median
(unreduced W126)
Tree Species
Aspen
Black Cherry
Douglas Fir
Ponderosa Pine
Red Alder
Red Maple
Sugar Maple
Tulip Poplar
Virginia Pine
Eastern White Pine
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Tree Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Tree Seedling Biomass
N
8,241
19,860
538
432
103
17,889
11,396
13,551
3,632
6,874
Max
8.5%
26.8%
0.1%
22.8%
0.0%
2.6%
24.4%
22.0%
1.0%
18.4%
Loss- 21
Max
5.1%
11.3%
0.0%
15.7%
0.0%
1.0%
2.6%
7.5%
0.4%
6.1%
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Mean
0.6%
4.2%
0.0%
0.4%
0.0%
0.3%
0.7%
2.9%
0.3%
2.3%
ppm-hr W126 Rollback
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Gain from Baseline -
Max
10.4%
24.2%
0.1%
22.7%
0.0%
2.4%
24.2%
20.9%
0.9%
17.2%
Min
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Mean
0.0%
0.1%
0.0%
0.2%
0.0%
0.0%
0.0%
0.1%
0.0%
0.1%
Median
0.0%
1.2%
0.0%
0.0%
0.0%
0.1%
0.1%
1.7%
0.3%
0.0%
(unreduced
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
STD
1.5%
5.4%
0.0%
6.1%
0.0%
0.4%
1.8%
3.6%
0.2%
3.4%
W126)
STD
0.3%
0.8%
0.0%
3.4%
0.0%
0.1%
0.1%
0.5%
0.1%
0.5%
13 ppm-hr W126 Rollback
Mean
0.7%
2.9%
0.0%
0.4%
0.0%
0.2%
0.7%
2.2%
0.2%
1.8%
Median
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.7%
0.2%
0.0%
STD
1.8%
4.4%
0.0%
5.9%
0.0%
0.4%
1.7%
3.2%
0.2%
2.9%
Abt Associates Inc
Appendix
H-19
-------
>
^-
>
<&
%
n
I
c.
a
o
fl
2 "°
*
n
0.
o
3
o
9
O.
n
O
VI
TS
O
O.
n
CTQ
i
O
n
o
o.
rt
o
O
VI
Biomass ( % loss)
I | <=1 (min = 0)
| | 1 < Biomass <=2
I | 2 < Biomass <=A
HI 4 < Biomass <=6
•I 6 < Biomass <=10
^•>10 (max = 27.6)
No production value reported
C/5
re
re
a
5'
era
a
o°
5
89
re
VI
•a
o
9
O
"I
C/5
re
a
H
re
re
C/5
T3
re
n
VI
-------
o.
SB
CTQ 7*
fB -^
S- ca
n-
O as
l£
H 9
3' ?
e o
3 3
8- 60
i?
3 o
o. »
g »
s-I
Sa
l>> g
O\ fD
O
9
CL
fB
o
a.
3
p^-
n'
I
Biomass (% gain)
I | <=1 (min = 0)
I 11 < Biomass <=2
I 12 < Biomass <=4
^H 4 < Biomass <=6
HI 3 < Biomass <=10
^B >IO (max= 10.1)
No production value reported
-------
era rr
?f *
H S
3>
e o
3 3
o »
5?
o- S
s S-
^ t«
5*
o.
O
9
CL
n
o
SB
P.
Biomass (% gain)
I | <=1 (min = 0)
I 11 < Biomass <=2
I 12 < Biomass <=4
^H 4 < Biomass <=6
HI 3 < Biomass <=10
^B >IO (max= 19.1)
No production value reported
a
KJ
I
-------
o.
a
Oi
o ^*
a a
»" 4^
t/5 •'
I I
ON ^3
=?
i O
O. -*i
c ?
r> "«
c Biomass (% gain)
^
TS ~| <=1 (min = 0)
^
S D 1 * Biomass <=2
^ ~j 2 < Biomass <=4
o
2 • 4 < Biomass <=6
O
»
rt-
n'
I
3 < Biomass <=10
>IO (max = 20.9)
No production value reported
-------
o.
V1 2
B &2
o "^
- S
^ (71
in ••
d S3
si
8" s
K?
3 3^
H
c S3
o. -*i
c o
o i
o\ o
O
9
a.
n
O
p.
3
p^-
o'
I
Biomass (% gain)
I | <=1 (min = 0)
I 11 < Biomass <=2
I 12 < Biomass <=4
^H 4 < Biomass <=6
HI 3 < Biomass <=10
^B >IO (max = 22.8)
No production value reported
-------
. 8:
iTS
if
lo
"* £
'e* 3'
s 2»
a S
^1
o.
fB
t V
I
/
c.
CL
fB
O
89
a.
Siomass(%
I I -==1 (mil = 0)
1
2 < Biomass <=4
6 < Biorns&s <=
>10 in-.3>: = 19.1}
No produdion value reported
a
M
ys
o
-------
~ ffi
o ',
o.
w S3
11
sf %
§ O
"t M
9s 5'
° 5?
S o
&B
fB
O
9
O.
n
o
9
CL
O
p.
3
,. V
|
/
L3 uiiuifts ;-•;., gain)
~j «1 (min*0)
I I 1 < Biomas-s <=2
I I 2 < Biomas-s <='<
^H ^ < B\OTK::-, •:-f-l
^B € < Komas« <=10
H s-10 (max • 22.7}
No production value reported
o_
cr
-------
2
5"
ffi
05
a
o
3
r
o
S
S"
o
n
j
^
9
O.
e
c.
Biomass (% loss)
I |<=5 (min=1.4)
I 15 < Biomass <=10
I I 10 < Biomass <=20
(•I 20 < Biomass <=30
^H >30 (max = 54)
No production value reported
n
O
n
a
-4
-------
3 ffi
oo
^ O
c^ 2.
9 9
3 =T
M
o B
2. 03
o.
2
5"
o
?r
O
O
»
o.
o_ Biomass (% gain)
S1 I I <=5 (min = 0)
£. I |5 < Biomass <=1Q
*• d3>10 (max =17.6)
M No production value reported
era'
a
00
oo
9-
-------
3 ffi
a ^
3 o
rt-
^ o'
II
O" ™
^ O
C O
n a
fB B
a 03
5- y
^ ss
H- n
t^t cr.
o\ a
•t
o.
O
o.
p
ffi
era'
Biomass {% gain)
I I <=5 (min = 0)
I 15 < Biomass <=10
I | 10 < Biomass <=20
^HJ 20 < Biomass <=30
^H>30 (max = 31.8)
No production value reported
-------
"I
1 ffi
O. M
SS
O. 3
03
-!
o
c.
I
33 Biomass (% gain)
B f I <=5 (min = 0)
g I 15 < Biomass <=10
£ •i>10 (max =15.3)
g No production value reported
o
9
in
-------
-
S •§
1 ffi
o.
a M
SS
a s
03
-!
n
o
I
n
ys
3
o
9
Biomass (% gain)
I | <=5 (min = 0)
I 15 < Biomass <=10
I I 10 < Biomass <=20
•• >20 (max = 27.5)
No production value reported
05
-------
3
3
O.
a
5'
H
3
O
03
VI
2.
5"
n
03
S"
O
a.
o.
Biomass(% gain)
I I <=5 (min = 0)
I I 5 < Biomass <=10
r~l>10 (max =12.4)
| No production value reported
a
£>
KJ
-------
ffi
o'
3
o
03
VI
2.
5"
n
03
S"
o
o.
-s
S- Biomass (% gain)
?3 I | <=5 (min = 0)
== I 15 < Biomass <=10
S • 10 < Biomass <=20
o.
>20 (max = 25.6)
No production value reported
-------
r
o
o
I
§
1
o.
c.
Biomass (% loss)
I l<=1 (min = 0)
I 11 < Biomass <=2
I | 2 < Biomass <=4
^HJ 4 < Biomass <=6
IB 3 < Biomass <=10
^B >10 (max =30)
No production value reported
n
O
9
•§
-------
o.
3 ffi
3 ^
3 o\
oo 5'
-
e o
n a
fB B
a 03
O) 3
S*
>
O.
n
o
Jti.
rt-
ffi
era'
n
oo
Biomass (% gain)
I I <=1 (min = 0)
I 11 < Biomass <=2
[ | 2 < Biomass <=4
^H 4 < Biomass <=6
^B @ < Biomass <=10
•i>10 (max = 12.3)
No production value reported
-------
B'
|
1
o.
3 ffi
c i^
3 .^
^ 'mm*
If
^ O
3 5,
O. !?
e o
o s
h- £B
TS
rs
o.
o
o
Jti.
rt-
ffi
ore'
oo
Biomass (% gain)
I I <=1 (min = 0)
I 11 < Biomass <=2
[ | 2 < Biomass <=4
^H 4 < Biomass <=6
^B @ < Biomass <=10
•i>10 (max = 19.6)
No production value reported
-------
a
1 1
o.
o.
o
n
$
>
O.
n
o
65
rs
o
O
Vi
i
o
Biomass (% gain)
I | <=1 (min = 0)
I | 1 < Biomass <=2
I | 2 < Biomass <=4
H 4 < Biomass <=6
^B 6 < Biomass <=10
^H>10 (max = 23.6)
No production value reported
-------
o.
a
6
00
C H-
NJ \O
fa
1 1
o.
o
n
$
>
O.
n
o
65
rs
o
O
Vi
i
o
<-f-
o
h^
(71
Biomass (% gain)
I | <=1 (min = 0)
I | 1 < Biomass <=2
I | 2 < Biomass <=4
H 4 < Biomass <=6
^B 6 < Biomass <=10
^H>10 (max = 23.6)
No production value reported
-------
o.
15
2 "°
I?
r> •>>
O
3
n
o1
O.
n
o
65
rs
Biomass (% gain)
I | <=1 (min=0)
I 11 < Biomass <=2
I I 2 < Biomass <=4
Bl 4 < Biomass <=6
^B 6 < Biomass <=1G
^H>10 (max = 17.0)
No production value reported
3
i
cr
o
-------
15
2 "°
I?
r> •>>
n H-
O. ••
o
3
cs
n
o1
a.
n
o
65
rs
o.
Biomass (% gain)
I | <=1 (min=0)
I 11 < Biomass <=2
I I 2 < Biomass <=4
Bl 4 < Biomass <=6
^B 6 < Biomass <=10
^H>10 (max = 22.3)
No production value reported
a
o
3"
O
-------
Appendix I AGSIM© Model Specifications
AGSIM© is an econometric-simulation model that is based on a large set of statistically estimated
demand and supply equations for agricultural commodities produced in the United States. This model has
been peer-reviewed and utilized in many pesticide and other major agricultural policy evaluations
(Taylor et al., 1993). The present version of the model includes supply and utilization of corn, sorghum,
barley, oats, wheat, soybeans, cotton, all hay, rice, peanuts, fresh and processed peaches, walnuts, fresh
and processed apples, lettuce, onions, fresh and processed berries, fresh and processed oranges, fresh and
processed grapes, and fresh and processed tomatoes. Supply of the major field crops is regionalized for
the nine USD A production regions. Demand for each commodity is separated into various components,
including stocks. Imports and exports are modeled separately.
The model is capable of analyzing the effects of changes in policies that affect crop yields or production
costs. This is achieved by estimating how farmers will adjust crop acreage between commodities when
relative profitability changes as a result of policy-induced crop yield and/or production cost changes53.
Acreage and yield changes from various scenarios will affect total production of crops, which
simultaneously affects both commodity prices and consumption. Commodity price changes, in turn, affect
profitability and cropping patterns in subsequent years. Federal farm program and conservation reserve
effects are also incorporated into the model.
Model Specification
AGSIM© is based on a set of dynamic supply and demand equations for major crops. Commodities are
generally linked on both the supply side and demand side of markets. The simulation component of the
model finds the set of prices for all commodities endogenous to the model that simultaneously clear all
markets in each year over the simulation period. Dynamics are incorporated into the econometric
specification and thus incorporated into the simulation model. All equations in the model were
econometrically estimated, except a few policy equations that were based on legislated formula.
Supply Components
The crop supply component of AGSIM© is based on a set of supply equations for the major field crops
produced in the United States. Effects of farm programs, specifically the 1985 Food Security Act (FSA),
the 1990 Food Agricultural Conservation and Trade Act (FACTA), and the 1996 Federal Agricultural
Improvement and Reform Act (FAIR), are reflected in the econometric specification of the supply
component of the model, and thus are included in the simulation model.
Ex ante simulation of environmental policy will likely involve an assumption of continuation of the 1996
FAIR Act indefinitely. However, since most of the historical observations on which supply equations
were econometrically estimated occurred under different programs, it is important to consider how
historical equations reflect the 1996 FAIR Act. The basic philosophy that guided inclusion of farm
program features into the supply component of the model follow. First, beginning with the 1985 FSA,
continuing with the 1990 FACTA, and now with the 1996 FAIR Act, North American Free Trade
Agreement (NAFTA) and the General Agreement on Tariffs and Trade (GATT), farm and international
trade policy has moved U.S. agriculture to a market orientation. Although the 1985 FSA and the 1990
To the extent that the Rule increases diesel prices, shipping prices for some agricultural products may increase, and may cause some
farmers to change their production decisions. The magnitude of such an impact is likely to be small. Time and resources did not permit
modeling this possible impact on their decision-making.
Abt Associates Inc Appendix 1-1
-------
FACTA had price support and acreage diversion features, they embodied a strong market orientation. For
all major program crops (in AGSIM©), the acreage devoted to the crop exceeded the acreage under
government programs. Thus, at the margin, market prices (and not support prices) influenced crop
acreage. Another way of looking at this is that farm programs have influenced crops at the intra-margin,
while the market has influenced crops at the margin. Thus, after accounting for acreage diverted under
farm programs, expected prices determine acreage. For these reasons, AGSIM© should be valid for
simulating agricultural markets under the market conditions established under the 1996 FAIR Act.
Sets of equations that comprise the supply component of the current version of the model include: (1)
acreage planted to each crop, (2) acreage harvested of each crop, (3) acreage in annual set-aside or
acreage reduction programs (ARP) by crop, (4) acreage in cultivated summer fallow, (5) crop yields per
harvested acre, (6) rate of participation in Federal farm programs by crop, and (7) annual set-aside rates
by crop under past farm programs, as related to stock levels (historically legislated) and thus related to
market price. Identities in the model are: (a) production is the product of acreage harvested and yield per
harvested acre, and (b) the quantity supplied equals the quantity demanded for each commodity (market
clearing). Specification of each of these sets of equations follows.
Acreage Planted Equations. Acreage planted is the key behavioral relationship in the supply component
of the model. Acreage planted of a particular crop depends on expected per-acre net returns for that crop,
expected per-acre net returns for competing crops, and farm program variables. In algebraic (and Fortran)
form, the acreage planted equation is:
(1) acresp(ic,it,irun) = bc(ic)+ bap(ic)*acresp(ic,it-l,irun)+ bcrp(ic)*acrp(ic,it,irun) +
bdiv(ic)*acrediv + brm(ic)*rerntm(ic,it,irun) +
ber(ic)*rerentnp(it,irun) + byr(ic)*time(it) +
bd83(ic)*dumb83(it)
where:
acresp(ic,it,irun) = acreage planted to the icth crop in the itth year and in simulation
"irun",
acrp(ic,it,irun) = acreage of crop "ic" that was placed in the conservation reserve
program,
acrediv = acreage diverted under annual set-aside programs,
rerentm(ic,it,irun) = real expected per acre returns over variable costs for the icth crop,
rerentnp(it,irun) = real expected per acre returns over variables costs computed as a
weighted average54 of rerentm(ic,it,irun) over all endogenous
crops,
time(it) = a time-trend variable, and
dumb 8 3 (it) = a binary dummy variable to account for the PIK program in crop
year 1983.
The remaining variables in equation (1) represent estimated coefficients. A single run of AGSIM©
involves two simulations, one for the baseline (irun=0) and one for the policy scenario (irun=l). These
two simulations are then compared to estimate the economic impacts of the policy scenario.
Weights used in computing a composite expected return variable were the acreage harvested of each crop the previous year divided
by total acreage harvested the previous year.
Abt Associates Inc Appendix 1-2
-------
Expected returns over variable costs, rerentm(ic,it,irun), is defined as:
(la) rerntm(ic,it,irun) = rp(ic,it-l,irun)*ey(ic,it,irun) - rcost(ic,it,irun)
where:
rp(ic,it-l,irun) = real price the previous crop year (actual or simulated, depending on the
time period),
ey(ic,it,irun) = expected crop yield, and
rcost(ic,it,irun) = real variable production cost.
Expected yield is based on trend-line regressions:
(Ib) ey(ic,it,irun) = [cint(ic) + by(ic)*time(it)]
where:
cint(ic) and by(ic) are estimated coefficients.
In the policy run, expected yield is adjusted for exogenously specified percentage yield changes ("dyld"):
(Ic) ey(ic,it,irun) = [cint(ic) + by(ic)*time(it)]*(1.0 + dyld(ic,it)/100.)
Changes in real variable costs of production can also be exogenously specified for the policy simulation
run. Thus, yield and cost changes directly impact acreage planted through equation (1), and indirectly
impact acreage planted because of the resulting impact on prices in equation (la) and thus in equation (1).
Given signs and magnitudes of estimated coefficients in equation (1), an increase in expected returns of
the icth crop will increase acreage planted of that crop, while an increase in expected returns of other
endogenous crops will decrease acreage of the icth crop. The estimated coefficient on lagged acreage
planted in equation (1) is positive and less than one in value for all crops, which means that acreage
planted is dynamically stable. The estimated coefficient on the set-aside acreage is negative and less than
one in absolute value for all crops except oats, which reflects acreage slippage in the ARP program. Oats
were typically planted to set-aside acreage, thus the estimated coefficient on set-aside acreage is positive
in the oats equation, as expected. Further comments will be made on the acreage diverted effects on
planted acreage after participation rate and acreage diverted equations, which are endogenous, are
presented below.
Acreage Harvested Equations. Acreage harvested depends primarily on acreage planted:
(2) acresh(ic,it,irun) = bch(ic) + baph(ic)*acresp(ic,it,irun) + byrh(ic)*time(it) +
bdvh(ic)*acrediv
where:
acresh(ic,it,irun) = the acreage harvested of the icth crop in the itth year and in
simulation "irun",
and other variables are as defined previously.
The estimated coefficient baph(ic) is positive and less than one, indicating that not all planted acreage is
harvested, as expected. The coefficient bdvh(ic) on the acreage diverted variable is non-zero for oats
only, in which case it is negative. This adjusts oat acreage harvested for the complexity of oats being
planted (but not harvested) on ARP acreage. A time-trend variable for corn and grain sorghum, but not
Abt Associates Inc Appendix 1-3
-------
other crops shows how harvested acreage as a percentage of planted acreage has been increasing slightly
overtime.
Participation Rate in Farm Programs. Participation rates in the annual set-aside programs under the
1985 FSA and the 1990 FACTA were endogenized in the model with the set of equations:
(3) part(ic,it,irun) = bcp(ic) + brmp(ic)*rerntm(ic,it,irun) + brpp(ic)*rerntp(ic,it,irun) +
byr(ic)*time(ic) + bpart(ic)*part(ic,it-l,irun) + bedpp(ic)*redp(ic,it,irun)
+ bd83p(ic)*dumb83(it)
where:
part(ic,it,irun) = the participation rate in the farm program for the icth crop in the itth year
and in simulation "irun",
rerntp(ic,it,irun)= real expected returns over variable costs based on the support (target)
price for that crop,
redp(ic,it,irun) = real effective acreage diversion payment rate,
and other variables are as defined previously.
Estimated coefficients brpp(ic) are non-negative, indicating that an increase in expected returns based on
support price will increase participation, while estimated coefficients brmp(ic) are non-positive,
indicating that an increase in expected returns based on expected market price will decrease participation.
Lagged participation rate in equation (3) shows strong dynamics with respect to farm program
participation.
Acreage Diverted under Farm Programs. Acreage diverted under annual set-aside (or ARP) programs
is modeled as:
(4) adiv(ic,it,irun) = bcd(ic) + bd83d(ic)*dumb83(it) + bedpd(ic)*redp(ic,it,irun) +
byrd(ic)*time(it) + bpsa(ic)*sa(ic,it,irun)*part(ic,it,irun)
where:
adiv(ic,it,irun) = acreage diverted under annual diversion programs for the icth crop in the
itth year and in simulation "irun",
sa(ic,it,irun) = the set-aside rate specified by the Secretary of Agriculture under 1985
FSA and 1990 FACTA,
and other variables are as defined previously.
Acreage slippage (with respect to historical set-aside) in farm programs is implicit in the model
specification, and results from the complex simultaneity of farm program variables in sets of equations
(1), (3), and (4).
Acreage in Cultivated Summer Fallow. Acreage in cultivated summer fallow is modeled by the
equation:
(5) afl(it,irun) = bcfl + bafl*afl(it-l,irun) + berfl*rerentnp(it,irun) + byrfl*time(it) +
bd83fl*dumb83(it)
where:
afl(it,irun) = acreage fallowed in year it in simulation run "irun".
Abt Associates Inc Appendix 1-4
-------
Although the acreage in cultivated summer fallow is highly inelastic, this equation shows that an increase
in expected returns based on expected market price results in a small decrease in acreage fallowed.
Fruit & Vegetable Supply. Fruit and vegetable supply in AGSIM© is modeled as a set of linear supply
response equations. Supply depends on expected per acre returns, including dynamics.
Demand Components
The crop demand component of AGSIM© is based on a set of demand equations for each crop for
utilization categories of (a) imports, (b) exports,(c) livestock feed, (d) food, fiber, ethanol production and
other domestic uses, (e) ending stocks, and (f) residual use. Each demand component depends on current
market price for that commodity and, where relevant, prices of other commodities. The model
specification of each utilization category follows.
Imports. Imports of agricultural commodities are modeled by the set of equations:
(6) qd(ic,it,irun,l) = bim(l,ic)+ bim(2,ic)*rp(ic,it,irun)*xrate(ic,it-l,irun) +
bim(3,ic)*qd(ic,it-l,irun,l) + bim(4,ic)*time(it) +
bim(5,ic)*uspop(it,irun)
where:
qd(ic,it,irun, 1) = the quantity of crop ic imported in year it in simulation run
"irun",
rp(ic,it,irun) = real market price,
xrate(ic,it-l,irun) = the real trade-weighted exchange rate,
uspop(it,irun) = the United States population,
and bim(j,ic) are estimated coefficients. Lagged imports in equation (6) reflects dynamic adjustments.
Exports. Exports of agricultural commodities are modeled by the set of equations:
(7) qd(ic,it,irun,2) = bex(l,ic)+ bex(2,ic)*rp(ic,it,irun)*xrate(ic,it-l,irun)+ bex(3,ic)*
qd(ic,it-l,irun,2) + bex(4,ic)*time(it) + bex(5,ic)*wpop(it,irun)
where:
qd(ic,it,irun,2) = the quantity of crop ic exported in year it in simulation run "irun", and
wpop(it,irun) = world population.
Feed, Fiber and Crushing Use. Domestic utilization of crops for feed, fiber or crushing (depending on
the crop) is modeled by the set of equations:
(8) qd(ic,it,irun,3) = bfd(l,ic) + £jcbfdcross(ic,jc)*rp(jc,it,irun) + bfd(2,ic)*qd(ic,it-l,irun,3) +
bfd(3,ic)*time(it)
where:
qd(ic,it,irun,3) = utilization for feed, fiber or crushing.
Note that cross-price effects are incorporated into this set of equations through the set of estimated
coefficients bfdcross(icjc). Symmetry of cross-price effects, consistent with microeconomic theory, was
imposed on estimation so that bfdcross(icjc) = bfdcross(jc,ic) for ic * jc. Own-price effects are all
negative, as expected.
Abt Associates Inc Appendix 1-5
-------
Domestic Food Use. The set of equations to represent domestic food use is:
(9) qd(ic,it,irun,4) = bfo(l,ic) + bfo(2,ic)*rp(ic,it,irun) +bfo(3,ic)*qd(ic,it-l,irun,4) +
bfo(4,ic)*time(it) + bfo(5,ic)*uspop(it,irun) + bfo(6,ic)*rdincome(it,irun)
where:
rdincome(it,irun) = real per-capita disposable income in the United States,
and other variables are as defined previously. In the case of peanuts, the real market price is replaced by
the fixed quota price that applies to all domestically consumed peanuts. This quota price for peanuts
applies to the 1985 FSA, the 1990 FACTA, and continues with the 1996 FAIR Act.
Ending Stocks. Ending stocks are viewed as another component of demand. Although commodities are
often held to maintain pipeline inventories, commodities are also held for speculative purposes. Thus,
stock levels respond strongly to prices, so the stock relationships were specified and estimated as
(10) qd(ic,it,irun,5) = bst(l,ic) + bst(2,ic)*rp(ic,it,irun) + bst(3,ic)*qd(ic,it-l,irun,5) +
bst(4,ic)*time(it)
where qd(ic,it,irun,5) is ending stocks in year t.
Residual Use. For some crops (rice, peanuts, and cottonseed), supply and utilization data show a residual
category, which is modeled as,
(11) qd(ic,it,irun,6) = brs(l,ic) + brs(2,ic)*rp(ic,it,irun)+ brs(3,ic)*time(it)
where:
qd(ic,it,irun,6) = residual use.
Although quantities in this residual use category are never used, the level of the residual does respond
negatively to the real price, and is thus viewed as another utilization (demand) category.
Market Clearing Identities
In supply and demand specification outlined above, supply generally depends on past prices, while
demand depends on current prices. In simulating these econometrically estimated equations into the
future, simulated prices are solved by simultaneously solving the market clearing identities
(12) qs(ic,it,irun)+qd(ic,it-l,irun,5) = qd(ic,it,irun,l) + qd(ic,it,irun,2) + qd(ic,it,irun,3) +
qd(ic,it,irun,4) + qd(ic,it,irun,5) + qd(ic,it,irun,6)
where:
qs(ic,it,irun) = the quantity produced of crop ic in year it in simulation "irun".
Production is defined to be qs(ic,it,irun) = acresh(ic,it,irun)*ey(ic,it,irun). The left hand side of the equal
sign in (12) gives total supply (production plus beginning stocks), while the right-hand side of (12) gives
total utilization, including ending stocks.
In the simulation model this set of simultaneous equations are numerically solved to get the market
clearing prices in a given year. This process is continued, considering the dynamics of the model,
indefinitely into the future.
Abt Associates Inc Appendix 1-6
-------
Historical Observation Period
Many econometric relationships in the model were estimated with data for the 1975-1995 time period.
However, where structural change was apparent, such as with stock holding behavior and international
trade, some of the early years were dropped from statistical analysis so that the simulation model would
better reflect the future.
Alternative Specifications Considered
Many different specifications of how farm programs influence crop acreage have been considered in the
evolution of AGSIM©, including: (a) acreage depends on support price, (b) acreage depends on the
maximum of expected market price and support price, (c) acreage depends on a weighted average of
support and expected market prices, with weights based on program and non-program acreage of the crop,
and (d) acreage depends on expected market price. Models for expected market price have considered
complex distributed lags that go back several years in time, to a simple model that expected market price
is actual price the previous year.
Acreage equations have also been specified to depend on expected returns of: (1) all competing individual
crops with no parameter restrictions, (2) all competing individual crops with full symmetry of cross-
effects imposed on estimation, (3) major competing individual crops, and (4) a weighted average of all
expected returns for all other crops. Many different ways of incorporating participation rates and acreage
diverted into the model have also been considered. Several alternative functional forms (linear, log-
linear, nonlinear share equations, asymptotic) have also been considered.
Theoretical specifications considered have ranged from ad hoc models to very tightly specified and
detailed theoretical economic models based on complex assumptions. The present model draws from
economic theory (e.g. symmetry of cross-price effects in demand and homogeneity of degree zero of all
supply and demand equations with respect to prices), but does not specify the model so tightly with
untested assumptions and functional forms that empirical data has almost no role in the resulting
estimates. Alternative estimation techniques, ranging from simultaneous equations techniques, to
Zellner's seemingly unrelated regressions, to ordinary least squares regression have been used. The
current version of AGSIM© reflects a degree of subjective judgment of what best reflects supply and
demand of agricultural commodities based on microeconomic theory, traditional statistical criteria, and
substantive direct contact with farmers and ranchers in most regions of the United States.
Baseline
The current version of AGSIM© is designed to estimate changes in the agricultural sector resulting from
pesticide or other policy. Changes in economic variables are computed by comparing a policy simulation
of the model with a baseline simulation of the model. For ex post (retrospective) evaluations, the baseline
reflects actual farm programs, prices, acreages, etc. However, for ex ante evaluations, AGSIM© is
calibrated to an external baseline. The calibration is done by comparing an internally generated baseline
to the external baseline and computing adjusted intercepts for all of the relevant demand and supply
relationships in AGSIM©.
For the 1999 version of AGSIM© the externally specified year 2010 baseline is forecasted from the 2007
baseline reported by USDA (1988b). A few endogenous variables in AGSIM© were not included in the
USDA baseline. In those cases, the 1997 FAPRI baseline was used (FAPRI, 1997).
It should be noted that the baseline is not especially critical to estimates of changes in the agricultural
sector, except for the case of price support policy, which is not relevant here. That is, sensitivity analyses
with previous versions of AGSIM© have shown that estimates of changes in variables are not very
Abt Associates Inc Appendix 1-7
-------
sensitive to baseline absolute values of variables. Use of the USDA baseline to the extent possible
assures consistency with other governmental mandated agricultural policy analyses.
A USDA baseline was not available for specific fruit and vegetable commodities included in the present
version of the model. For commodities for which there was no USDA baseline, an internally generated
linear trend line based on historical values of the endogenous variables was used as a baseline. This
internally generated baseline for fruit and vegetables is included as part of the output from AGSIM©.
Regional Effects Sub-Model
AGSIM© subroutines are also available to combine AGSIM© output with production cost information to
estimate net farm income impacts for the policy scenario at the regional level (or farm, representative
farm, area or state level). Required information for this type of evaluation includes for each farm or area:
(a) yield and cost changes (which often differ from the national yield and cost changes for the policy
scenario), (b) baseline production costs, and(c) acreages of each crop. This information is combined with
price impacts estimated with AGSIM©, and regional supply elasticities from a prior version of AGSIM©
(or from other sources) to estimate net farm income changes for the farms or areas considered.
The conceptual foundation for regional evaluation in this version of AGSIM© begins with a net farm
income formula,
where:
IIir = net farm income in region ir,
Aic,ir = acreage harvested of the icth crop in that region, and
Ric,ir = per-acre net return in that region.
Based on equation (13), it can be shown that the theoretically appropriate formula for computing net farm
income changes for different regional situations is:
TI^ ^ A4., A/;,, y^,,
= " "'
where:
A represents a discrete change,
AZ represents the discrete policy change,
ic and jc are crop indices,
and other variables are as previously defined.
Equation (14) can be expressed in acreage elasticity (with respect to per-acre income) form,
(15) ATIr _ „ R,,,r ARrr _ ^Afi,g,r
-- — - -
where:
£ic,ij,ir
elasticity of acreage of the icth crop in the irth region with respect to per-
acre income of the jcth crop in that region.
The term AR1C)1J/AZ in equations (14) and (15) can be further expanded to give
Abt Associates Inc Appendix 1-8
-------
(16) AS^£
AZ " "'" "'" Ii:
Formula (15) along with (16) can be empirically implemented to estimate the change in regional (or farm,
representative farm, area or state level) farm income with the following information for each region: (a)
crop budgets, (b) the change in yield and cost associated with the policy in question, price impacts
estimated with AGSIM©, and externally specified (from an older version of AGSIM©, from subjective
estimates, or from the literature) elasticities.
The first term on the right-hand side of (14) and (15) represents the change in net income resulting from
increased or decreased acreage, while the last term on the right-hand side of (14) and (15) represents the
change in net farm income on existing acreage of crops in the region. Since acreage response is generally
inelastic, the last term on the right-hand side of (14) and (15) dominates the change in net farm income in
a region; thus, elasticities generally will not have a major impact on regional net farm income changes
estimated with the above approach.
AGSIM© Output
The major outputs from AGSIM© are changes in crop acreage, production, price, income, foreign
consumer benefits, domestic consumer benefits, and farm program costs. The traditional method of
economic welfare analysis (which is based on the concept of economic surplus) of policy changes is used
to compute the sum of changes in producer surplus (net farm income) plus changes to all consumers
(changes in consumers surplus) plus any changes in farm program payments (zero under 1996 FAIR). To
avoid the possibility of inappropriately comparing a baseline with a policy scenario that was actually
based on another baseline, a single run of AGSIM© produces both the baseline tables and the policy
scenario tables, then computes economic surplus and price changes based on these two runs of the model.
Output from each run of the model includes two sets of tables for each crop; one set of tables for supply
variables and another set of tables for supply and utilization variables. Each table includes historical
statistics as well as simulations into the future. These tables are constructed for the baseline and the
policy scenario.
Uncertainty
From a theoretical viewpoint, the types of uncertainty about results from an econometric-simulation
model like AGSIM© run the full gamut from specification bias to estimation bias to measurement bias to
functional form bias. Since these potential biases are covered extensively in econometrics texts, they are
not be repeated here. From a practical standpoint, much of the uncertainty about estimated economic
impacts can be viewed in terms of uncertainty about estimated demand and supply elasticities. Generally,
the demand and supply elasticities in the present version of AGSIM© are within the range of elasticities
reported in the literature for the same commodities. Furthermore, estimated elasticities (or estimated
coefficients on which elasticities are based) are generally highly significant.
The AGSIM© simulation model is keyed to the USDA baseline. Although the USDA gives point
estimates of relevant endogenous and exogenous variables, there is nevertheless some uncertainty about
these future values. This uncertainty about the baseline has not, to our knowledge, been quantified. Thus,
uncertainty about the baseline cannot be quantitatively translated into uncertainty about economic impacts
estimated with AGSIM©. Qualitatively, however, the estimated economic impacts are not highly
sensitive to the baseline because they are computed as changes in economic variables.
Theoretically one could compute an overall goodness of fit statistic for the model, but such a statistic
would be essentially meaningless since the statistical properties would be largely unknown. More
Abt Associates Inc Appendix 1-9
-------
importantly, while one can theoretically compute such statistics for large-scale models, they cannot be
compared for different models due to fundamental differences in the structure of alternative large-scale
models and due to different sets of endogenous variables from model to model.
Another unresolved theoretical issue pertains to the combined effects of uncertainty about AGSIM© and
uncertainty about yield and cost estimates provided by others. This is a very messy and complicated issue,
particularly when one appropriately considers non-zero covariances between economic variables and crop
yield variables.
Yet another source of uncertainty relates to implications of massive consolidation and integration of the
agricultural sector that has occurred in the past two decades. AGSIM© implicitly assumes competition; to
the extent that imperfect competition exists, the econometric results are somewhat biased as is the
theoretical interpretation of economic surplus.
At best, we can only subjectively assess uncertainty about quantitative results from a large-scale model
like AGSIM©. Based on over 30 years of developing and applying large-scale models including national
programming models and econometric-simulation, our subjective assessment is that there is a modest
amount of uncertainty about the AGSIM© results given changes in yield and cost. Overall, our subjective
estimate is that estimated changes in economic surplus are within 50% of their true values, but estimated
effects are more uncertain for some commodities and less uncertain for others.
Abt Associates Inc Appendix 1-10
-------
Appendix J TREGRO simulations of red maple and
yellow poplar trees under scenarios of reduced ozone
exposure at two locations in the southern Appalachian
Mountains
REPORT TO ABTASSOCIATES, INC.
D. A. Weinstein, Ph.D.
May 18, 2006
The response of total tree growth of two species, red maple and yellow (or tulip) poplar was simulated in
each of two locations in the southern Appalachian Mountains to five scenarios of ozone (03) reduction
provided by Abt Associates. These simulations were done using the computer model, TREGRO
(Weinstein et al. 1991). This report provides the details of the methodology and results from this
examination.
Procedure
1. The TREGRO model was used to simulate the growth of a single mature yellow poplar (Liriodendron
tulipifera L.) tree over three years under climate conditions characteristic of the cove hardwood forests
(USDA Forest Type, Hansen et al., 1992) of the Cranberry region of North Carolina, east of Great Smoky
Mountains National Park (Table 1) and the cove hardwood forests of the Shenandoah National Park
region of Virginia. Yellow poplar was chosen for analysis because controlled studies and previous
analysis indicated a sensitivity to ozone (O3) (Cannon et al. 1993). These sites were selected, in
consultation with personnel at Abt and EPA, because previous analysis indicated yellow poplar growing
at those sites was sensitive to OT, exposure (Weinstein et al. 2002). The cove hardwood stands are widely
viewed as some of our most treasured forests because their protected, rich, and moist set of conditions
historically permits trees to grow to magnificent size with very high growth rates. Yellow poplar is one
of the most abundant species in the southern Appalachian forest, and comprises approximately 10% of the
cove forest (unpublished data from USFS FIA Eastwide Database, Hansen et al. 1992). The simulations
were done using a parameter set established previously based on the following method:
The parameterization for yellow poplar (tulip poplar; (Liriodendron tulipifera L.)) was originally reported
in Weinstein et al. 2001. Parameters were established to permit TREGRO to simulate the growth of a
mature yellow-poplar tree of approximately 30 m in height, 40 cm in diameter, and 50 years old (Beck
1990, Clark and Schroeder 1977). Initial biomass of individual tree components (foliage, branch, stem,
and coarse and fine root) was calculated using diameter at breast height (dbh)-based allometric equations
derived by Clark and Schroeder (1977). Biomass of branch, stem and coarse roots was divided into
structure (living) and wood (dead) according to Panshin and de Zeeuw (1980). This distinction is
important because wood cannot be used to store reserve carbon that can be drawn on to meet plant needs
when the supply of newly fixed carbon coming from photosynthesis falls to insufficient levels. The
reserve carbon, or total nonstructural carbohydrate (TNC), was assumed to represent 30% of structural
Abt Associates Inc Appendix J-l
-------
mass in the stem, branch, and roots, and 20% of structure in the foliage based on reported starch
concentrations in roots (Jensen and Patton 1990) and TNC concentrations in foliage (Wullschleger et al.
1992). This percentage in the structural tissue, therefore, establishes an upper limit of the quantity of
reserve carbon the tree can maintain.
Leaf growth was set to be initiated on April 7 and last until May 7 (Britton 1878, Kienholz 1941, Lamb
1915) at which point height and radial growth were started (Kienholz 1941, Morrow and McKee 1963,
Mowbray and Costing 1968), continuing until approximately September 10th (Lieth and Radford 1971,
Morrow and McKee 1963). Leaf fall was set to occur at the end of the second week in October (Lamb
1915). Maximum net carbon assimilation rate was set at 0.01718 gOgC leaf~l*hr~l (7.67 umol*m~2«s~l)
(Cannon et al. 1993, Chappelka et al. 1988, Gunderson et al. 1993, Neufeld et al. 1985, Norby and O'Neill
1991, Tjoelker and Luxmoore 1992, Wullschleger et al. 1992) to approximate the conditions found on a
midsummer day (PPFD > 1500 umol«m~2«s~l; ambient air temperature 25 - 30° C). Leaf respiration rate
was set to approximate a respiration-to-net photosynthesis ratio of 0.2215 (mean value from Cannon et al.
1993 and Wullschleger et al. 1992) and finalized at 0.04% of net carbon assimilation rate. Both flushes of
leaves produced in a given year were assumed to have identical photosynthetic and respiratory rates.
Trees of this dbh typically exhibit an annual dbh increment of 0.30 cm yr'l and a height increment of 23
cm yr'l under the environmental conditions that existed on average from 1940 to 1990 (Beck 1990).
Calculating the expected size of tissues from the previously described allometric relationships, the amount
of growth expected by each type of tissue was estimated by subtracting their estimated initial size before
this three year growth period. The maximum potential rate of growth of each tissue was then adjusted
until the simulated tree predicted the appropriate growth for each type of tissue, with the proportion of
TNC, structure, and wood in each of the tree components remaining unchanged. The fine root senescence
was set at the maximum rate possible given excess available carbon.
2. The TREGRO model was used to simulate the growth of a single mature red maple (Acer rubrum L.)
tree over three years under climate conditions characteristic of the mixed central hardwood forests
(USDA Forest Type, Hansen et al., 1992) of the Cranberry region of North Carolina, east of Great Smoky
Mountains National Park (Table 1) and the mixed central hardwood forests of the Shenandoah National
Park region of Virginia. Red maple was chosen for analysis because controlled studies and previous
analysis indicated a sensitivity to O3 (Samuelson 1994). These sites were selected, in consultation with
personnel at Abt and EPA, because previous analysis indicated red maple at these locations was sensitive
to 03 exposure (Weinstein et al. 2002). Red maple is one of the most abundant species in the eastern
forest, comprising between 10% and 20% of the forests studied (unpublished data from USFS FIA
Eastwide Database, Hansen et al. 1992). The simulations were done using a parameter set established
previously based on the following method:
The parameterization for red maple (Acer rubrum L.) was originally reported in Weinstein et al. 2001.
Parameters were established to permit TREGRO to simulate a mature red maple tree based on the
diameter at breast height (dbh = 26 cm) of dominant and codominant trees measured by Erdmann et al.
(1985) in upper Michigan as part of a crown release study of red maple. The age of a tree at this dbh is
approximately 50 years old (Erdmann et al. 1985), with a height of approximately 19 m (Erdmann et al.
1985), and mean crown radius of 3.34 m (Oilman 1988).
Initial biomass of individual tree components (foliage, branch, stem, and coarse and fine root) was
calculated using dbh-based allometric equations derived from red maple trees originally growing in
Maine and New Hampshire (Young et al. 1980, Hocker and Early 1983). In the absence of available data,
the proportion of wood in the initial tree's stem, branches, and coarse (woody) roots was set to be 20%
based on a value determined for the stem of sugar maple by Chapman and Gower (1991). The remainder
of the initial biomass was split between structure and TNC using the assumption that TNC was 30% of
Abt Associates Inc Appendix J-2
-------
structure in the stem, branch, and roots and TNC was 20% of structure in the foliage, using the same
assumption reported above for yellow poplar (Jensen and Patton 1990; Wullschleger et al. 1992).
Seasonal development of red maple was set to begin in the late winter/early spring (approximately March
1st - May 1st) with flower bud swell and bloom, with foliage bud break and foliage flush occurring in the
early spring (approximately May 1st - May 22nd), and with height and radial growth beginning soon after
foliage bud break and continuing until approximately July 15th (Walters and Yauney 1990). Foliage
senescence occurred in mid-October (approximately October 15th) (Lamb 1915). Net carbon assimilation
under high light conditions (1000 (iE m~2 s~l; 25 - 30 C, midsummer day, day-of-year 200-210) was set
to approximately 0.00776 gOgC leaf~l*hr~l based on values recorded by Reich et al. (1991). Leaf
respiration was set at 14% of gross carbon assimilation based on values recorded by Kloeppel et al.
(1993).
Red maple trees of this dbh typically exhibit diameter growth of 3.6 cm in seven years in the absence of
O3 (Erdmann et al. 1985). Calculating the expected size of tissues from allometric relationships, the
amount of growth expected by each type of tissue was estimated by subtracting their estimated initial size.
The maximum potential rate of growth of each tissue was then adjusted until the simulated tree predicted
the appropriate growth for each type of tissue, with the proportion of TNC, structure, and wood in each of
the tree components remaining unchanged. The fine root senescence was set at the maximum rate
possible given excess available carbon.
3. An O3 response was established for both species. The TREGRO model calculates the hourly uptake of
O3 through the stomata as a function of stomatal conductance and cumulative uptake over the leafs
lifespan. In the model, the potential maximum rate of photosynthesis during any given hour is reduced in
direct proportion to the cumulative uptake of OT, over the course of the growing season, based on work by
Hansen et al. (1994). Although 03 does not accumulate in the leaf tissue, the effect of cumulative OT,
uptake is proportional to the sum of the hourly 03 concentration (exposure) multiplied by the foliar rate
of 03 stomatal conductance. The slope of the described response was set so that the simulated reduction
in photosynthesis matched the photosynthesis reduction and cumulative 03 exposure observed at the end
of an open-top chamber exposure experiment.
The red maple response to 03 was based on the work of Samuelson (1994), who reported a reduction in
current leaf net photosynthesis of seedlings of 25% relative to charcoal filtered air after a total cumulative
03 exposure of 175,000 ppb'hrs (a simulated uptake of 0.0138 g 03 g"l leaf C), which amounted to a
14.3% drop in photosynthesis for every 100 ppm-hrs 03 exposure. The yellow-poplar response to 03
was set to match the results reported by Cannon et al. (1993), who measured a reduction in net
photosynthesis of seedlings of 10% relative to charcoal-filtered air after a total cumulative 03 exposure of
75,600 ppb'hrs (a simulated uptake of 0.0044 g 03 g"l leaf C), a 13.2% drop in photosynthesis for every
100 ppm-hrs 03 exposure. It was assumed that all trees of a given species would experience the same
average reduction in photosynthesis in response to 03 as was reported in the aforementioned studies.
Mature trees were assumed to have the same 03 exposure responses as those measured in experiments
with seedlings. Experiments by Samuelson and Edwards (1993) and by Hanson et al. (1994) have
demonstrated that mature red oak tree leaves are more sensitive to 03 than seedlings. However, no data
from controlled exposures of mature yellow poplar or red maple were available. Consequently, the data
of Samuelson (1994) and Cannon et al. (1993) were the only relevant studies on which to draw. At worst,
the use of this data to represent mature tree responses gives a conservative estimate of the actual tree
response.
Abt Associates Inc Appendix J-3
-------
4. Meteorology input for TREGRO consisted of base files for the hourly conditions of the period 1993-
1995.
5. Air quality input for TREGRO consisted of 63 scenarios provided by Abt Associates, which
represented rollbacks of hourly 63 values for 1993-1995 to meet current and alternative standards.
Growth of each tree species was simulated for three years to account for the accumulative effect of
repeated injury. For some of the scenarios 03 levels did not exceed the defined standard for each year;
therefore, theoretically the trees were not injured during every year of the analysis. The hourly 03 values
in each scenario were used in place of the 03 values in the original base meteorology data. Therefore, the
simulations were run with weather variables for 1993-1995 and 03 values from each of the scenarios.
6. TREGRO simulations were run for each scenario of 63 reduction supplied by Abt Associates:
1) Scenario 1. Rolled back hourly 03 values for 1993-1995 to the current EPA 03 standard (as
expressed in western TREGRO analysis) as 4th highest daily maximum 8-hour average, not to exceed
0.085 ppm). The 8-hour mean of 0.085 ppm was exceeded only at the Big Meadows site in
Shenandoah National Park, Virginia, and only for the years 1993 and 1995; therefore, the quadratic
rollback (Rizzo 2006) was only performed on this site for these two site-years.
2) Scenario 2. Rolled back hourly 03 values for 1993-1995 to a SUM06 of 25 ppm-hours cumulated
over a consecutive 3-month period during the 12 (8:00 am to 8:00 pm) daylight hours.
3) Scenario 3. Rolled back hourly 03 values for 1993-1995 to the 1st highest daily maximum 8-hour
average, not to exceed 0.070 ppm.
4) Scenario 4. Rolled back hourly 03 values for 1993-1995 to the 4th highest daily maximum 8-hour
average, not to exceed 0.070 ppm.
5) Scenario 5. Rolled back hourly 63 values for 1993-1995 to a SUM06 of 15 ppm-hours cumulated
over a consecutive 3-month period during the 12 (8:00 am to 8:00 pm) daylight hours.
Results
The simulations produced a prediction of average annual total tree growth over the 3-year period for each
scenario. These results were compared to the base scenario, which consisted of a prediction of growth
under the hourly meteorology and 03 conditions for the period 1993-1995.
The predictions indicated substantial increases in 3-year total tree growth increments with reduction of 03
exposure, particularly under Scenario 3, a rollback to conform to the standard of the 1st highest maximum
8-hour average being no greater than 0.070 ppm. Yellow poplar had nearly a twenty percent increase in
growth in response to this scenario, an average annual increase of 6.5%.
Table J-l Predicted percent increases in total tree growth over a 3-year period under the 4 ozone (O3)
reduction scenarios.
Red maple, Yellow
Red maple, Cranberr Yellow poplar, poplar,
Shenandoah y Shenandoah Cranberry
Abt Associates Inc Appendix J-4
-------
Scenario_l
Scenario_2
Scenario_3
Scenario_4
Scenario 5
1.22%
1.02%
8.14%
6.72%
4.49%
6.92%
4.15%
2.99%
0.08%
0.20%
1.15%
1.03%
0.60% 8.26%
19.61%
11.73%
Scenario 1: Rollback current EPA standard 4th highest max. 8-hr avg. 0.085 ppm
Scenario 2: Rollback SUM06 25 ppm-hr
Scenario 3: Rollback 1st highest max. 8-hr avg. 0.070 ppm
Scenario 4: Rollback 4th highest max. 8-hr avg. 0.070 ppm
Scenario 5: Rollback SUM06 15ppm-hr
Figure J-l Tree growth response of red maple and yellow poplar in forests of Shenandoah National Park,
Virginia and Cranberry, North Carolina to ozone (O3) reduction scenarios.
20%
Scenario_l
IScenario_2
Scenario_3
Scenario_4
I Scenario 5
Red maple was simulated to have a similar response to Scenario 3 in Shenandoah and in Cranberry.
However, it had nearly twice the increase to Scenario 4 at Shenandoah as it did at Cranberry. The
response to Scenario 5 was slightly less than Scenario 4 for red maple and for yellow poplar in both
Shenandoah and Cranberry. The response of yellow poplar at Cranberry to Scenario 5 was still very
large, with growth projected to increase more than 8% under this level of O3 reduction.
Yellow poplar had a very different response to 03 reduction at Shenandoah compared to Cranberry. The
temperatures at Cranberry are more in the middle of the range of temperatures over which yellow poplar
is found than are the cool temperatures of Shenandoah, making conditions at Cranberry more ideal for
growth. Higher growth rates may cause greater sensitivity to 63. Red maple has a much larger
geographical distribution, so that the temperature differences between Shenandoah and Cranberry are less
likely to affect the growth response. This phenomenon was reflected in the simulations.
Finally, Scenarios 1 and 2 produced very little growth response in either species. These scenarios
produced no change in the predicted 63 exposure at Cranberry, so they were not even simulated. At
Shenandoah, the change in 03 exposure to Scenarios 1 and 2 was very slight.
Abt Associates Inc
Appendix
J-5
-------
Uncertainty
Any simulation result is dependent on the accuracy with which the parameters used in the model can be
estimated. In a model with as many parameters as TREGRO, it is nearly impossible to conduct a
comprehensive uncertainty analysis, in which each parameter is allowed to vary throughout it's potential
distribution to assess the impact of different values on the model's predictions. Despite the absence of an
uncertainty analysis, it is also incorrect to assume that in a deterministic model such as TREGRO (with
no stochastic elements) all parameters were correctly estimated and therefore the prediction is the only
one possible. Two trees of the same species, identical size, and growing under the same conditions can
vary in growth rate considerably.
The predictions of the response to 03 are dependent on the relationship used in the model between
cumulative exposure and photosynthesis reduction. This relationship, established in a controlled chamber
experiment, had only a small amount of variability surrounding it caused by within experiment variability.
The amount the relationship might vary when environmental conditions are changed from those of the
experiment to those of the Cranberry or Shenandoah forests is unknown. However, the relationship does
not seem to be very sensitive to these types of shifts in environmental variables. The principle variable
that appears to affect the responsiveness of growth of a tree to 63 is the historical growth of that tree
under a given set of environmental conditions. The TREGRO model explicitly incorporates this effect by
calculating the energy and carbon balance of the tree. A tree with a poorer energy balance, i.e. one where
energy demands do not greatly exceed energy supplies, has proven to be more susceptible to injury from
03 in TREGRO (Weinstein et al. 2001), and this appears to mirror patterns observed in real forests.
The effect of 03 changes as a tree continues to be exposed over a succession of years. TREGRO attempts
to account for some of this effect by simulating over three consecutive years. However, as stated above,
the effect of 03 is larger in a tree that is in a poorer energy balance, and a year of significant 03 exposure
will place a tree into a poorer energy balance for the next year. Therefore, if 03 levels remain high for
many years in a row, the effect of a given level of 03 will increase with each year. The choice of three
years attempts to capture the nature of this effect, but simulating five or ten successive years of high 03
exposure would lead to a prediction of higher average annual effects. Historically, the variability in year-
to-year 03 exposure has made it unlikely that there would be a period of high 03 this lengthy without
intervening years of lower 03.
Abt Associates Inc Appendix J-6
-------
Appendix K Report from the Preliminary Evaluation
of O3 Interpolation Approaches using 53 monitored
dropout sites
This appendix contains the full text of a memorandum delivered by to EPA by Abt Associates on 4-25-
06, detailing the preliminary investigation of 10 interpolation approaches being considered. Since this
early investigation was redone in greater depth for the writing of the current report, much of the material
in this appendix appears earlier in the body of this report. Specifically, material coming before section
K.2.4 has already appeared in the body of this report.
K.l. Introduction
During the last review of the ozone (O3) NAAQS, as part of the development of the 1996 O3 Staff Paper
(SP), EPA conducted analyses that assessed national O3 air quality, vegetation exposures and risk, and
impacts to economic benefits. At the time of the last review, large rural sections of the country had little
or no monitor coverage, including important growing regions for agricultural crops and forested
ecosystems. Since O3 monitor coverage in agricultural and rural/remote sites has changed little since the
last review, EPA must again rely on generated O3 air quality in non-monitored areas to provide national
O3 exposure coverage. Given a number of recent air quality related developments, including the
refinement of the Community Multiscale Air Quality (CMAQ) model, inclusion of a spatial interpolation
tool in EPA's Environmental Benefits Mapping and Analysis Program (BenMAP), and updated monitored
air quality, EPA has decided to use a different method to generate a national exposure surface in this
review.
In this memorandum we evaluate approaches for generating a national O3 surface, based on their ability to
make predictions for sites where we have O3 monitoring data. To test these approaches, we chose a
number of "dropout" monitors across the country (see section K.2.4 for details), removed them from our
data set, and used each of the approaches to make O3 predictions for the dropout locations using the
remaining data. We then compared these predictions to the actual data from the dropout monitors to
evaluate the effectiveness of the different approaches.
The approaches we considered are summarized in Section K.3. Our results suggest that enhanced
Voronoi Neighbor Averaging (eVNA) offers substantial improvements over the CMAQ modeling data,
and in certain cases, substantial improvements over the traditional VNA approach (as used in BenMAP)
as well. In the results section (Section K.3) we provide a range of quantitative measures that underlie our
evaluation.
K.2 Methods
Below we describe the data that we used in our analysis, how we processed the data, and finally the
approach that we used to evaluate our results.
Abt Associates Inc Appendix K-l
-------
K.2.1 Monitor Data
The monitor data used in this analysis was taken from the Air Quality System (AQS) and Clean Air
Status and Trends Network (CASTNet) for the year 2001. AQS O3 data was taken from the file
RD_501_44201_2001.zip, and information on the monitors was taken from the file
AMP500_1994_FEB05.zip. Both are available at
http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm. CASTNet O3 data was taken from
www.epa.gov/camdisOI/prepack/ozone 2001 .zip, and information on CASTNet monitors can be found at
http://cfpub.epa.gov/gdm/index.cfm.
Completeness Criteria
Data from a given monitor was only used if the monitor was deemed "complete", i.e. if it had valid O3
values for at least 50 percent8 of the hours during its region's O3 season. Note that O3 seasons vary by
geography and range from year-round (in California) to periods as short as June-September (Montana). In
all states except Texas, O3 regions follow state boundaries. Texas is unique in that it contains parts of 2
different O3 regions, each with its own O3 season. To simplify matters, we have applied the shorter of
these seasons to the entirety of Texas.
Definition of Os Metrics (SUM06 and Annual 4th Highest Daily Maximum 8-hour Average)
For hourly O3 data, we considered two O3 metrics (SUM06 and 8-hour maximum average) that can be
calculated on the daily-level ( "SUM06" and "daily 8-hour maximum average") or the yearly-level
("annual maximum 3-month SUM06" and "annual 4th highest daily maximum 8-hour average").
The daily SUM06 metric is the sum of all O3 values greater than or equal to 0.06 parts per million (ppm)
observed from 8am-8pm. In order for a day to have a valid SUM06 value, 75 percent of the hours from
8am-8pm must be valid. To adjust for missing hourly O3 values, we scale SUM06 by the ratio of (number
of possible hourly O3 values) / (number of valid hourly O3 values).1
The yearly SUM06 metric is the annual maximum 3-month SUM06. To compute it, we calculate the sum
of all daily SUM06 values over all possible 3-month periods. And then to adjust for missing days, we
scale each monthly SUM06 value by the ratio of (number of days in the month) / (number of valid days)."
The greatest of these 3-month SUM06 values is the annual maximum 3-month SUM06. In order for a 3-
month period to have a valid "3-month SUM06" value, each month in the 3-month period must have at
least 75 percent valid days. In order for a year to have a valid yearly SUM06 value (annual maximum 3-
month SUM06), it must have at least one 3-month period with a valid 3-month SUM06 value.
s For example, if the ozone season were May-September, then a valid monitor would have to have at least 1,836
hourly observations out of a potential total of 3762 (= 153 days x 24 hours). Out of 1,194 monitors, all but two
of them have at least 50 percent valid readings during their O3 season. We are considering raising this threshold
to 75 percent. This would eliminate an additional 79 monitors, leaving about 93 percent of the original monitors
remaining.
4 When interpolating hourly ozone values (as opposed to daily ozone metrics), we do not follow this exact method.
Instead, we apply a scaling factor at the monthly level (valid observations during the month) / (12hours x
number of days in the month). This simplifies the process and does not alter the quantitative analysis in any
meaningful way. For further explanation, see subsection in Section 2.3, Interpolating Hourly Data vs. Metrics.
u This is done when interpolating metrics, but not when interpolating hours. See previous footnote for details.
Abt Associates Inc Appendix K-2
-------
The daily 8-hour maximum is calculated from rolling 8-hour averages of hourly O3 data, where a valid 8-
hour average must have 75 percent of a potential of eight hours in any given 8-hour period (i.e., at least
six hours out of eight) .v The daily 8-hour maximum is the greatest of the day's 8-hour averages. Note that
for a daily maximum to be considered valid, the day must have at least 75 percent of the potential 8-hour
averages (i.e., 18 out of a potential of 24).w
The yearly metric associated with the 8-hour maximum is the annual 4th highest daily maximum 8-hour
average. This is defined to be the 4th highest value amongst all of the valid daily 8-hour maximums
throughout the year.
K.2.2 Model Data
We used two CMAQ modeling datasets, one with a resolution of 12km x 12km, the other with a
resolution of 36km x 36km. The 12-km CMAQ grid consists of 188 x 213 cells covering the eastern U.S.
(bounded approximately on the west by the 99 line of longitude) excluding the northernmost parts of
Maine, Wisconsin, Minnesota, and South Dakota, and the southernmost parts of Florida and eastern
Texas. The 36-km CMAQ grid consists of 112 x 148 cells covering the entire continental US. Each
dataset gives hourly O3 values for each cell.x
K.2.3 Interpolation Approaches
We used two interpolation approaches: Voronoi Neighbor Averaging (VNA) and enhanced Voronoi
Neighbor Averaging (eVNA). The former is based only on monitor data, and the latter uses both
monitoring and modeling data. We describe each below; in addition, Appendix C (above) provides a
detailed numerical example.
Voronoi Neighbor Averaging (VNA)
VNA uses distance-weighted averages of neighboring monitor data to arrive at predictions for a dropout
site. It identifies neighboring monitors for each dropout site using a Voronoi Neighbor Algorithm, and
takes an inverse-distance-weighted average of each neighbor's value for the data point in question (hourly
O3 value, daily metric, etc) to arrive at a prediction for that data point corresponding to the dropout site in
question.
Enhanced Voronoi Neighbor Averaging (eVNA)
v When there are six or seven hours in any given 8-hour period, we sum the available hourly measurements and then
divide by the number of available measurements (as opposed to always dividing by eight).
w There are 24 possible 8-hour measurements in a day, starting with 12:00 midnight to 7:59 am, and going through
11:00 pm to 6:59pm. This allows 8-hour measurements to straddle days, following the approach in the Federal
Register (40 CFR, Part 58). We consider an 8-hour measurement to be part of a day if it starts during that day.
So Ilpm[dayl]-6am[day2] is part of dayl.
x All of the CMAQ data was provided by EPA in netCDF (Network Common Data Form) format. Steve Howard
from EPA provided a program to convert from netCDF to text.
Abt Associates Inc Appendix K-3
-------
The eVNA approach attempts to improve the accuracy of VNA predictions by taking into consideration
modeling predictions for the areas involved. To illustrate the rationale behind eVNA, we consider a
simple fictional example.
Suppose we wish to predict the O3 level at a hypothetical monitor at location X for a given hour. Location
X has two equidistant neighboring monitors, monitor A and monitor B. Monitor A reports 32 ppb, and
monitor B reports 20 ppb. A simple VNA approach would calculate the O3 at location X to be 26 ppb (the
average of 32 and 20, with equal weights given to the two equidistant neighbors).
Suppose, however, that CMAQ modeling data shows O3 levels at location X to be about twice that of O3
levels at either location A or location B. For example, suppose the average CMAQ O3 values for locations
A and B are 15 ppb, whereas average CMAQ O3 values for location X are 30 ppb. If CMAQ accurately
captures the relationship between locations A, B, and X, then we would expect the O3 value to be twice as
high at X, compared to A and B. That is we would expect a value closer to 52 ppb - the weighted-
distance average of 2*32 ppb and 2*20 ppb. The eVNA approach formalizes this technique over large
and more complicated sets of data.
Unlike VNA, which averages the "raw" monitor predictions, eVNA first "adjusts" the individual monitor
predictions, multiplying by an adjustment factor that reflects the relationship between the neighbor's and
the dropout location's O3 levels, as determined by modeling data. For example, if the modeling data
suggested that O3 levels at neighbor A were generally twice as high as at the dropout location, and that O3
levels at neighbor B were generally half as high as the dropout location, we would multiply neighbor A's
O3 value by 1A and multiply neighbor B's O3 value by 2 before proceeding to take a distance weighted
average over the two neighbors.
Four Approaches to Condition-Specific Adjustment Factors
The eVNA approach, as we have described it thus far, is imperfect in that it assumes the O3-level
relationship between two locations to be constant throughout the year. In fact, the relationship may vary
with the season, or with the time of day, or with numerous other factors. To take this into account, we
have added an additional layer of complexity. Rather than condensing a year's worth of model data into a
single relationship between the O3 levels of two locations (and thus a single adjustment factor for each
neighbor-dropout pair), we determine the relationship for a number of different conditions. This allows us
to tailor our adjustments to the conditions at hand.
If we are adjusting a monitor value in January, we can use an adjustment factor that specifically reflects
the modeled relationship between the locations at hand during the month of January. Similarly, if we are
adjusting an O3 value that is particularly high, we can use an adjustment factor that describes the general
modeled relationship between the locations in question when O3 levels are high.
One can imagine many such ways to divide the data into subsets that reflect the particular conditions of
the data in question. We have chosen four such divisions, herein referred to as "approaches", which we
outline below. Each approach represents a separate and distinct effort to generate an O3 surface; i.e. each
of these four approaches can each be applied separately to the data to yield a different set of O3
predictions.
Month-Decile
We first sort CMAQ modeled hourly values into 12 groups by month. In each month-group we
split evenly the ordered hourly values into ten rank-ordered deciles. This gives us 120 groups of
hourly O3 values (for every CMAQ grid cell). We calculate the average of the hourly values in
each gridcell-month-decile combination, and use this average as the "representative value" of that
Abt Associates Inc Appendix K-4
-------
CMAQ grid cell for that month-decile. In order to adjust a neighboring monitor value to reflect
the modeled relationship to the unmonitored or dropout site, the appropriate month-decile
adjustment factor must be used. For example, to adjust a monitor value that falls into the 10th
decile of January monitor values, we multiply by the ratio of [the representative value of the
dropout's gridcell forthe 10th decile of January] over [the representative value of the neighbor's
gridcell forthe 10th decile of January].
,. , ., , „ representativeCMAQd gndcellmonthtdeale
adjusted monitor value = monitor _valueneigKbor *
representativeCMAQnei
ignor_gndcellimontMecne
Season-Decile
We first sort CMAQ modeled hourly values into four groups by season (Jan-Mar, Apr-Jun, July-
Sep, Oct-Dec). In each season-group we split evenly the ordered hourly values into ten rank-
ordered deciles. This gives us 40 groups of hourly O3 values (for every CMAQ grid cell). We
calculate the average of the hourly values in each gridcell-season-decile combination, and use this
average as the representative value of that CMAQ gridcell for that season-decile. In order to
adjust a neighboring monitor value to reflect the modeled relationship to the unmonitored or
dropout site, the appropriate season-decile adjustment factor must be used. For example, to adjust
a monitor value that falls into the 10th decile of the Jan-Mar monitor values, we multiply by the
ratio of [the representative value of the dropout's modeled O3 data for the 10th decile of Jan-Mar]
over [the representative value of the neighbor's modeled O3 data forthe 10th decile of Jan-
March].
Month-Hour
We first sort CMAQ modeled hourly values into 12 groups by month. In each month-group we
split evenly the ordered hourly values into 24 groups by time of day. This gives us 288 groups of
hourly O3 values (for every CMAQ grid cell). We calculate the average of the hourly values in
each gridcell-month-hour combination, and use this average as the representative value of that
CMAQ grid cell for that month-hour. In order to adjust a neighboring monitor value to reflect the
modeled relationship to the unmonitored or dropout site, the appropriate month-hour adjustment
factor must be used. For example, to adjust a monitor value from 9am in the month of January,
we multiply by the ratio of [the representative value of the dropout's gridcell forthe 9am hour in
January] over [the representative value of the neighbor's gridcell for the 9am hour in January].
Season-Hour
We first sort CMAQ modeled hourly values into four groups by season. In each season-group we
split evenly the ordered hourly values into 24 groups by time of day. This gives us 96 groups of
hourly O3 values (for every CMAQ grid cell). We calculate the average of the hourly values in
each gridcell-season-hour combination, and use this average as the representative value of that
CMAQ grid cell for that month-hour. In order to adjust a neighboring monitor value to reflect the
modeled relationship to the unmonitored or dropout site, the appropriate season-hour adjustment
factor must be used. For example, to adjust a monitor value from 9am in the Jan-Mar season, we
multiply by the ratio of [the representative value of the dropout's gridcell forthe 9am hour in the
Jan-Mar season] over [the representative value of the neighbor's gridcell for the 9am hour in the
Jan-Mar season].
Abt Associates Inc Appendix K-5
-------
Interpolating Hourly Data vs. Metrics
So far we have been speaking only of interpolating hourly O3 values from neighbor sites to a dropout
location. In principle, the exact same techniques can be used to interpolate daily (or even yearly) metrics
from neighbors to dropout.
For example, suppose we had a day's worth of hourly O3 values for monitor A and monitor B. We wanted
to predict the daily SUM06 value for location X, situated at the midpoint between monitors A and B. We
have two options. We can use eVNA to generate hourly O3 predictions for location X, then calculate the
daily SUM06 from these hourly predictions. Alternately, we can calculate daily SUM06 at each of the
neighbor sites, and then interpolate daily SUM06 values using eVNA.
Our work examines both of these methods. For each of the four eVNA approaches outlined above, we
generate a set of predictions based on interpolating hourly O3 values, and a set of predictions based on
interpolating daily metrics.
To interpolate daily metrics, we class hourly data according to some approach (month-decile, month-hour,
season-decile, season-hour). As with hourly-techniques, we adjust neighboring monitor values at the
hourly level (scale by a ratio of representative CMAQ values). However, before taking a distance-
weighted average over the set of neighbors, we compute daily metrics (SUM06 and 8-hour maximum
average) from the adjusted hourly neighbor data. These metrics are then distance-weight averaged to
produce daily metric predictions at the dropout site.
K.2.4 Choice of Dropout Monitor Sites
To test the validity of the different approaches, we dropped some monitors from our monitor sample, and
then used the remaining O3 monitors to predict O3 levels at these "out-of-sample" monitor sites. We
chose monitor sites for our out-of-sample testing that are isolated from other monitors and that appear to
be in relatively rural areas (Figures K-l and K-2). There are fewer monitors in the western United States
(i.e., west of 99 degrees longitude) than in the East, and the monitors tend to be a greater distance from
each other, with the exception of California, which has a large number of closely located monitors. As a
result, we chose fewer dropout sites in the West (12) than in the East (41).
Figure K-l Location of "Dropout" Monitor Sites (Triangle = West; Pentagon = East)
Abt Associates Inc Appendix K-6
-------
Figure K-2 Location of "Dropout" Monitor Sites and Other AQS and CASTNet Monitor Sites
K.2.5 Evaluation Criteria
In evaluating the different options for generating an O3 surface, we used two summary statistics
(normalized bias and normalized error) for both the annual maximum 3-month SUM06 and the annual 4th
highest daily maximum 8-hour average/ We define mean normalized bias and mean normalized error
formulaically for annual maximum 3-month SUM06. (The definitions of mean normalized bias and mean
normalized error for the annual 4th highest daily maximum 8-hour average follow the same principles, so
we have not presented them here.)
Mean SUM 06 normalized bias = averageiedmpouts (100 *
Mean SUM 06 normalized error = averageiedropouts(WO
actualSUM06t
\predictedSUM06!. - actualSUM06;
actualSUMQ6i
)
y See the discussion on the definition of model performance statistics starting on page 6 of the recent analysis by
EPA: CMAQ Performance Evaluation for 2001: Updated March 2005.
Abt Associates Inc
Appendix
K-7
-------
The mean normalized bias gives an average of the signed error of each dropout prediction. The mean
normalized error gives an average of the absolute error of each dropout prediction. A negative mean
normalized bias indicates predictions that tend to underestimate the actual values. A positive mean
normalized bias indicates predictions that tend to overestimate the actual values. All mean normalized
errors are positive. Their magnitude grows with the inaccuracy of the predictions.
To clarify these statistics, we give the following example dataset. Note that individual normalized biases
and errors differ only in sign. However, when averages are taken over normalized biases and normalized
errors, the results differ in magnitude as well as sign. A normalized error and a normalized bias which are
equal in absolute value indicate that predictions consistently underestimate (or overestimate, depending
on the sign) actual values:
eVNA value monitor value
5 7
9 8
3 5
6 5
bias error
-2
1
-2
1
normalized bias normalized error
2 -29% 29%
1 13% 13%
2 -40% 40%
1 20% 20%
mean normalized bias
mean normalized error
-9%
25%
It is erroneous to say that either mean normalized bias or mean normalized error is more "important" or
"meaningful" than another. Rather, they work in conjunction to tell a particular story. The error (from
here forward, we use bias and error to refer to mean normalized bias and mean normalized error)
describes the general level of accuracy of its corresponding set of predictions. The bias then indicates the
general distribution of this inaccuracy between under-predictions and over-predictions.
In choosing approaches for generating a POES, it is preferable to have low error (and thus more reliable
predictions), but the bias must play a role in the decision as well. For example, if one is attempting to
establish a lower bound on crop damage estimates, it is unwise to choose an approach with positive bias -
as this may overestimate O3 pollution.
It is worth noting, however, that bias and error cannot paint a complete picture of the results. Observe
below:
-30% -20% -10% 0% 10% 20% 30%
-30% -20% -10% 0% 10% 20% 30%
Both scenarios yield the same error and bias (16.7 percent error, 0 percent bias), even though the
distribution of predictions is quite different. This suggests that in addition to comparing the bias and
error, it may be useful to look at the distribution of the results themselves. To this end, we have included
individual prediction data for the hourly approaches in Section K.3.3 below.
Abt Associates Inc
Appendix
K-8
-------
K.3. Results
The approaches we considered are summarized in Table K-l, and ranked on a subjective scale of good /
fair / poor. The "good" ranking is assigned to the top-performing approaches. Often several approaches
receive a "good" ranking, because one performs slightly belter in terms of error, and the other performs
slightly better in terms of bias. Approaches that perform almost as well as "good" approaches (but
generally yield slightly less accurate results) receive a "fair" rating. The "poor" rating is given to
approaches that perform significantly worse than "good" approaches. Bear in mind that in this subjective
analysis, more weight was given to small variations in error than small variations in bias. However,
significant biases factored in strongly when compared to smaller variations in error. Sections K.3.1 and
K.3.2 provide a range of quantitative measures that underlie our evaluation. Section K.3.3 presents the
raw data on which these quantitative measures are based.
Table K-l Summary of O3 Prediction Accuracy by Region and Metric
Approach
VNA
VNA
Model
Month-Decile
Month-Hour
Season-Decile
Season-Hour
Month-Decile
Month-Hour
Season-Decile
Season-Hour
What gets
interpolated
Hour
Metric
Hour
Hour
Hour
Hour
Metric
Metric
Metric
Metric
SUM 06 - Eastern
Dropouts (12km
model Grid)
SUM06
Fan-
Poor
Poor
Good
Fair
Good
Fair
Poor
Poor
Poor
Poor
8-hour Maximum
— Eastern
Dropouts (12km
model grid)
Fair
Good
Poor
Fair
Good
Fair
Good
Good
Fair
Good
Fair
SUM 06 -
Western Dropouts
(36km model grid)
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
8-hour Maximum
—Western
Dropouts (36km
model grid)
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
Poor
We observe generally that in the east (on a 12km grid), the hour-month-decile and hour-season-decile
approach are best for predicting annual maximum 3-month SUM06 metrics, whereas the hour-month-
hour approach is best for predicting annual 4th highest daily maximum 8-hour average metrics. The data
further suggest (though they do not guarantee) that better predictions might be achieved through a blend
of VNA and eVNA techniques; applying VNA to monitors within 100km distance of the location in
question, and eVNA to monitors which are further away. We recommend further exploration of this
mixed approach.
The predictions in the west are much less accurate, which is not surprising given concerns expressed by
EPA staff about the accuracy of western modeling results. Section K.3.1 presents quantitative results for
the east; Section K.3.2 presents similar results for the west; Section K.3.3 presents the individual
predicted and actual values for each of the 41 eastern monitors.
K.3.1 Eastern Dropout Results
Abt Associates Inc
Appendix
K-9
-------
Table K-2 through K-4 summarize the results of our predictions for the selected dropout sites and for each
of the approaches. They used two summary statistics, (mean normalized) bias and (mean normalized)
error for both the annual maximum 3-month SUM06 and the annual 4th highest daily maximum 8-hour
average. For simplicity of presentation, these tables (and their discussion) use the terms SUM06 and 8-
hour maximum (or 8hr) in lieu of annual maximum 3-month SUM06 and annual 4th highest daily
maximum 8-hour average. Our results suggest that the best approach would be hour-season-hour.
Table K-2 summarizes the results from the entire set of 41 eastern dropouts for each of our 11
approaches. Based on our initial evaluation of the data, we observe the following:
• SUM06 predictions generally have greater bias and error compared to 8-hour.
• The model by itself predicts relatively poorly.
• Hourly VNA performs relatively well. However, some potentially significant gains in accuracy
are still possible by including an eVNA approach.
• For SUM06, interpolating hourly values is more accurate than interpolating metrics. (Model
predictions fall between these two approaches in terms of accuracy.)
• For the best overall performance on SUM06 (in the east), we would recommend the hour-month-
decile approach or the hour-season-decile approach. Hour-month-decile has a slightly lower bias;
hour-season-decile has a slightly lower error. The choice may depend on the specific application
of the data (see section 2.5 for a more in-depth discussion of error and bias)
• For the best overall performance on 8-hour (in the east), we would recommend the hour-month-
hour approach. Metric-month-decile may be appropriate in certain circumstances where its
tendency towards underestimation is not a concern.
Table K-2 Evaluation Statistics for Eastern Dropout Monitors
Adjustment Method
model-predictions
What gets interpolated SUM06 Bias SUM06 Error 8-hour Bias 8-hour Error
37.59
69.22
-6.72
9.26
VNA (no adjustment)
VNA (no adjustment)
hour
metric
4.18
116.69
32.61
118.89
-2.42
-1.37
6.23
5.86
month-decile
month-hour
season-decile
season-hour
hour
hour
hour
hour
-4.67
10.31
-5.09
14.33
27.60
36.31
27.02
39.05
-3.09
-0.64
-3.14
-0.70
6.09
6.17
5.92
6.30
month-decile
month-hour
season-decile
season-hour
metric
metric
metric
metric
108.47
115.06
109.30
115.11
111.35
117.02
111.98
117.08
-1.96
1.33
-1.86
1.18
5.80
7.02
5.73
6.93
Table K-3 separates the data into subsets based on the distance between neighbor monitor and dropout
site. To isolate distance as a variable, we must present the data at a pre-interpolation phase (interpolation
combines data from several neighbors, each of varying distance). Interpolation significantly reduces error
and bias; as a result, inaccuracies are significantly overstated in these data. We observe the following:
• Predictions generally become less accurate as distance increases. The one exception to this is the
transition from 0-5Okm to 50-100km distance. Here, SUM06 errors increase slightly, but SUM06
biases decrease. 8-hour errors and biases both decrease.
Abt Associates Inc
Appendix K-10
-------
VNA outperforms eVNA between 0-50 km. VNA also does quite well from 50-100km, with
very low relative error (though its bias is significantly higher than that of the month-decile and
season-decile approaches).
Above 100km, VNA's performance is significantly worse than that of eVNA. Season-decile
performs the best at these distances, with month-decile as a close second.
Table K-3 Evaluation Statistics for Adjusted Neighbor Values At Eastern Dropout Monitors, by Distance
from Neighbors
Adjustment Method
What gets interpolated
SUM06 Bias
SUM06 Error
8-hour Bias
8-hour Error
0-50 km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
8.18
10.32
13.71
11.04
16.05
29.65
31.48
33.60
33.26
36.62
1.39
1.40
4.35
1.78
4.78
7.32
8.11
8.84
7.55
9.28
50-1 00 km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
7.53
-2.71
6.67
-1.06
10.98
30.07
32.75
37.89
33.81
40.03
1.16
-0.29
3.37
-0.13
3.70
6.53
6.44
8.71
6.51
9.15
100- 150 km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
33.81
23.28
29.41
18.81
31.09
56.32
47.81
52.50
44.37
53.69
3.15
1.66
5.02
1.55
4.91
7.29
6.94
9.44
6.49
9.17
1 50+ km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
149.45
106.31
147.25
106.36
145.41
180.71
141.00
173.95
140.24
173.24
5.93
6.12
11.82
5.21
11.21
13.30
10.48
16.86
10.67
16.93
COMBINED
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
45.20
29.33
43.98
28.95
46.02
69.49
59.99
70.68
59.60
72.07
2.73
1.82
5.76
1.71
5.79
8.30
7.67
10.69
7.55
10.88
Note: To allow for distance-based comparison, the results from this table are calculated pre-interpolation, i.e. using the predicted
O3 values at each neighbor site, rather than the interpolated value at the dropout site. Because interpolation eliminates a great
deal of error, our actual results would be much more accurate than suggested in this exhibit. Compare the "Combined" results
from this exhibit with the hour-based results from exhibit 3-2 for more a quantitative look at this phenomenon.
Table K-4 separates the monitors between those with a low O3 concentration at the dropout site (as
determined by the dropout monitor's SUM06 value), and those with a high O3 concentration at the
dropout site. It then gives the four statistics for each of subset. The split between "low" and "high" is
made at the 50th percentile of SUM06 O3 values. We observe the following:
All approaches (Model, VNA, and eVNA) tend to overestimate low SUM06 values and underestimate
high SUM06 values, i.e. they reduce extremes. This trend is also generally observed for 8-hour values.
The model's SUM06 predictions are more accurate in high-O3 situations. In low O3 situations they
perform poorly.
VNA performs relatively poorly in low-O3 conditions. It performs very well in high O3 conditions - it is
one of the better approaches for high-O3 SUM06, and it performs relatively well on high-O3 8-hour
predictions.
Abt Associates Inc
Appendix K-ll
-------
Season-decile appears to be the best approach in low-O3 conditions. Season-hour appears to be the best
approach in high-O3 conditions, with season-decile as a close second.
Table K-4 Evaluation Statistics for Eastern Dropout Monitors, by Low/High O3 Levels
Adjustment Method What gets interpolated SUM06 Bias SUM06 Error 8-hour Bias 8-hour Error
LOW
Model
vna
month_decile
month_hour
season_decile
season hour
hour
hour
hour
hour
hour
98.29
26.66
11.92
39.48
11.58
45.66
111.17
46.33
33.33
53.79
31.50
59.53
-2.99
0.92
0.15
3.41
-0.01
3.30
8.19
6.49
5.48
6.93
5.26
6.85
HIGH
Model
vna
month_decile
month_hour
season_decile
season hour
hour
hour
hour
hour
hour
-20.21
-17.23
-20.48
-17.46
-20.96
-15.51
29.26
19.53
22.15
19.65
22.75
19.55
-10.28
-5.61
-6.18
-4.50
-6.13
-4.50
10.28
5.98
6.66
5.44
6.55
5.77
COMBINED
Model
vna
month_decile
month_hour
season_decile
season hour
hour
hour
hour
hour
hour
37.59
4.18
-4.67
10.31
-5.09
14.33
69.22
32.61
27.60
36.31
27.02
39.05
-6.72
-2.42
-3.09
-0.64
-3.14
-0.70
9.26
6.23
6.09
6.17
5.92
6.30
Abt Associates Inc
Appendix K-12
-------
K.3.2 Results of Western Monitors
Below we present data from the west that corresponds to the data from the east above. See section K.3.1
for explanatory text.
Table K-5 Evaluation Statistics for Western Dropout Monitors
Adjustment Method
model-predictions
What gets interpolated
SUM06 Bias
143.03
SUM06 Error
149.98
8-hour Bias
22.23
8-hour Error
22.81
VNA (no adjustments)
VNA (no adjustments)
hour
metric
-10.91
203.76
83.23
203.76
3.00
4.77
12.68
12.65
month-decile
month-hour
season-decile
season-hour
hour
hour
hour
hour
-18.97
-17.11
-19.50
-15.53
73.49
73.40
71.62
71.64
1.95
4.50
1.69
3.83
11.78
13.47
12.03
13.10
month-decile
month-hour
season-decile
season-hour
metric
metric
metric
metric
163.44
161.56
154.73
160.90
163.44
164.58
155.54
162.65
3.92
7.87
3.64
7.44
13.03
13.98
12.49
13.86
Table K-6 Evaluation Statistics for Adjusted Neighbor Values At Western Dropout Monitors, by Distance
from Neighbors
Adjustment Method What gets interpolated SUM06 Bias SUM06 Error 8-hour Bias 8-hour Error
0-50 km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
-42.41
-36.12
-32.18
-36.96
-33.52
57.13
62.94
67.35
62.58
66.02
-15.12
-13.60
-13.07
-13.82
-12.88
15.63
15.37
16.85
15.78
16.66
50-1 00 km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
195.88
174.62
166.77
169.45
170.27
214.91
190.13
179.62
183.58
183.12
17.40
17.73
26.63
17.44
25.66
21.02
21.08
29.83
21.09
29.04
100-1 50 km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
18.22
24.80
49.29
27.25
44.36
58.40
53.47
65.68
54.32
61.23
15.91
17.71
18.57
14.93
16.06
20.62
18.08
18.57
15.43
16.55
150+ km
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
18.21
-6.60
-1.43
-9.38
0.23
85.27
62.25
62.65
62.38
63.17
1.88
1.68
6.09
1.21
5.49
11.35
9.89
12.44
9.86
12.00
COMBINED
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
46.79
26.10
30.78
23.43
32.09
104.07
83.03
82.75
82.09
83.28
5.46
5.58
10.33
4.92
9.50
13.95
12.67
16.04
12.40
15.40
Table K-7 Evaluation Statistics for Western Dropout Monitors, by Low/High O3 Levels
Abt Associates Inc
Appendix K-13
-------
Adjustment Method What gets interpolated SUM06 Bias SUM 06 Error 8-hour Bias 8-hour Error
LOW
Model
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
277.28
33.36
10.71
11.37
9.46
14.64
277.28
123.39
103.78
100.75
102.53
99.72
29.29
10.40
8.43
10.98
8.38
10.49
29.29
15.23
15.25
15.94
15.03
15.59
HIGH
Model
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
31.15
-47.80
-43.70
-40.84
-43.63
-40.67
43.90
49.75
48.26
50.61
45.87
48.24
11.37
-7.34
-7.38
-4.84
-7.87
-5.68
12.54
8.57
7.88
10.41
8.25
9.93
COMBINED
Model
vna
month decile
month hour
season decile
season hour
hour
hour
hour
hour
hour
143.03
-10.91
-18.97
-17.11
-19.50
-15.53
149.98
83.23
73.49
73.40
71.62
71.64
20.33
1.53
0.52
3.07
0.26
2.41
20.92
11.90
11.56
13.17
11.64
12.76
K.3.3 Monitor-Level Prediction Data
This section contains individual predictions for the 41 eastern dropout monitors and the associated error
and bias values. We present this data for model predictions and all hour-based predictions. We have
omitted the equivalent tables for the less accurate metric-based predictions. For SUM06 predictions an
additional datum is given which specifies whether or not the 3-month period used to predict the annual
SUM06 metric corresponds to the 3-month period used to calculate the actual annual SUM06 metric. Its
value is 1 if the 3-month periods match, 0 otherwise.
Abt Associates Inc
Appendix K-14
-------
Table K-8 Model Predictions
Monitor ID SUM06 Acutal SUM06 Correct First Month Normalized Bias Normalized Error
8hr max Actual 8hr max Normalized Bias Normalized Error
0111900021
1200130111
1302100121
1308500012
1700100061
1704910012
1719710111
1805500011
1901700111
2205500051
2302100031
2700310011
2800100041
2918600051
3110900161
340273001 1
3604100051
3706500991
4002190021
400670671 1
4007190031
4213300081
4500300041
450890001 2
4707500031
4833900891
4846900031
5507300121
ABT147
CAD 150
CKT136
CND125
CTH110
ESP127
LRL117
LYK123
PED108
SND152
SUM156
UVL124
WST109
9.98
11.62
19.04
27.89
18.25
24.22
30.37
25.85
21.25
9.47
4.65
14.48
13.78
13.75
15.80
23.12
14.22
19.30
13.74
18.87
15.15
24.72
14.27
14.24
18.74
20.14
4.00
11.61
18.54
13.40
19.09
18.38
15.32
17.83
29.22
25.14
19.89
18.62
7.57
16.38
9.31
8.53
21.81
20.74
20.18
13.27
15.33
19.37
30.13
12.31
14.84
5.14
9.07
14.16
23.42
0.92
31.80
12.84
27.27
19.31
29.79
34.29
35.09
23.39
8.38
26.09
18.15
8.58
10.10
22.11
13.95
22.38
22.03
16.49
15.11
15.05
30.63
22.64
29.17
9.02
20.70
8.35
0
0
1
1
0
0
0
0
0
0
1
0
1
0
1
1
0
1
0
0
0
0
0
1
0
0
0
0
1
0
1
1
0
0
0
0
1
0
0
0
1
17.05
-46.74
-8.18
38.25
37.50
58.01
56.77
-14.20
72.59
-36.21
-9.57
59.71
-2.66
-41 .29
1619.52
-27.28
10.78
-29.22
-28.84
-36.67
-55.83
-29.56
-38.97
70.02
-28.17
10.95
-53.33
14.87
-16.14
-3.91
-14.70
-16.57
-7.09
17.98
94.10
-17.94
-12.15
-36.16
-16.07
-20.87
11.52
17.05
46.74
8.18
38.25
37.50
58.01
56.77
14.20
72.59
36.21
9.57
59.71
2.66
41.29
1619.52
27.28
10.78
29.22
28.84
36.67
55.83
29.56
38.97
70.02
28.17
10.95
53.33
14.87
16.14
3.91
14.70
16.57
7.09
17.98
94.10
17.94
12.15
36.16
16.07
20.87
11.52
69.38
68.19
81.41
73.87
72.97
76.00
77.93
77.87
71.48
72.25
64.35
79.91
69.10
75.00
70.52
81.06
71.75
71.49
71.31
71.87
69.47
78.79
70.37
69.45
72.88
77.96
67.13
67.69
86.13
68.81
69.93
73.17
70.45
69.06
77.09
76.50
71.53
72.69
67.03
80.59
66.68
72.57
79.00
86.25
77.75
78.00
77.13
78.43
85.75
69.88
77.25
68.00
73.67
75.13
75.13
51.50
101.13
82.38
85.64
75.00
81.25
81.88
87.75
79.88
67.88
81.50
90.63
73.50
72.13
106.00
74.28
76.69
80.08
85.23
74.36
75.58
86.05
84.64
83.20
68.51
85.24
69.39
-4.39
-13.68
-5.62
-5.00
-6.45
-1.46
-0.64
-9.19
2.30
-6.48
-5.37
8.47
-8.03
-0.16
36.94
-19.84
-12.89
-16.52
-4.92
-11.55
-15.15
-10.22
-11.90
2.31
-10.57
-13.98
-8.66
-6.15
-18.74
-7.35
-8.81
-8.62
-17.33
-7.13
2.01
-11.09
-15.48
-12.64
-2.16
-5.45
-3.90
4.39
13.68
5.62
5.00
6.45
1.46
0.64
9.19
2.30
6.48
5.37
8.47
8.03
0.16
36.94
19.84
12.89
16.52
4.92
11.55
15.15
10.22
11.90
2.31
10.57
13.98
8.66
6.15
18.74
7.35
8.81
8.62
17.33
7.13
2.01
11.09
15.48
12.64
2.16
5.45
3.90
Abt Associates Inc
Appendix K-15
-------
Table K-9 Hour-VNA Predictions
Monitor ID SUM06 Acutal SUM06 Correct First Month Normalized Bias Normalized Error
8hr max Actual 8hr max Normalized Bias Normalized Error
0111900021
1200130111
1302100121
1308500012
1700100061
1704910012
1719710111
1805500011
1901700111
2205500051
2302100031
2700310011
2800100041
2918600051
3110900161
340273001 1
3604100051
3706500991
4002190021
400670671 1
4007190031
4213300081
4500300041
4508900012
4707500031
4833900891
4846900031
5507300121
ABT147
CAD 150
CKT136
CND125
CTH110
ESP127
LRL117
LYK123
PED108
SND152
SUM156
UVL124
WST109
12.54
10.09
20.28
15.81
14.00
18.04
17.52
20.02
3.24
7.82
6.64
9.28
9.64
18.21
4.13
26.06
11.78
21.88
10.16
27.51
35.23
32.77
18.85
14.53
25.83
14.72
11.59
10.04
17.32
16.81
13.18
26.20
13.85
25.09
22.45
24.83
21.56
18.48
11.10
21.22
9.29
8.53
21.81
20.74
20.18
13.27
15.33
19.37
30.13
12.31
14.84
5.14
9.07
14.16
23.42
0.92
31.80
12.84
27.27
19.31
29.79
34.29
35.09
23.39
8.38
26.09
18.15
8.58
10.10
22.11
13.95
22.38
22.03
16.49
15.11
15.05
30.63
22.64
29.17
9.02
20.70
8.35
1
0
1
1
1
1
1
1
1
0
0
1
0
1
0
1
1
1
0
1
1
1
1
1
1
0
0
0
1
1
1
0
0
1
1
1
1
1
0
1
0
47.09
-53.75
-2.20
-21 .63
5.51
17.72
-9.55
-33.56
-73.67
-47.33
29.15
2.28
-31 .95
-22.28
349.67
-18.04
-8.24
-19.79
-47.37
-7.65
2.73
-6.61
-19.40
73.45
-0.98
-18.89
35.03
-0.61
-21 .66
20.52
-41.14
18.94
-16.05
66.02
49.13
-18.95
-4.78
-36.65
23.13
2.49
11.26
47.09
53.75
2.20
21.63
5.51
17.72
9.55
33.56
73.67
47.33
29.15
2.28
31.95
22.28
349.67
18.04
8.24
19.79
47.37
7.65
2.73
6.61
19.40
73.45
0.98
18.89
35.03
0.61
21.66
20.52
41.14
18.94
16.05
66.02
49.13
18.95
4.78
36.65
23.13
2.49
11.26
73.62
71.56
80.07
72.62
71.69
71.43
75.16
77.99
62.41
73.24
70.36
72.28
71.16
73.06
65.11
92.79
78.99
75.03
65.88
79.61
81.86
90.16
75.01
74.36
75.80
85.76
80.22
72.66
93.67
72.61
73.44
80.98
80.59
75.69
81.80
84.01
80.83
73.29
72.36
85.30
74.67
72.57
79.00
86.25
77.75
78.00
77.13
78.43
85.75
69.88
77.25
68.00
73.67
75.13
75.13
51.50
101.13
82.38
85.64
75.00
81.25
81.88
87.75
79.88
67.88
81.50
90.63
73.50
72.13
106.00
74.28
76.69
80.08
85.23
74.36
75.58
86.05
84.64
83.20
68.51
85.24
69.39
1.45
-9.42
-7.16
-6.59
-8.09
-7.38
-4.17
-9.05
-10.68
-5.19
3.47
-1.89
-5.27
-2.74
26.43
-8.24
-4.10
-12.39
-12.16
-2.02
-0.01
2.75
-6.09
9.55
-6.99
-5.37
9.15
0.75
-11.63
-2.25
-4.23
1.13
-5.43
1.79
8.24
-2.37
-4.50
-11.91
5.62
0.08
7.61
1.45
9.42
7.16
6.59
8.09
7.38
4.17
9.05
10.68
5.19
3.47
1.89
5.27
2.74
26.43
8.24
4.10
12.39
12.16
2.02
0.01
2.75
6.09
9.55
6.99
5.37
9.15
0.75
11.63
2.25
4.23
1.13
5.43
1.79
8.24
2.37
4.50
11.91
5.62
0.08
7.61
Abt Associates Inc
Appendix K-16
-------
Table K-10 Hour-Month-Decile Predictions
Monitor ID SUM06 Acutal SUM06 Correct First Month Normalized Bias Normalized Error
8hr max Actual 8hr max Normalized Bias Normalized Error
0111900021
1200130111
1302100121
1308500012
1700100061
1704910012
1719710111
1805500011
1901700111
2205500051
2302100031
2700310011
2800100041
2918600051
3110900161
340273001 1
3604100051
3706500991
4002190021
400670671 1
4007190031
4213300081
4500300041
450890001 2
4707500031
4833900891
4846900031
5507300121
ABT147
CAD 150
CKT136
CND125
CTH110
ESP127
LRL117
LYK123
PED108
SND152
SUM156
UVL124
WST109
10.26
9.59
20.32
16.01
15.37
16.28
17.95
19.79
3.39
7.02
5.24
12.66
9.72
13.66
1.97
32.39
12.80
22.46
2.77
23.34
33.88
31.26
18.14
12.65
26.15
24.46
14.65
10.70
18.44
12.23
15.21
25.43
12.92
19.18
22.98
26.21
19.67
15.92
6.94
19.77
9.39
8.53
21.81
20.74
20.18
13.27
15.33
19.37
30.13
12.31
14.84
5.14
9.07
14.16
23.42
0.92
31.80
12.84
27.27
19.31
29.79
34.29
35.09
23.39
8.38
26.09
18.15
8.58
10.10
22.11
13.95
22.38
22.03
16.49
15.11
15.05
30.63
22.64
29.17
9.02
20.70
8.35
1
0
1
1
1
1
1
1
0
0
1
1
1
1
0
1
1
1
0
1
1
1
1
1
1
0
1
0
1
1
1
1
0
1
1
0
1
1
1
1
0
20.36
-56.02
-2.00
-20.64
15.79
6.21
-7.33
-34.34
-72.50
-52.74
1.89
39.62
-31 .39
-41 .69
114.11
1.85
-0.36
-17.65
-85.66
-21 .65
-1.20
-10.91
-22.43
51.00
0.24
34.79
70.77
5.91
-16.59
-12.32
-32.06
15.45
-21 .70
26.94
52.68
-14.43
-13.10
-45.42
-23.05
-4.50
12.46
20.36
56.02
2.00
20.64
15.79
6.21
7.33
34.34
72.50
52.74
1.89
39.62
31.39
41.69
114.11
1.85
0.36
17.65
85.66
21.65
1.20
10.91
22.43
51.00
0.24
34.79
70.77
5.91
16.59
12.32
32.06
15.45
21.70
26.94
52.68
14.43
13.10
45.42
23.05
4.50
12.46
73.20
68.91
81.79
72.77
72.09
69.78
75.43
78.92
62.66
71.52
68.15
76.76
72.24
71.38
63.01
95.82
80.99
76.01
62.49
78.17
83.22
89.60
74.52
68.27
77.57
93.59
78.70
71.73
92.99
70.61
72.16
77.62
79.07
74.21
81.05
84.19
76.86
72.68
69.98
86.42
75.24
72.57
79.00
86.25
77.75
78.00
77.13
78.43
85.75
69.88
77.25
68.00
73.67
75.13
75.13
51.50
101.13
82.38
85.64
75.00
81.25
81.88
87.75
79.88
67.88
81.50
90.63
73.50
72.13
106.00
74.28
76.69
80.08
85.23
74.36
75.58
86.05
84.64
83.20
68.51
85.24
69.39
0.87
-12.77
-5.17
-6.41
-7.58
-9.53
-3.83
-7.97
-10.33
-7.42
0.21
4.20
-3.84
-4.99
22.34
-5.24
-1.68
-11.24
-16.67
-3.79
1.65
2.10
-6.70
0.58
-4.82
3.27
7.08
-0.55
-12.27
-4.94
-5.90
-3.06
-7.23
-0.20
7.24
-2.16
-9.19
-12.65
2.14
1.38
8.43
0.87
12.77
5.17
6.41
7.58
9.53
3.83
7.97
10.33
7.42
0.21
4.20
3.84
4.99
22.34
5.24
1.68
11.24
16.67
3.79
1.65
2.10
6.70
0.58
4.82
3.27
7.08
0.55
12.27
4.94
5.90
3.06
7.23
0.20
7.24
2.16
9.19
12.65
2.14
1.38
8.43
Abt Associates Inc
Appendix K-17
-------
Table K-ll Hour-Month-Hour Predictions
Monitor ID SUM06 Acutal SUM06 Correct First Month Normalized Bias Normalized Error
8hr max Actual 8hr max Normalized Bias Normalized Error
0111900021
1200130111
1302100121
1308500012
1700100061
1704910012
1719710111
1805500011
1901700111
2205500051
2302100031
2700310011
2800100041
2918600051
3110900161
340273001 1
3604100051
3706500991
4002190021
400670671 1
4007190031
4213300081
4500300041
450890001 2
4707500031
4833900891
4846900031
5507300121
ABT147
CAD 150
CKT136
CND125
CTH110
ESP127
LRL117
LYK123
PED108
SND152
SUM156
UVL124
WST109
12.55
9.42
20.16
17.41
16.92
16.68
20.07
21.36
3.87
6.57
5.64
11.06
13.51
15.00
4.10
33.94
13.15
22.93
3.88
22.45
33.97
30.01
17.82
13.82
25.95
27.36
22.41
11.84
19.39
22.42
16.87
24.82
14.38
20.56
24.67
29.25
20.56
17.76
8.89
20.41
9.55
8.53
21.81
20.74
20.18
13.27
15.33
19.37
30.13
12.31
14.84
5.14
9.07
14.16
23.42
0.92
31.80
12.84
27.27
19.31
29.79
34.29
35.09
23.39
8.38
26.09
18.15
8.58
10.10
22.11
13.95
22.38
22.03
16.49
15.11
15.05
30.63
22.64
29.17
9.02
20.70
8.35
0
1
1
1
1
1
1
1
0
0
0
1
1
1
0
1
0
1
0
1
1
1
1
1
1
0
1
0
1
1
1
0
0
1
1
0
1
1
1
1
0
47.20
-56.80
-2.77
-13.73
27.48
8.79
3.59
-29.09
-68.60
-55.76
9.63
21.98
-4.57
-35.98
346.07
6.73
2.37
-15.90
-79.92
-24.64
-0.93
-14.48
-23.80
64.97
-0.52
50.77
161.21
17.18
-12.28
60.68
-24.65
12.69
-12.81
36.08
63.88
-4.52
-9.17
-39.13
-1.45
-1.41
14.42
47.20
56.80
2.77
13.73
27.48
8.79
3.59
29.09
68.60
55.76
9.63
21.98
4.57
35.98
346.07
6.73
2.37
15.90
79.92
24.64
0.93
14.48
23.80
64.97
0.52
50.77
161.21
17.18
12.28
60.68
24.65
12.69
12.81
36.08
63.88
4.52
9.17
39.13
1.45
1.41
14.42
75.63
70.72
80.61
75.04
74.18
70.25
77.97
79.53
63.54
70.95
68.85
75.31
75.78
71.77
66.23
101.91
82.15
76.19
62.75
81.11
83.26
89.04
74.67
71.60
76.57
99.74
88.58
73.44
95.83
77.52
75.48
77.69
81.93
74.90
83.25
90.82
78.28
75.06
71.37
85.61
75.58
72.57
79.00
86.25
77.75
78.00
77.13
78.43
85.75
69.88
77.25
68.00
73.67
75.13
75.13
51.50
101.13
82.38
85.64
75.00
81.25
81.88
87.75
79.88
67.88
81.50
90.63
73.50
72.13
106.00
74.28
76.69
80.08
85.23
74.36
75.58
86.05
84.64
83.20
68.51
85.24
69.39
4.21
-10.48
-6.54
-3.48
-4.90
-8.91
-0.59
-7.26
-9.06
-8.15
1.25
2.23
0.88
-4.46
28.59
0.77
-0.27
-11.04
-16.33
-0.17
1.69
1.47
-6.51
5.48
-6.05
10.05
20.52
1.83
-9.59
4.37
-1.58
-2.98
-3.86
0.72
10.16
5.55
-7.52
-9.78
4.17
0.43
8.92
4.21
10.48
6.54
3.48
4.90
8.91
0.59
7.26
9.06
8.15
1.25
2.23
0.88
4.46
28.59
0.77
0.27
11.04
16.33
0.17
1.69
1.47
6.51
5.48
6.05
10.05
20.52
1.83
9.59
4.37
1.58
2.98
3.86
0.72
10.16
5.55
7.52
9.78
4.17
0.43
8.92
Abt Associates Inc
Appendix K-18
-------
Table K-12 Hour-Season-Decile Predictions
Monitor ID SUM06 Acutal SUM06 Correct First Month Normalized Bias Normalized Error
8hr max Actual 8hr max Normalized Bias Normalized Error
0111900021
1200130111
1302100121
1308500012
1700100061
1704910012
1719710111
1805500011
1901700111
2205500051
2302100031
2700310011
2800100041
2918600051
3110900161
340273001 1
3604100051
3706500991
4002190021
400670671 1
4007190031
4213300081
4500300041
450890001 2
4707500031
4833900891
4846900031
5507300121
ABT147
CAD 150
CKT136
CND125
CTH110
ESP127
LRL117
LYK123
PED108
SND152
SUM156
UVL124
WST109
10.52
8.86
19.19
16.09
14.68
16.57
17.49
20.08
3.40
7.29
5.13
12.25
10.09
14.27
2.09
32.46
12.71
22.44
2.70
23.12
33.87
31.13
18.35
12.32
25.98
22.04
12.84
10.31
18.13
14.81
14.66
25.71
12.96
19.81
23.69
26.67
19.50
16.04
6.80
19.71
9.30
8.53
21.81
20.74
20.18
13.27
15.33
19.37
30.13
12.31
14.84
5.14
9.07
14.16
23.42
0.92
31.80
12.84
27.27
19.31
29.79
34.29
35.09
23.39
8.38
26.09
18.15
8.58
10.10
22.11
13.95
22.38
22.03
16.49
15.11
15.05
30.63
22.64
29.17
9.02
20.70
8.35
1
0
1
1
1
1
1
1
0
0
0
1
1
1
0
1
1
1
0
1
1
1
1
1
1
0
1
0
1
1
1
0
1
1
1
0
1
1
1
1
0
23.33
-59.39
-7.47
-20.26
10.65
8.07
-9.71
-33.37
-72.41
-50.90
-0.20
35.09
-28.73
-39.09
127.35
2.09
-0.99
-17.72
-86.00
-22.39
-1.23
-11.29
-21 .53
47.03
-0.42
21.44
49.65
2.08
-18.00
6.16
-34.52
16.70
-21.41
31.12
57.40
-12.93
-13.86
-45.01
-24.56
-4.81
11.48
23.33
59.39
7.47
20.26
10.65
8.07
9.71
33.37
72.41
50.90
0.20
35.09
28.73
39.09
127.35
2.09
0.99
17.72
86.00
22.39
1.23
11.29
21.53
47.03
0.42
21.44
49.65
2.08
18.00
6.16
34.52
16.70
21.41
31.12
57.40
12.93
13.86
45.01
24.56
4.81
11.48
72.10
70.44
80.80
72.17
72.47
70.91
76.12
78.72
62.94
72.93
67.51
74.99
71.92
70.61
63.12
96.12
80.17
75.01
62.32
78.55
81.82
90.01
74.14
69.00
75.57
94.10
76.55
71.67
94.17
70.82
73.38
78.90
79.43
73.21
81.62
85.60
78.54
70.81
70.27
86.77
74.89
72.57
79.00
86.25
77.75
78.00
77.13
78.43
85.75
69.88
77.25
68.00
73.67
75.13
75.13
51.50
101.13
82.38
85.64
75.00
81.25
81.88
87.75
79.88
67.88
81.50
90.63
73.50
72.13
106.00
74.28
76.69
80.08
85.23
74.36
75.58
86.05
84.64
83.20
68.51
85.24
69.39
-0.66
-10.83
-6.32
-7.17
-7.09
-8.06
-2.94
-8.20
-9.93
-5.60
-0.71
1.79
-4.26
-6.01
22.57
-4.95
-2.68
-12.41
-16.91
-3.33
-0.07
2.57
-7.19
1.66
-7.28
3.84
4.15
-0.63
-11.17
-4.65
-4.32
-1.47
-6.80
-1.55
7.99
-0.52
-7.20
-14.89
2.57
1.80
7.93
0.66
10.83
6.32
7.17
7.09
8.06
2.94
8.20
9.93
5.60
0.71
1.79
4.26
6.01
22.57
4.95
2.68
12.41
16.91
3.33
0.07
2.57
7.19
1.66
7.28
3.84
4.15
0.63
11.17
4.65
4.32
1.47
6.80
1.55
7.99
0.52
7.20
14.89
2.57
1.80
7.93
Abt Associates Inc
Appendix K-19
-------
Table K-13 Hour-Season-Hour Predictions
Monitor ID SUM06 Acutal SUM06 Correct First Month Normalized Bias Normalized Error
8hr max Actual 8hr max Normalized Bias Normalized Error
0111900021
1200130111
1302100121
1308500012
1700100061
1704910012
1719710111
1805500011
1901700111
2205500051
2302100031
2700310011
2800100041
2918600051
3110900161
340273001 1
3604100051
3706500991
4002190021
400670671 1
4007190031
4213300081
4500300041
450890001 2
4707500031
4833900891
4846900031
5507300121
ABT147
CAD 150
CKT136
CND125
CTH110
ESP127
LRL117
LYK123
PED108
SND152
SUM156
UVL124
WST109
12.93
8.88
19.14
17.65
15.28
16.84
20.59
21.80
4.16
6.33
5.71
11.00
15.20
16.50
5.04
36.31
13.14
22.88
4.21
24.85
33.96
29.95
18.39
13.33
26.78
25.03
21.38
11.37
19.96
26.56
16.95
25.23
14.80
21.26
26.66
30.60
20.09
18.02
8.58
21.70
9.82
8.53
21.81
20.74
20.18
13.27
15.33
19.37
30.13
12.31
14.84
5.14
9.07
14.16
23.42
0.92
31.80
12.84
27.27
19.31
29.79
34.29
35.09
23.39
8.38
26.09
18.15
8.58
10.10
22.11
13.95
22.38
22.03
16.49
15.11
15.05
30.63
22.64
29.17
9.02
20.70
8.35
1
0
1
1
1
1
1
1
1
0
0
1
0
1
0
1
1
1
0
1
1
1
1
1
1
0
1
0
1
1
1
0
0
1
1
0
1
1
1
1
0
51.66
-59.27
-7.68
-12.52
15.12
9.88
6.28
-27.64
-66.20
-57.34
11.09
21.25
7.31
-29.57
448.65
14.18
2.31
-16.12
-78.19
-16.60
-0.97
-14.66
-21 .36
59.19
2.65
37.94
149.17
12.49
-9.72
90.42
-24.28
14.54
-10.28
40.66
77.07
-0.12
-11.26
-38.23
-4.84
4.80
17.67
51.66
59.27
7.68
12.52
15.12
9.88
6.28
27.64
66.20
57.34
11.09
21.25
7.31
29.57
448.65
14.18
2.31
16.12
78.19
16.60
0.97
14.66
21.36
59.19
2.65
37.94
149.17
12.49
9.72
90.42
24.28
14.54
10.28
40.66
77.07
0.12
11.26
38.23
4.84
4.80
17.67
74.49
70.12
79.31
73.46
73.28
70.97
78.00
79.32
63.84
70.71
68.25
75.87
76.89
71.23
66.31
104.07
81.88
75.86
63.78
80.43
81.87
89.35
74.25
71.23
75.19
98.01
88.07
72.74
96.28
78.99
76.34
79.90
82.43
74.03
83.83
90.93
79.39
73.49
71.52
87.73
75.59
72.57
79.00
86.25
77.75
78.00
77.13
78.43
85.75
69.88
77.25
68.00
73.67
75.13
75.13
51.50
101.13
82.38
85.64
75.00
81.25
81.88
87.75
79.88
67.88
81.50
90.63
73.50
72.13
106.00
74.28
76.69
80.08
85.23
74.36
75.58
86.05
84.64
83.20
68.51
85.24
69.39
2.64
-11.24
-8.05
-5.52
-6.05
-7.98
-0.55
-7.49
-8.63
-8.47
0.37
2.99
2.35
-5.18
28.76
2.91
-0.61
-11.41
-14.97
-1.01
-0.01
1.82
-7.05
4.94
-7.74
8.15
19.83
0.85
-9.17
6.35
-0.45
-0.22
-3.28
-0.45
10.92
5.67
-6.20
-11.67
4.39
2.92
8.95
2.64
11.24
8.05
5.52
6.05
7.98
0.55
7.49
8.63
8.47
0.37
2.99
2.35
5.18
28.76
2.91
0.61
11.41
14.97
1.01
0.01
1.82
7.05
4.94
7.74
8.15
19.83
0.85
9.17
6.35
0.45
0.22
3.28
0.45
10.92
5.67
6.20
11.67
4.39
2.92
8.95
Abt Associates Inc
Appendix K-20
-------
Appendix L Comparison of Hour and Metric
Interpolation Approaches
This appendix offers an initial discussion of the tendency observed (In appendix K) of metric-based
interpolation to over-predict SUM06, and hour-based interpolation to slightly under-predict SUM06 (here
we once again use the terms SUM06 and 8-hour maximum to refer to the associated annual metrics).
> When interpolating metrics, any neighbor's O3 value at or above 60 ppb will have a positive
effect on the final SUM06 statistic, even if the rest of the neighbors do not show any O3 levels
greater than or equal to 60 ppb during the hour in question. This leads to a tendency to over-
predict SUM06.
> When interpolating hours, neighboring O3 values above 60 only have an effect on the final daily
SUM06 statistic if the other neighbors have high enough O3 values in the exact hour of the first
neighbor's high value. This leads to lower values, as made clear in the examples that follow.
Note that the examples that we present below are only for SUM06. Similar arguments (not elaborated
here) would seem to apply to the 8-hour maximum, and indeed suggest that the metric approach would
give somewhat higher values than the hour approach. However, our reported results (e.g., Exhibit 3-1)
show the reverse, with the metric approach predicting lower values than the hour approach. This suggests
that the arguments presented here do not directly apply to the 8-hour maximum, or there are some errors
in our reported results, or, perhaps, that there are some other issue that we have not yet identified that
explain the results that we have presented. As a result, we need to view our conclusions as tentative, until
we can better understand our results.
Example of Over-prediction in Metric-based SUM06 Interpolation
Suppose location X has 2 equidistant neighbors, monitor A, and monitor B.
Suppose monitor data looks like this:
Time Monitor A Monitor B
Sam 0.01 0.01
9am 0.02 0.01
10am 0.03 0.01
11am 0.04 0.01
12pm 0.05 0.01
1pm 0.06 0.01
2pm 0.05 0.01
3pm 0.04 0.01
4pm 0.03 0.01
5pm 0.02 0.01
6pm 0.01 0.01
7pm 0.01 0.01
Under the hourly approach (using VNA), daily SUM06 is predicted to be 0:
Abt Associates Inc Appendix L-l
-------
Time Hourly VNA Interpolation to Location X
Sam 0.01
9am 0.015
10am 0.02
11am 0.025
12pm 0.03
1pm 0.035
2pm 0.03
3pm 0.025
4pm 0.02
5pm 0.015
6pm 0.01 SUM06=0
7pm 0.01
But under the Metric Approach, daily SUM06 is predicted to be .03:
SUM06 A 0.06
SUM06 B 0
Interpolation to SUM06 at Location X 0.03
An even more extreme example can be seen where a monitor reports an unlikely high value:
Time Monitor A Monitor B
Sam 0.01 0.01
9am 0.02 0.01
10am 0.03 0.01
11am 0.04 0.01
12pm 0.05 0.01
1pm 0.06 0.01
2pm 0.05 0.01
3pm 0.04 0.01
4pm 0.03 0.01
5pm 0.02 0.01
6pm 0.01 0.06
7pm 0.01 0.01
This "fluke" .06 doesn't have an effect on daily SUM06 in the hourly approach, since it doesn't correlate
with the data from monitor A very well:
Time Hourly VNA Interpolation to Location X
Sam 0.01
9am 0.015
10am 0.02
11am 0.025
12pm 0.03
1pm 0.035
2pm 0.03
3pm 0.025
4pm 0.02
5pm 0.015
6pm 0.035 SUM06=0
7pm 0.01
But in the metric approach, it has a large effect:
Abt Associates Inc Appendix L-2
-------
SUM06 A 0.06
SUM06 B 0.06
Interpolation to SUM06 at Location X 0.06
In this case, .06 seems to be an over-approximation for location X. This agrees with our findings that the
metric approach tends to dramatically over predict SUM06 values.
An Example of Under-prediction in Hour-based SUM06 Interpolation
The hourly approach tends towards a slight underestimation of SUM06 values. We can understand this as
follows:
Time Monitor A Monitors Dropout @X
Sam 0.01 0.01 0.01
9am 0.02 0.01 0.01
10am 0.03 0.01 0.02
11am 0.04 0.02 0.03
12pm 0.05 0.03 0.04
1pm 0.06 0.04 0.05
2pm 0.05 0.05 0.06
3pm 0.04 0.06 0.05
4pm 0.03 0.05 0.04
5pm 0.02 0.04 0.03
6pm 0.01 0.03 0.02
7pm 0.01 0.02 0.01
We could interpret this as a pocket of high O3 levels moving across the terrain with the regional wind
pattern. It reaches A first, then the dropout at X, then B.
In a case such as this, hourly interpolation would under-predict daily SUM06, due to the failure of the
above-threshold O3 values to fall in the same hour at the neighbor monitors:
Time Hourly VNA Interpolation to Location X Location X
Sam
9am
10am
11am
12pm
1pm
2pm
3pm
4pm
5pm
6pm
7pm
0.01
0.015
0.02
0.03
0.04
0.05
0.05
0.05
0.04
0.03
0.02
0.15
0.01
0.01
0.02
0.03
0.04
0.05
0.06
0.05
0.04
0.03
0.02
0.01
SUM06 0 0.06
Abt Associates Inc Appendix L-3
-------
-------
References
Abt Associates Inc (1995) Ozone NAAQS Benefits Analysis: California Crops, July 5, 1995.
Beck, D.E.. 1990. Yellow-Poplar (Liriodendron tulipifera L.). in Silvics of North America. Vol. 2.
Hardwoods. USDA For. Serv. Agriculture Handbook 654. USDA,Washington, DC.
Britton, N.L. 1878. When the leaves appear. Bulletin of the Torrey Botanical Club 6:235-236.
Cannon, W.N., Jr., B.R. Roberts, and J.H. Barger. 1993. Growth and physiological response of water-
stressed yellow-poplar seedlings exposed to chronic ozone fumigation and ethylenediurea. Forest
Ecology and Management 61:61-73.
Chapman, J.W., and S.T. Gower, 1991. Aboveground production and canopy dynamics in sugar maple
and red oak trees in southwestern Wisconsin. Can J. For. Res. 21:1533-1543.
Chappelka, A.H., B.I. Chevone, and J.R. Seiler. 1988. Growth and physiological responses of Yellow-
Poplar seedlings exposed to ozone and simulated acidic rain. Environ Pollut 49:1-18.
Clark, A., and J.G. Schroeder. 1977. Biomass of yellow-poplar in natural stands in western North
Carolina. USDA-FS Research Paper SE-165.
EPA (1996) Air Quality Criteria For Ozone And Related Photochemical Oxidants. United States
Environmental Protection Agency (EPA), pp. 1-1 to 1-33.
Erdmann, G.G., R.M. Peterson, Jr., and R.R. Oberg. 1985. Crown releasing of red maple poles to shorten
high-quality sawlog rotations. Can J For Res 15:694-700
ESRI (2005) Annual World Temperature Zones, available on-line at
http://www.geographynetwork.com/ESRI_Temp_Yr/
FAPRI 1997. Meyers, William H. & Hayes, Dermot J. & Smith, Darnell B. & Mohanty, Samarendu &
Fuller, Frank H. &, 1998. FAPRI 1997 U.S. Agricultural Outlook, Staff General Research Papers
1102, Iowa State University, Department of Economics.
Oilman, E.F. 1988. Predicting root spread from trunk diameter and branch spread. J. Aboric 14:85-89.
Gunderson, C.A., R.J. Norby, and S.D. Wullschleger. 1993. Foliar gas exchange responses of two
deciduous hardwoods during 3 years growth in elevated CO2: no loss of photosynthetic
enhancement. Plant Cell and Environment 16:797-807.
Hansen, Mark H., Thomas Frieswyk, Joseph F. Glover, and John F. Kelly. 1992. The Eastwide forest
inventory database: Users manual. Gen. Tech. Rep. NC-151. St. Paul, MN: U.S. Department of
Agriculture, Forest Service, North Central Forest Experiment Station. 48 p.
Hanson, P.J., L.J. Samuelson, S.D. Wullschleger, T.A. Tabberer, and G.S. Edwards. 1994. Seasonal
patterns of light-saturated photosynthesis and leaf conductance for mature and seedling Quercus
mbraL. foliage: differential sensitivity to ozone exposure. Tree Physiol. 14:1351-1366.
Abt Associates Inc References R-l
-------
Hocker, H.W., Jr., and D.J. Early. 1983. Biomass and leaf area equations for northern forest species.
Research Report Number 102, NH Ag. Exp. Sta., U. of NH, Durham, NH.
IPM Center (various), available on-line at http://nepmc.org/rese_profiles.cfm ,
http://www.ipmcenters.org/cropprofiles/ , http://www.ncpmc.org/profiles/ .
Jensen, K.F. and R.L. Pattern. 1990. Response of yellow poplar (Liriodendron tulipifera L.) seedlings to
simulated acid rain and ozone. 1. Growth modifications. Environ. Exp. Bot. 30:59—66.
Kienholz, R. 1941. Seasonal course of height growth in some hardwoods in Conneticut. Ecology 22:249-
258.
Kloeppel, B.D., M.D. Abrams, and M.E. Kubiske. 1993. Seasonal ecophysiology and leaf morphology of
four successional Pennsylvania barrens species in open versus understory environments. Can J
For Res 23:181-189.
Lamb, G. 1915. A calendar of the leafing, flowering and seedling of common trees of the eastern United
States. USDA Weather Bureau, Monthly Weather Review Supplement 2:3-19.
Lee, E. H.; Hogsett, W. E. (1996) Methodology for calculating inputs for ozone secondary standard
benefits analysis: Part II. Report prepared for Office of Air Quality Planning and Standards, Air
Quality Strategies and Standards Division, U.S. Environmental Protection Agency, Research
Triangle Park, N.C., March 1996.
Lefohn A. S., J.A. Laurence, and R.J. Kohut (1988) A comparison of indices that describe the relationship
between exposure to O3 and reduction in the yield of agricultural crops. Atmospheric
Environment 22:1229-1240.
Lesser, V.M., Rawlings, J.O., Spruill, S.E., Somerville, M.C. (1990) Ozone Effects On Agricultural
Crops: Statistical Methodologies And Estimated Dose-Response Relationships. Crop Science 30,
pp. 148-155.
Lieth H., and J.S. Radford. 1971. Phenology, resource management, and synagraphic computer mapping.
BioScience 21:62-70.
Little, Elbert L. Jr. (1978) Digital Representations of Tree Species Range Maps from "Atlas of United
States Trees", available on-line at http://esp.cr.usgs.gov/data/atlas/little/
Morrow, R.R. and W.W. McKee. 1963. Seasonal radial growth of hardwoods. Cornell Plantations 18:68-
70.
Mowbray, T.B., and H.J. Oosting. 1968. Vegetation gradients in relation to environment and phenology
in a southern Blue Ridge gorge. Ecological Monographs 38:309-344.
Neufeld, H.S., J.A. Jernstedt, and B.L. Haines. 1985. Direct foliar effects of simulated acid rain. I.
Damage, growth and gas exchange. New Phytol. 99:389-405.
Norby R.J., and E.G. O'Neill. 1991. Leaf area compensation and nutrient interactions in CO2-enriched
seedlings of yellow-poplar (Liriodendron tulipifera L.). New Phytol. 117:515-528.
Olszyk DM, Thompson CR, Poe MP. (1988) Crop loss assessment for California: modeling losses with
different ozone standard scenarios. Environmental Pollution 53 (1-4), pp. 303-11.
Abt Associates Inc References R-2
-------
Panshin, A.J., and C. de Zeeuw. 1980. Textbook of wood technology. Structure, identification, properties
and uses of the commercial woods of the United States and Canada. 4th Edition. McGraw-Hill.
New York, NY. pp 722.
Reich, P.B., M.B. Walters, and D.S. Ellsworth. 1991. Leafage and season influence the relationships
between leaf nitrogen, leaf mass per area and photosynthesis in maple and oak trees. Plant, Cell
and Environ 14:251-259.
Samuelson, L.J. and G.S. Edwards. 1993. Tree versus seedling sensitivity to ozone in Quercus rubra L.
NewPhytol. 125:373-379.
Samuelson, L.J. 1994. Ozone-exposure responses of black cherry and red maple seedlings. Environ and
ExpBot 34:355-362.
Taylor, C.R., K.H. Reichelderfer and S.R. Johnson. 1993. Agricultural Sector Models for the United
States: Descriptions and Selected Policy Applications. Iowa State University Press, Ames, IA,
1993.
Tjoelker, M.G., and R.J. Luxmoore. 1992. Soil nitrogen and chronic ozone stress influence physiology,
growth and nutrient status ofPinus taeda L. and Liriodendron tulipifera L. seedlings. New
Phytol. 119:69-81.
USDA (1997), Usual Planting and Harvesting Dates for U.S. Field Crops, National Agricultural Statistics
Service Agricultural Handbook Number 628, December 1997.
Walters, R.S., and H.W. Yawney. 1990. Red maple (Acer rubrum L.). in Silvics of North America,
Volume 2, Hardwoods. R.M. Burns and B.H. Honkala Tech. Coordinators. USDA For. Ser.,
USDA Agriculture Handbook 654. USDA,Washington, DC.
Weinstein, D.A., P. B. Woodbury, B. Gollands, P. King, L. Lepak, D. Pendleton. 2002. Assessment of
Effects of Ozone on Forest Resources in the Southern Appalachian Mountains. Final Report.
Southern Appalachian Mountain Initiative. Asheville, N.C. 1032 pgs. (47 p text, 70 figures, 31
tables, 11 appendices).
Weinstein, D. A., Gollands, B; and Retzlaff, W. A. 2001. The effects of ozone on a lower slope forest of
the Great Smoky Mountain National Park: Simulations Linking an Individual Tree Model to a
Stand Model. Forest Science 47 (1): 29-42.
Wullschleger, S.D., R.J. Norby, and D.L. Hendrix. 1992. Carbon exchange rates, chlorophyll content, and
carbohydrate status of two forest tree species exposed to carbon dioxide enrichment. Tree Phys.
10:21-31.
Young, H.E., J.H. Ribe, and K. Wainwright. 1980. Weight tables for tree and shrub species in Maine. Life
Sci. and Ag. Exp. Sta., U. Maine at Orono. Miscellaneous Report 230. 84 p.
Abt Associates Inc References R-3
-------
United States Office of Air Quality Planning and Standards Publication No. EPA 452/R-07-002
Environmental Protection Health and Environmental Impacts Division January 2007
Agency Research Triangle Park, NC
------- |