Evaluation of the 1990 Baseline Ozone Exposure Estimates Used in EPA’s Economic Benefits Assessment of Alternative Secondary Ozone Standards Paper 99-375 Bret A. Schichtel, Stefan R. Falke, and Luis A. Vasconcelos Center for Air Pollution Impact and Trend Analysis (CAPITA), Washington University ABSTRACT EPA’s 1990 three month maximum daily SUM06 ozone exposure field was generated by spatially interpolating SUM06 values calculated from 1990 AIRS hourly ozone data. The exposure estimates were evaluated by first verifying the SUM06 calculations, then assessing the performance of the SUM06 exposure field in non monitored agricultural and forested areas using SUM06 exposure values calculated from data not included in EPA’s analysis (non-EPA SUM06). Last, the representativeness of the 1990 SUM06 exposure estimates was assessed by comparing exceedances of the 25 ppm-hrs proposed secondary SUM06 ozone standard during 1990 to the 10 year period 1986 - 1995. EPA’s SUM06 calculations suffered from two methodological errors which caused them to be underestimated by 2% on average. EPA’s exposure field systematically underestimated the non-EPA SUM06 values by an average of 25% or 5.5 ppm-hrs. This caused a number of nonEPA stations to be misclassified by the EPA SUM06 exposure field as being below the proposed secondary standard of 25 ppm-hrs. The underestimation appears to be due to a combination of the non-EPA SUM06 values having smaller values than the closest AIRS SUM06 value, and the spatial interpolation model suppressing the SUM06 values in some agricultural and forested regions. The evaluation of the representativeness of the 1990 SUM06 ozone exposure estimates showed that in the Western US, 1990 had below average exposure values. In the East, the 1990 exposure values were below average in the Northeast, while in the Southeast they were above average. The Eastern US also displayed a high degree of temporal variability with regions of elevated SUM06 values, >25 ppm-hrs, shifting from year to year. This high degree of variability precludes finding a single representative year for the East. Under the constraint of using only one year of data for the analysis, 1990 appears to be an adequate year for the ozone exposure baseline. INTRODUCTION The U.S. Environmental Protection Agency (EPA) partly based its proposed secondary ozone standard on an economic analysis assessing the benefits of reduced ozone on agricultural yields. The economic evaluation essentially compared the level of economic benefits under one ozone standard of air quality with the benefits level under another standard of air quality. The benefits are estimated by comparing projected ozone damage to crops under a baseline condition to damages under conditions consistent with attainment of an alternative secondary standard. If the baseline ozone exposure estimates are too high the economic analysis will likely overestimate any benefits of alternative secondary standards. Conversely, if ozone exposure estimates are too low the economic analysis will likely underestimate any benefits of alternative secondary standards. 1 The economic analysis was conducted using 1990 baseline ozone exposure fields generated by EPA’s National Health and Environmental Effects Research Laboratory – Western Ecology Division (NHEERL-WED) in Corvallis, Oregon. This was accomplished by first calculating ozone exposure estimates from 1990 measured hourly ozone data stored in the Aerometric Information Retrieval System (AIRS) database. These exposure estimates were then spatially interpolated to a 10 km grid over the United States using NHEERL-WED’s Geographic Information System (GIS) model.1, 2, 3 The accuracy of the exposure estimates in non-monitored regions is primarily dependent on the measured ozone concentrations at the monitoring sites being consistent with the ozone concentrations at the non-monitored regions. Most AIRS monitoring sites are in populated areas, leaving large agricultural and forested areas unmonitored, precisely the regions in which the economic analysis was most concerned.4, 5, 2 Ozone is a dynamic pollutant with distinct spatial and temporal differences between urban and rural regions.6, 7, 8, 9 For example, urban centers have pronounced ozone diurnal cycles which peak in the afternoon with low concentrations in the evening and morning, while rural and elevated locations generally display a less pronounced diurnal cycle.6, 7 Lefohn and Pinkerton10 and Lefohn and Lucier4 compared ozone data collected in forested regions to data collected in the nearest populated centers, and concluded that ozone exposure is most likely larger in populated centers than forested areas. These differences between urban and rural ozone coupled with the lack of AIRS monitoring sites in many rural regions have raised concerns over the suitability of using AIRS data to estimate ozone exposure over agricultural and forested areas. In this paper, the suitability of the AIRS data and NHEERL-WED’s GIS model to estimate 1990 baseline ozone exposures in non-monitored agricultural and forested areas is evaluated. This evaluation is conducted using the 1990 three-month maximum daily SUM06 ozone exposure estimates generated by NHEERL-WED. The three month maximum daily SUM06 exposure estimate is the maximum sum of daily SUM06 values over three consecutive months within each state’s EPA defined ozone season. The daily SUM06 value is the summation of all hourly ozone concentrations above 0.06 ppm between 8am and 8pm.5, 1 This exposure index was chosen for the evaluation because it was EPA’s recommended form of the proposed secondary standard with a recommended level between 25 – 38 ppm-hrs.2 In the remainder of this paper, SUM06 refers to three month maximum daily SUM06, unless otherwise specified. Note that the three month maximum daily SUM06 exposure estimates are not directly comparable to the baseline SUM06 exposure estimates used in EPA’s economic analysis. The baseline exposure estimates were generated by summing the daily SUM06 values over each crop’s growing season in each state, while the three month maximum SUM06 values summed the daily SUM06 values over a consecutive three month time period within each state’s EPA ozone season. The three month maximum time periods fell within each crop’s growing seasons, so that the three month maximum daily SUM06 exposure estimates are lower bounds of the baseline SUM06 exposure estimates.11 The evaluation was conducted in two phases. First, EPA’s SUM06 calculations were validated. Second, the 1990 SUM06 exposure grid was compared to SUM06 values calculated from measured hourly ozone data collected in agricultural and forested areas. The evaluation used multiple data sets, data aggregation and interpolation techniques, and comparisons. To aid in following the various data trails, Figure 1 displays a data flow diagram. The components of this diagram are defined in the glossary and are discussed below. 2 As a final point of investigation the representativeness of the 1990 SUM06 exposure estimates was assessed by comparing the 1990 values to the 10 year period, 1986 - 1995. This was to determine how “typical” 1990 was in terms of ozone exposure, since ozone concentrations can vary substantially from year to year. MEASURED OZONE DATA Evaluation of the EPA SUM06 ozone exposure estimates was conducted using the same measured hourly ozone concentrations EPA used to create the exposure fields for the economic analysis, and additional measured data from agricultural and forested monitoring sites. These two data sets are referred to as EPA and non-EPA ozone data respectively. EPA Ozone Data The EPA ozone data set consisted of 728 monitoring sites from the AIRS database (Figure 2). Large metropolitan areas, such as the Eastern Seaboard, Los Angeles, CA, and St. Louis, MO, have a number of monitoring sites in and around them, while expansive rural regions in the West, Midwest and Southeast are unmonitored. In Figure 2, the monitors have been grouped into urban, suburban, rural, and agricultural and forest categories based upon AIRS site setting and land use classifications. These classifications are assigned by each state that submits data for inclusion in the AIRS database, and consistent criteria are not adhered to for every site. Rural monitors occur within cities and Census Metropolitan Statistical Areas,2 and often display diurnal cycles characteristic of urban sites.12 Nonetheless, for this analysis the AIRS site setting and land use classifications were used to group the monitoring sites due to a lack of a better classification scheme. This classification designates 162 monitoring sites as agricultural and forested sites and 138 sites as urban (Table 1). The average SUM06 ozone exposure estimates for each site classification are presented in Table 1. These SUM06 values were calculated from the raw AIRS data using the same procedure reported to be used in the economic analysis.1, 2 The lowest SUM06 values occur for the urban sites with an average of 18 ppm-hrs, while the monitoring sites in the suburban, rural, and agricultural and forested regions have average exposure estimates of ~22 ppm-hrs (Table 1). The standard deviations for all regions are similar at ~15 ppm-hrs. Non-EPA Ozone Data The non-EPA ozone data set consists of 61 monitoring sites located in agricultural and forested regions (Figure 3). This data set fills in some of the “holes” in the EPA data, such as the nonmonitored regions in West Virginia, northwestern Pennsylvania, and northern Mississippi. These data came from a database prepared for the deliberations of the Ozone Transport Assessment Group’s (OTAG) Air Quality Analysis Workgroup,8 and include data from the CASTNet monitoring network, the AIRS database, as well as several other smaller networks. The CASTNet network accounted for 46 of the 61 monitoring sites. This network was established to monitor the air quality in rural areas, and all monitoring sites followed strict siting criteria restricting their proximity to ozone precursor sources, such as population centers, industry, and roadways. The average non-EPA SUM06 value is 21.9 ppm-hrs with a standard deviation of 12.2 ppm-hrs (Table 2). These statistics are similar to those for the non urban EPA data in Table 1. This suggests 3 that on average the EPA monitoring data are consistent with the non-EPA SUM06 exposure estimates. EPA 1990 Three Month Maximum Daily SUM06 Ozone Exposure Estimates The EPA generated the 1990 SUM06 ozone exposure grid by spatially interpolating three month maximum daily SUM06 ozone exposure data values calculated from the EPA ozone data to a 10 km grid covering the conterminous US (Figure 4). These SUM06 ozone exposure grid and data values are referred to as the EPA SUM06 grid and the EPA SUM06 data, respectively. The spatial interpolation model used a Geographical Information Systems (GIS) approach developed at EPA’s NHEERL-WED.3, 1 The model is based on an inverse distance squared weighted interpolation technique for ozone that is enhanced by incorporating factors that influence ozone formation and transport, such as NOX emissions, temperature, and wind speed and direction, to provide additional information for non-monitored regions. These factors are used to produce a potential exposure surface (PES) that reflects the likelihood of a region having elevated ozone concentrations. For example, areas that experience large numbers of days with elevated temperature, low cloud cover, and are downwind of large sources of ozone precursor species will have a greater PES value than other areas not situated near any large sources of ozone precursors. The PES then acts as an added weight in the spatial averaging of the SUM06 ozone data for each grid cell. The GIS model “honors” the monitoring data, reproducing the measured values at all monitoring sites that fall at the center of a given grid cell. However, monitoring sites are rarely located at the center of a grid cell, so the gridded values only approximate the monitoring data. The GIS uncertainties are difficult to quantify, but they are mainly due to the spatial resolution and quality of the many input data sets.1 Another source of uncertainty is that the PES surface was calculated for the three months June -August, while the three month maximum SUM06 ozone values for the monitoring data could span any contiguous three month period during the EPA ozone season. The three month maxima for ~50% of the EPA monitoring sites were different from the June – August period. Uncertainties also arise from the assumptions used to generate the PES surface. The results of the GIS model correspond well with other interpolation methods, but no formal performance evaluation has been conducted.2 The 1990 EPA SUM06 grid is presented in Figure 4. As shown, the largest SUM06 values occur in Southern California, exceeding 50 ppm-hrs. Elevated exposure estimates also exist in the Southeast from Tennessee to South Carolina with levels between 25 - 35 ppm-hrs. Low values, <10ppm-hrs, exist throughout the West, excluding southern California, and the Northeast. Verification of EPA Three Month Maximum Daily SUM06 Ozone Calculations The calculation of the three month maximum daily SUM06 exposure values from measured data follows a complex procedure to account for missing hourly ozone data; such methodologies are prone to error. Therefore, the EPA SUM06 data were verified to ensure that EPA used the proper calculation procedure. Also, EPA’s exact calculation procedure needed to be identified for use in evaluating the EPA SUM06 grid. The verification was conducted by comparing the EPA SUM06 data values to SUM06 values calculated from the EPA ozone data set (see data flow 1 in Figure 1). Our calculated SUM06 values were generated using the procedure reported to be used for the EPA economic analysis.1, 2 These 4 data will be referred to as CAPITA SUM06 data. Figure 5A presents a scatter plot between the CAPITA and EPA SUM06 data. As shown, the two calculation procedures do not match, with the EPA SUM06 data ~2% lower on average than CAPITA SUM06 data, and two out of the 728 sites with differences greater than 6 ppm-hrs. It was found that the EPA analysis suffered from two errors. First, EPA did not properly account for missing hourly ozone concentrations in months with fewer than 31 days. This error resulted in site specific EPA SUM06 data values being up to 4% less than the corresponding CAPITA SUM06 data. The second error was that EPA did not estimate SUM06 values for missing months of data, but used the next available month with valid data in the three month maximum calculation. This second error was the more serious, resulting in SUM06 values calculated from non consecutive months, and was responsible for the two sites with deviations greater than 6 ppm-hrs. Appendix B contains a full description of EPA SUM06 calculation errors. Recalculating the CAPITA SUM06 data using the flawed EPA procedure produced virtually identical results to the EPA SUM06 values (Figure 5B). Small differences in the two SUM06 data sets are expected due to round off errors, but two SUM06 values differ by more than 10%. The causes for these two large differences are not known. Evaluation of EPA’s 1990 Three Month Maximum Daily SUM06 Ozone Grid Comparison of EPA 1990 Three Month Maximum Daily SUM06 Ozone Exposure Grid Estimates with EPA SUM06 Data The EPA SUM06 ozone exposure grid was generated by spatially interpolating the EPA SUM06 data. Comparing the input EPA SUM06 data to the output EPA SUM06 grid provides a consistency check on NHEERL-WED’s GIS model. The SUM06 grid estimates the ozone exposure at the center of the grid cells, and assumes that this value is consistent with the exposure throughout the ten km grid cell. In actuality ozone concentrations can vary substantially within the grid cells, and comparing SUM06 point values from stations located anywhere within a grid cell to the grid cell value will produce differences. These differences are primarily due to intra-cell variability of the SUM06 grid values, and can be used to assess the error associated with using the grid cell values to approximate the SUM06 value over the 10 km region. The comparison was conducted by scattering the EPA SUM06 data against the corresponding EPA SUM06 grid values (Figure 6), see data flow 2 in Figure 1. As shown in Figure 6, the EPA SUM06 grid values correlate well with the EPA SUM06 data, r = 0.97. However, significant differences exist with values from ten out of 684 EPA monitoring sites differing by more than 10 ppm-hrs from the EPA SUM06 grid, and one site almost 40 ppm-hrs greater than the EPA SUM06 grid. The intra-cell variability in the EPA SUM06 grid can be approximated by the root mean square (RMS) error. Table 3 contains the RMS error for each site category and all sites grouped together. As shown, the RMS error for all sites is 3.2 ppm-hrs with the lowest RMS error occurring in agricultural and forested regions at 2.7 ppm-hrs and the highest RMS error occurring in the rural regions. The high rural RMS error is due to the EPA SUM06 grid underestimating a rural site located at the southern tip of Lake Michigan by 38 ppm-hrs. With this site removed, the rural RMS error decreases to 1.9 ppm-hrs. 5 Performance of the 1990 EPA Three Month Maximum SUM06 Ozone Exposure Grid in Agricultural and Forested Regions The performance of the 1990 EPA SUM06 grid in agricultural and forested regions was evaluated by comparing the grid values to SUM06 exposure values calculated from the non-EPA data (data flow 3 in Figure 1). These SUM06 data values are referred to as non-EPA SUM06 data. The nonEPA SUM06 data were calculated using the flawed EPA procedure. The hourly ozone concentrations for all non-EPA monitoring sites were sufficiently complete that the issue of missing months in the EPA calculation procedure was not a problem. The evaluation of the EPA SUM06 grid is presented as a scatter plot in Figure 7. As shown, the reproduction of the measured values deteriorates substantially compared to the EPA data (Figure 6). The correlation has decreased to r = 0.73 (Figure 7), and the baseline values underestimate the measured data by an average of 5.5 ppm-hrs or 25% of the measured average (Table 4). Also, the RMS error increased to 10 ppm-hrs, which is more than three time the RMS error between the EPA SUM06 grid and data in agricultural and forested regions. The underestimated values are distributed throughout the United States with half of the non-EPA monitoring sites being underestimated by 5 ppm-hrs or more (Figure 8). The bias is large enough that it can potentially influence the designation of agricultural and forested regions as in or out of attainment of a secondary standard. For example, 27 of the 61 non-EPA monitoring sites have SUM06 values greater than the proposed 25 ppm-hrs standard, 16 of these sites are estimated to be below the proposed 25 ppm-hrs standard by the EPA SUM06 grid. The systematic underestimation of the non-EPA SUM06 data could be due to the closest EPA monitoring sites to the non-EPA monitoring sites generally having lower SUM06 values, and/or errors in the PES surface used in the GIS model. The potential cause was examined by computing a new 1990 SUM06 exposure grid by spatially interpolating the EPA SUM06 data using an inverse distance squared weighted interpolator (1/r2), see data flow 4 in Figure 1. The 1/r2 interpolator is identical to the NHEERL-WED GIS model assuming a constant PES surface. With the influence of the PES surface removed, the 1/r2 SUM06 grid is dependent only on the EPA SUM06 data and provides a measure of the consistency between the closest EPA data and the non-EPA data. The comparison of the non-EPA SUM06 data to corresponding 1/r2 SUM06 grid values in Figure 9 shows that the 1/r2 SUM06 grid underestimates the non-EPA SUM06 data by an average of 3 ppmhrs or 13% of the average non-EPA SUM06 values. This is about half of the underestimation by the EPA SUM06 exposure grid. Therefore, it appears that on average about half of the underestimation of the non-EPA SUM06 data values by the EPA SUM06 grid is due to the EPA SUM06 data values being lower than the non-EPA SUM06 data, and the other half is attributable to NHEERL-WED’s GIS interpolation scheme suppressing the SUM06 values, when in fact it should amplify them. Representativeness of 1990 As A Typical Ozone Exposure Year The economic analysis for the secondary ozone standard was performed using baseline ozone exposure estimates for only 1990. According to the EPA,2 the 1990 air quality baseline was chosen in order to be consistent with other data in the economic analysis. In addition to consistency, an important issue is whether 1990 was a typical ozone year, i.e. did 1990 have high, low, or average ozone concentrations relative to historical ozone records? If 1990 was an extreme year with very high or low ozone concentrations then the economic analysis would produce higher or lower benefits, respectively, compared to a typical ozone year. 6 In this section, the representativeness of the 1990 SUM06 ozone exposure estimates are assessed by first establishing the spatial and temporal variability of the SUM06 ozone exposure estimates over a ten year period, 1986 - 1995, and then comparing the 1990 SUM06 values to this historical record. The analysis focuses on exceedances of the proposed secondary standard that had recommended three month maximum daily SUM06 ozone exposure estimates between 25 - 38 ppm-hrs, since benefits in the economic analysis were only realized if the ozone exposure estimates exceeded the level of the proposed secondary ozone standard. The historical ozone data consisted of two data sets. One data set contained five years, 1988 – 1992, of SUM06 ozone exposure grids. These SUM06 grids were generated by NHEERL-WED using the GIS model to spatially interpolate SUM06 data calculated from AIRS hourly ozone concentrations.13 These SUM06 data values suffered from the same calculation errors as those using in the EPA SUM06 Data. The other data set consisted of 10 years, 1986-1995, of measured hourly ozone concentrations from the AIRS database and the CASTNet and several smaller monitoring networks. These data came from the ozone database prepared for the deliberations of the OTAG Air Quality Analysis Workgroup.8 The SUM06 values were calculated using the procedure reported by the EPA.5, 1 The five years of SUM06 grids are presented in Figure 10, and the ten years of SUM06 data are presented in Figure 11 as the average SUM06 value and the percentage of monitoring sites exceeding the proposed 25 ppm-hrs standard for six equal area regions of the US and the conterminous US. As shown in Figure 10, exceedances of the proposed secondary standard typically occur throughout California south of San Francisco, and in the Eastern US between the Great Lakes and the Gulf of Mexico. In the other regions, exceedances of the proposed standard occur near urban centers, such as Salt Lake City, UT, Denver, CO, and Dallas, TX. The regional SUM06 ozone values for the ten year data set in the West display moderate temporal variability, varying about 50% in the Northwest (8-16 ppm-hrs) and 30% in the Southwest (26-34 ppm-hrs) (Figures 11E and F). The East displays larger temporal variability with the Northeast and Southeast average SUM06 exposure estimates varying by almost a factor of three from ~40 ppm-hrs in 1988 to ~15 ppm-hr in 1992 (Figures 11A and B). During 1988, more than 85% of all Eastern monitoring stations exceeded the proposed 25 ppm-hrs standard, while fewer than 10% of the stations exceeded this proposed standard during 1992 (Figures 10, 11A and B). The Eastern US also displays substantial year-to-year spatial variability. During 1990, most Eastern exceedances occurred in the South from western Tennessee to South Carolina, while in 1991 they occurred in the North from Indiana to New Jersey (Figure 10). Comparing the 1990 SUM06 ozone exposure estimates to the other four years in Figure 10 shows that the West displays the typical pattern of high SUM06 values in Southern California and low values elsewhere. However, the average 1990 SUM06 exposure values in the Southwest were the lowest over the 10 year period at 26 ppm-hrs compared to 30 ppm-hrs average over the 10 years (Figure 11 F). Also, the number of stations exceeding the proposed 25 ppm-hrs standard was 38% compared to 45% on average. Overall, the 1990 Western SUM06 values can be considered consistent with the typical spatial pattern of SUM06 values in other years, but SUM06 exposure estimates were low. In the Northeast, the 1990 average exposure values were also low, 19.3 compared to 23.6 ppm-hrs averaged over the ten year period (Figure 11A), with only 27% of the Northeast monitoring stations exceeding the proposed 25 ppm-hrs standard compared to 42% on average. However, the Southeast 7 had above average exposure estimates at 25 ppm-hrs compared to 22.5 ppm-hrs on average, and 50% of the monitors exceeded the proposed standard compared to 37% on average. Neither the 1990 spatial pattern of SUM06 ozone estimates nor their magnitudes are consistent with typical SUM06 values in the Eastern US. In fact, the large temporal variability of the SUM06 exposure values precludes the ideal of finding a single representative year for the Eastern US. SUMMARY AND CONCLUSIONS The evaluation of EPA’s 1990 three month maximum daily SUM06 ozone exposure estimates was conducted by first verifying the SUM06 calculations, then assessing the performance of the EPA SUM06 grid in non monitored agricultural and forested areas, and finally assessing the representativeness of the 1990 SUM06 exposure estimates as compared to other years. The verification of SUM06 calculations revealed that the EPA SUM06 calculations suffered from two methodological errors. First, missing hourly ozone concentrations in months with fewer than 31 days were not properly accounted for, and second, SUM06 values were not estimated for missing months of data, but instead used the next available month. This second error was more severe resulting in SUM06 estimates calculated from non consecutive months. These errors caused the EPA SUM06 data values to underestimate the proper values by 2% on average. The EPA SUM06 grid performance evaluation was conducted by comparing the grid values to nonEPA SUM06 values calculated from measured ozone concentrations in agricultural and forested areas that were not used in EPA’s analysis. It was found that the EPA SUM06 grid systematically underestimated the non-EPA SUM06 data by an average of 25% or 5.5 ppm-hrs. This caused a number of non-EPA stations to be misclassified by the EPA SUM06 grid as being below the proposed secondary standard of 25 ppm-hrs. Also, the root mean square error was 10 ppm-hrs, more than three times the error expected from the intra-cell variability. It is believed that the underestimation is due to a combination of the non-EPA SUM06 data having smaller values than the closest EPA SUM06 data, and the NHEERL-WED GIS model suppressing the SUM06 values in some agricultural and forested regions. The representativeness of the SUM06 ozone exposure estimates during 1990 was evaluated by comparing exceedances of the 25 ppm-hrs proposed secondary SUM06 ozone standard during 1990 to the 10 year period 1986 - 1995. In the West, exceedances of the proposed standard occurred throughout Southern California every year and occasionally occurred near urban centers, such as Denver, CO and Salt Lake City, UT. The 1990 exposure estimates were similar to this pattern, but had below average exposure values. In the Eastern US, the 1990 exposure values were below average in the Northeast, while in the Southeast they were above average. The Eastern US also displayed a high degree of temporal variability with regions of elevated SUM06 values, >25 ppmhrs, shifting from year to year. For example, 1990 had elevated SUM06 values occurring primarily in the Southeast, 1991 had elevated exposure estimates primarily in the Northeast, and 1992 had almost no elevated exposure estimates. This high degree of variability precludes finding a single representative year for the East. Under the constraint of using only one year of data for the analysis, 1990 appears to be an adequate year for the ozone exposure baseline. While overall the 1990 SUM06 ozone exposure estimates were below average, large regions of Southern California and the Eastern US still had elevated values as would be expected in a “typical” year. However, it should be noted that the analysis 8 included all available ozone data. It is not known how the analysis would change if only ozone exposure estimates in agricultural and forested regions were considered. ACKNOWLEDGMENTS The authors would like to thank the Electric Power Research Institute (EPRI) for its support of this work. The authors gratefully acknowledge William Hogsett, A. Herstrom, and H. Lee of EPA’s NHEERL-WED for their assistance in obtaining the EPA baseline ozone exposure estimates and information on the data and methodologies used to calculate the ozone exposure estimates. We also acknowledge Ben Hartsell and Eric Edgerton for their assistance in obtaining data, documentation and information on the SCION and CASTNet ozone data. REFERENCES 1. Rodecap, K.; Lee H.; Herstrom A.; Broich S.; (June 29, 1995) Methodology for Calculation Inputs for Ozone Secondary Standard Benefits Analysis, Memorandum to Bill Hogsett, U.S. EPA. In Methodology for Calculation Inputs for Ozone Secondary Standard Benefits Analysis: Part II. Prepared for OAQPS by E. Henry Lee and William E. Hogsett, March 18, 1996. 2. U.S. Environmental Protection Agency. (1996b) Review of national ambient air quality standards for ozone assessment of scientific and technical information. Research Triangle Park, NC: Office of Air Quality Planning and Standards; EPA report nos. EPA-452/R-96-007. 3. Hogsett, W.E.; Weber J.E.; Tingey D.; Herstrom A.; Lee E.H.; and Laurence J.A. (1997). Environmental Auditing: An Approach for Characterizing Tropospheric Ozone Risk to Forests. Environmental Management 21:105-120. 4. Lefohn, A. S.; Lucier A. A. (1991) Spatial and temporal variability of ozone exposure in forested areas of the United States and Canada: 1978-1988. J Air Waste Manage. Assoc. 41: 694-701. 5. U.S. Environmental Protection Agency. (1996a) Air quality criteria for ozone and related photochemical oxidants. Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report nos. EPA/600/p93/004aF-cF. 6. Kelly, N A.; Wolff, G. T.; Ferman, M. A. (1984) Sources and sinks of ozone in rural areas. Atmos. Environ, 18:1251 7. Wolff G. T.; Lioy P. J.; Taylor R. S. (1987) The diurnal variations of ozone at different altitudes on a rural mountain in the eastern United States. JAPCA 37: 45 8. Husar, R. B. (1996) Spatial pattern of daily maximum ozone over the OTAG region. Presented to the OTAG Air Quality Analysis Workgroup. Available on the world wide web at “http://capita.wustl.edu/OTAG/Reports/otagspat/otagspat.html” 9. Husar, R. B. (1997) Seasonal Pattern of Ozone over the OTAG Region. Presented to the OTAG Air Quality Analysis Workgroup. Available on the world wide web at “http://capita.wustl.edu/otag/reports/otagseas/otagseas.html” 10. Lefohn A. S.; Pinkerton, J. E. (1988) High resolution characterization of ozone data for sites located in forested areas of the United States. JAPCA 38: 1504 11. Rodriguez, Rosalina (1997) Personal communication. 12. Stasiuk, W. N. and Coffey, P. E. (1974) Rural and urban ozone relationships in New York State. J. Air Pollut. Control Assoc. 24: 564-568. 9 13. Herstrom A. (1997) Personal communication. GLOSSARY CAPITA SUM06 Data: 1990 three month maximum daily SUM06 ozone exposure estimates calculated using the EPA ozone data and the reported SUM06 calculation procedure (See Appendix A) used in the secondary ozone economic analysis. EPA SUM06 Calculation Procedure: The flawed procedure the EPA used to calculate the EPA SUM06 data for the secondary ozone economic analysis. The calculation procedure is outlined in Appendix B. EPA Ozone Data: The 1990 hourly ozone concentrations from the AIRS database that the EPA used to calculate ozone exposure estimates for the secondary ozone economic analysis. EPA SUM06 Data: The 1990 three month maximum daily SUM06 ozone exposure estimates for each AIRS monitoring site in the EPA ozone data set. These data were created by NHEERLWED and used as input into their GIS model to generate the EPA SUM06 grid. EPA SUM06 Grid: 1990 three month maximum daily SUM06 ozone exposure estimates on a 10 km grid over the conterminous US. This grid was generated from the NHEERL-WED model using the EPA SUM06 data. Non-EPA Ozone Data: 1990 hourly ozone data from the CASTNet, AIRS, and other smaller monitoring networks that were not used in EPA’s secondary ozone economic analysis. Non-EPA SUM06 Data: 1990 three month maximum daily SUM06 ozone exposure estimates calculated from the non-EPA ozone data set using the EPA SUM06 calculation procedure outline in Appendix B. Reported SUM06 Calculation Procedure: The calculation procedure reported to be used by the EPA to calculate SUM06 ozone exposure estimates from the EPA ozone data for the secondary ozone economic analysis. The calculation procedure is outlined in Appendix A. 1/r2 SUM06 Grid: 1990 three month maximum daily SUM06 ozone exposure estimates on a 10 km grid over the conterminous US. This grid was generated by spatially interpolating the EPA SUM06 data using an inverse distance square weighting procedure. 10 TABLES Table 1. The average and standard deviation of the three month maximum daily SUM06 ozone exposure estimates calculated from the EPA data. The monitoring sites have been grouped into urban, suburban, rural , and agricultural and forest categories based on AIRS site setting and land use criteria. 20 unclassified sites were grouped in the All Sites category. SUM06 Ozone Values Calculated from EPA Data Monitoring Site Classification Urban Suburban Rural Agric. & Forest All Sites # Stations 138 324 84 162 728 Average ppmhrs 17.9 22.2 21.9 22.0 21.4 Standard Deviation 13.9 15.2 18.3 14.7 15.5 Table 2. The average and standard deviation of the three month maximum daily SUM06 ozone exposure estimates calculated from the non-EPA data. Non-EPA Measured Data Monitoring Site Classification # Stations Average ppmhrs Agric. & Forest 61 21.9 Standard Deviation ppm-hrs 12.2 Table 3. The average of the difference between the EPA SUM06 grid and data, and the root mean square error (RMS). The monitoring sites have been grouped into urban, suburban, rural , and agricultural and forest categories based on AIRS site setting and land use criteria. Note, the table does not contain all of the EPA SUM06 values. Grid cells whose center lay outside of the US border had been removed from the EPA SUM06 grid. 44 EPA monitoring sites were located in these grid cells. EPA SUM06 Grid - EPA SUM06 Data Monitoring Site Classification Urban Suburban Rural Agric. & Forest All Sites # Stations 131 297 78 158 684 Average ppm-hrs 1.05 0.55 -0.09 0.37 0.47 RMS ppm-hrs 3.35 2.76 4.75 2.74 3.16 Table 4. The average of the difference between the EPA SUM06 grid and non-EPA SUM06 data, and the root mean square error (RMS). EPA SUM06 Grid - Non-EPA SUM06 Data Monitoring Site Classification Agric. & Forest # Stations 61 Average ppm- RMS ppm-hrs hrs -5.49 9.98 11 FIGURES Figure 1. Data flow diagram for the evaluation of the EPA three month maximum daily SUM06 ozone exposure estimates. The four data flow paths are Data: Flow 1, validation of EPA’s calculation of the SUM06 data values, Data Flow 2, validation of the EPA SUM06 grid, Data Flow 3, evaluation of the EPA SUM06 grid in agricultural and forested regions, and Data Flow 4, evaluation of a 1/r 2 SUM06 grid in agricultural and forested regions. See the glossary for the definition of terms. 12 Figure 2. Location of the AIRS monitoring sites comprising the EPA data. The urban (squares), suburban (pentagon), rural (triangles), and agricultural and forested (circles) classifications are based on the AIRS setting and land use classifications. Figure 3. Location of the CASTNet, AIRS, and other ozone monitoring sites comprising the non-EPA data. All locations in the non-EPA data set were classified as agricultural and forested. 13 Figure 4. The EPA 1990 three-month maximum daily SUM06 ozone exposure grid. The exposure grid was calculated by EPA using NHEERL-WED’s Geographic Information System model to spatially interpolate SUM06 values calculated from the AIRS monitoring network. Figure 5. Verification of EPA’s SUM06 calculation procedure. A) EPA SUM06 data compared to SUM06 values calculated from the EPA ozone data using the reported SUM06 calculation procedure in Appendix A. B) EPA SUM06 data compared to SUM06 values calculated from the EPA ozone data using the flawed EPA calculation procedure. 14 Figure 6. Scatter plot comparing EPA SUM06 grid values to the EPA SUM06 data. The solid line is the linear regression line. The dotted line is the line of one-to-one correspondence. Figure 7. Scatter plot comparing EPA SUM06 grid values to the non-EPA SUM06 data. The non-EPA SUM06 data were calculated using the EPA procedure outlined in Appendix B. The solid line is the linear regression line. The dotted line is the line of one-to-one correspondence. 15 Figure 8. The difference between the 1990 EPA SUM06 grid and the non-EPA SUM06 data. The non-EPA SUM06 data were calculated using the EPA procedure. Figure 9. Scatter plot comparing the 1990 1/r 2 SUM06 grid values to the non-EPA SUM06 data. The 1990 1/r2 SUM06 grid was calculated by spatially interpolating the EPA SUM06 data using an inverse distance squared weighted interpolator. The solid line is the linear regression line. The dotted line is the line of one-to-one correspondence. 16 Figure 10. The three-month maximum daily SUM06 ozone exposure grid for 1988-92. The data were generated by EPA using NHEERL-WED’s GIS model. 17 Figure 11. The average 3-month maximum SUM06 ozone exposure estimates and percentage of monitoring sites exceeding the proposed 25 ppm-hrs standard from 1986 – 1995 for six regions of the US (A-F), the continental US (G). The six approximately equal area averaging regions (H). 20 40 10 20 0 0 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 50 Upper Midwest D 100 80 30 60 20 40 10 20 0 0 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 Northwest F 100 80 30 60 20 40 10 20 0 0 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 Conterminous US 40 80 30 60 20 40 10 20 0 0 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 Sum06, 3 Month Max 80 30 60 20 40 10 20 0 0 % Exceedance 30 20 10 1986 1987 1988 1989 1990 1991 E 50 40 30 20 10 0 1986 1987 1988 1989 1990 1991 % Exceedance Southwest G 100 40 80 30 60 20 40 10 20 0 0 Sum06, 3 Month Max 100 40 0 100 40 50 50 % Exceedance Lower Midwest 50 % Exceedance 50 40 30 20 10 0 1986 1987 1988 1989 1990 1991 % Exceedance Number of Valid Stations 500 Number of Stations 50 0 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 % Station Exceedance SUM06 (ppm-hrs) G 0 Sum06, 3 Month Max 40 Sum06, 3 Month Max 20 % Exceedance SUM06 (ppm-hrs) 50 10 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 % Station Exceedance SUM06 (ppm-hrs) E 40 Sum06, 3 Month Max 40 Sum06, 3 Month Max 20 % Exceedance SUM06 (ppm-hrs) SUM06 (ppm-hrs) C 60 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 % Station Exceedance Sum06, 3 Month Max 30 SUM06 (ppm-hrs) 60 80 SUM06 (ppm-hrs) 30 40 SUM06 (ppm-hrs) 80 C 100 % Station Exceedance 40 Southeast 50 % Station Exceedance B 100 % Station Exceedance Northeast SUM06 (ppm-hrs) 50 % Station Exceedance SUM06 (ppm-hrs) A 400 300 200 100 0 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 NE SE UMW LMW NW SW 18