|
|
|
|
SIP Suit ReportResponse to the EPA Brief in the SIP suitPrepared by G. Miller 2/27/00 In applying benchmark 1, the DEQ used an alternative ranking system that does not adhere to the requirements for alternative ranking systems in Sections 4.3.3 and 4.3.4 of the EPA's Attainment Guidelines. See EPA Guidelines at 22, 23. The way DEQ applied the alternative ranking system does not meet the requirements of these sections of the Guidelines. Specifically, there is no correlation between observed and predicted ozone values presented, there is no indication that the alternate ranking procedure's ability to predict variance exceeds that of the default procedure, and the adjusted R2 value is not given. In this circumstance the Guidelines require that the Deterministic Approach be used in the attainment demonstration instead of the Statistical Approach. See EPA Guidelines at 22, 23. The DEQ did not use a ranking system based on ozone forming potential, as required in the Guidelines. Instead, DEQ simply calculated the number of times a set of meteorological conditions would be expected to occur each year. See revised SIP, Appendix N, page 564. Since the DEQ did not apply a ranking system based on ozone forming potential, they didn't correctly calculate the needed ExEx values. The Guidelines require that ExEx must be calculated as ExEx = (Ranking -1)/(# of years used for ranking). See EPA Guidelines at page 12. The attainment demonstration for August 19, 1993 can be used to show the failure of DEQ's ranking system. The CART methodology placed August 19 in a meteorological category containing seventeen days that had similar meteorological characteristics. Of these seventeen days, only one, August 19, 1993, had an ozone exceedance. Therefore, these meteorological characteristics would be expected to produce an ozone exceedance only once in seventeen times. This indicates that these meteorological conditions have a low ozone forming potential. DEQ calculated the ExEx value for August 19, 1993, by simply taking the seventeen days and dividing it by the number of years in the study, which was nine. This is how DEQ arrived at the incorrect ExEx value of 1.9 for August 19. The value of 1.9 says nothing about the ozone forming potential of this class of meteorological conditions, but is simply the number of times these conditions are expected to occur in any given year. The actual ExEx for August 19 can be determined from information in Appendix N of the SIP. This can be done because DEQ did rank the study days by ozone forming potential, they just didn't report the results or use this information to correctly calculate the required ExEx values. There was also no evidence presented that this ranking procedure met the requirements of Sections 4.3.3 and 4.3.4 of the Guidelines. Appendix N states that, "Approximately 15 percent of the days within the CART dataset were estimated to have greater ozone forming potential" than August 19, 1993. See the revised SIP at 570. Since the dataset includes 2,205 days, March through October for nine years, the actual rank for August 19, 1993, can be calculated as 15 percent times 2,205. This gives the rank of August 19 as 330. The associated ExEx can then be determined as ExEx = (Ranking -1)/(# of years used for ranking) = (330-1)/2,205 = 25.4. Since an ExEx value of 25.4 is greater than 2.0, August 19, 1993, didn't pass benchmark 1 because no modeled exceedances are allowed in this case. See EPA Guidelines at 9. Therefore, either the alternative ranking system doesn't meet the requirements of Section 4.3.3 and 4.3.4 and the deterministic approach must be used, or benchmark 1 is failed if the statistical approach is used. The correct ExEx values for the other two primary episode days can also be determined from information in Appendix N. See the revised SIP at 568 and 570. These are; ExEx = 14.5 for August 16, 1989 and ExEx = 0.33 for May 25, 1990. This shows that benchmark 1 fails again for August 16, 1989, because no modeled exceedances are allowed for days with ExEx greater than 2.0. Benchmark 1 also fails the test for exceedances in a subregion. Since there is only one severe day, May 25, 1990, the value of N becomes 1. The required value for (N - 1) is then (1-1) = 0, meaning that no modeled subregion is allowed to have any exceedances. See EPA Guidelines at 9. The EPA brief incorrectly states that the August 19, 1993, model overestimated by 17.9%. See EPA Br. at 52. DEQ correctly identifies the overestimation. See revised SIP at 28. This misused 17.9% value is a comparison of the highest modeled ozone value on that day to the highest measured value. Since there was no ozone monitor at the site of the highest modeled value, calculation of an over or underprediction at that point is not possible. The value the EPA brief should have used for comparative purposes is the "Average accuracy of the peak (%)" given in Table 7-10 on page 560 of the SIP. This is the value the DEQ used. See the SIP at 28. Table 7-10 shows that the model overpredicted by only 6.2% for August 19, 1993. This invalidates the EPA's argument that the modeled exceedances for August 19 can be attributed to overprediction by the model. Decreasing the maximum modeled value on August 19 by 6.2% still gives an exceedance of over 0.133 ppm, well above the 0.124 ppm value needed to pass benchmark 2. Extending the EPA's argument, Table 7-10 shows that the August 16, 1989, model was underpredicted by 3.5%. Adjusting the peak modeled ozone value for this day gives a maximum ozone model value of over 0.131 ppm for the August 16, 1989 model. This too is significantly over the value of 0.124 required to pass benchmark 2. The EPA brief states that the "EPA also analyzed the trends based on the number of exceedances in the Baton Rouge area" and that regarding this issue "LEAN's information is incorrect, and its focus is too narrow." See EPA Br. at 55. Unfortunately, neither the EPA nor the DEQ presented data or graphs to support this view. If the EPA had bothered to present graphs and trend data they would have shown that the number of exceedances each year, the severity of the maximum exceedance each year, and the number of exceedance days per year showed no improvement from 1991 through 1998. Since there is no downward trend in any of this data, the EPA's position that attainment by the attainment date is likely can't be validated by the collected ozone data. The only trend the EPA presented data for was the trend associated with the design value. What the EPA didn't bring out is the fact that the design value has never been below that required for attainment and that the design value increased consistently from 1993 through 1997. It is also unexplained why the EPA didn't use the recommended trend analysis discussed in Section 5.3.3 of the Guidelines, page 28. Based on the Guidelines, this should have been the first trend analysis reported. The EPA brief states that the trend data "present compelling evidence that the area would attain the ozone standard by November 1999". See EPA Br. at 56. Though we disagree with this statement, November 1999 has come and gone. On November 15, 1999, there were two monitors in the Baton Rouge nonattainment area that were not in attainment for ozone, with six other monitors precariously close to nonattainment. The Baton Rouge ozone nonattainment area did not attain the ozone standard by the attainment date. This is exactly what the attainment demonstration predicts, and this is exactly what happened. Footnote 27 in the EPA brief implies that a precursor study strengthens the trend analysis. This is incorrect as the precursor study only shows that precursor concentrations went down from 1985 to 1996. Precursor concentrations would be expected to decrease in the same proportion as the drop in ozone from 1985 to 1996. The precursor study doesn't determine that precursor levels are low enough to achieve attainment nor does the study make any attempt to determine the precursor levels required to achieve attainment. Footnote 28 of the EPA brief infers that the current case is similar to the example problem discussed in the Guidelines. This is incorrect. The example problem consists of three hypothetical modeling demonstrations that pass all the benchmarks except for a single, small failure of benchmark 2. This is in contrast to the Baton Rouge modeling failures which don't pass any of the benchmarks. In the example problem the absolute worst model just barely misses passing the benchmarks. The opposite was found for Baton Rouge, where the models show that the very best that Baton Rouge can do will still fail the benchmarks. Back to Air Projects page.
|