                                       
                                 EPA REGION 6
                                       
                   2015 8-HOUR OZONE TRANSPORT SIP PROPOSAL
                                       
                          TECHNICAL SUPPORT DOCUMENT
                                       
                                       
                                       
                                       
                             February 1[st], 2022
                                       
                                       
                                      By
                                  Erik Snyder
                       Lead Regional Air Quality Modeler
                          EPA Region 6  -  Dallas, TX

Contents
1.	Maintenance Receptor Methodology	3
2.	Underestimation Concerns	36
a.	Underprediction in the base case 2012 period of Texas ozone at DFW/HGB monitors	37
  b.	Underprediction in the TCEQ's base case 2012 period at downwind areas in Midwest and West.	52
c.	Underestimation of Future year projections at DFW/HGB receptors	57
d.	Underestimation in TCEQ's 2023 Future year projections at downwind areas	62
e.	Summary	70
3.	Comparison of TCEQ and EPA 2023 future modeling and identified maintenance and nonattainment receptors	71
4. 	Other Potential Modeling Concerns	82
5.	TCEQ Other Factor Analysis (Weight of Evidence)	83
6.	EPA Summary	102


This Technical Support Document (TSD) provides additional analysis of TCEQ's air quality modeling relied upon in its August 17, 2018, SIP submission addressing interstate transport obligations for the 2015 ozone NAAQS. As discussed in the notice of proposed disapproval, EPA has identified several concerns with the reliability in TCEQ's modeling and application of those model results in assessing Texas' potential contributions to downwind ozone. States are free to develop their own modeling, but that modeling and analysis of the modeling and other technical analyses must be technically supportable, and the EPA is obligated to assess and evaluate the reliability of each state's technical demonstration when determining whether the Act's requirements are met. Based on the EPA's evaluation of the SIP submission, the EPA is proposing to find that TCEQ's August 17, 2018, SIP submission does not meet the State's obligations with respect to prohibiting emissions that contribute significantly to nonattainment or interfere with maintenance of the 2015 ozone NAAQS in any other state. EPA's assessment and evaluation of the reliability of the technical demonstration in TCEQ's SIP is discussed in this document. Oklahoma also relied on TCEQ's modeling analysis in their SIP. EPA's concerns with TCEQ's SIP also impact the approvability of Oklahoma's 2015 Ozone Transport SIP.

1.	Maintenance Receptor Methodology 
      As discussed in Section V.A of this action, in addition to the use of an alternative modeling platform, TCEQ also created its own method for identifying maintenance receptors. TCEQ has not adequately explained or justified how its method for identifying maintenance receptors reasonably identifies areas that will have difficulty maintaining the NAAQS. The EPA proposes to find that TCEQ has not provided a sufficient technical basis for how its chosen methodology gives meaning to the CAA's instruction that states submit good neighbor SIPs that prohibit their states' emissions from interfering with the maintenance of the NAAQS in another state.  
      In North Carolina v. EPA, 531 F.3d 896, 909-11 (D.C. Cir. 2008), the D.C. Circuit rejected the EPA's CAIR on the basis that the EPA had not adequately given meaning to the phrase "interfere with maintenance" in the good neighbor provision. Specifically, North Carolina argued that it had counties that were projected to attain the NAAQS in the future analytic year but were at risk of falling back into nonattainment due to interference from upwind sources, particularly given year-to-year variability in ozone levels. The court agreed, holding that the EPA's rule did not adequately protect "[a]reas that find themselves barely meeting attainment." Id. at 910. Consequently, the EPA has developed a methodology, as described elsewhere in this action and used in its 2011 CSAPR and its 2016 CSAPR Update and Revised CSAPR Update, for identifying areas that may struggle to maintain the NAAQS. See 76 FR at 48227-28. The EPA's approach to addressing maintenance receptors was upheld in the EME Homer City litigation. See 795 F.3d 118, 136-37. It was also upheld in Wisconsin. 938 F.3d at 325-26. In Wisconsin, the court noted that four upwind states were linked only to maintenance receptors and rejected the argument that application of the same control level as the EPA imposes for those states linked to nonattainment receptors was unreasonable or unlawful absent a particularized showing of overcontrol. Id. at 327.
      To explain the differences between TCEQ's and the EPA's methodology for identifying maintenance receptors, it is helpful to provide some additional context of how the EPA projects future air quality. The EPA's air quality modeling guidance has long recommended developing a base design value (DV) (i.e., the design value that will be used as a starting point to model and analyze for purposes of projecting future air quality concentrations) that is the average of three DVs spanning a five-year period, centered around one year for which an emissions inventory will be submitted (e.g., if 2011 was the base emissions inventory year, a state would use monitored values from 2009-2011, 2010-2012, 2011-2013 as the starting point for projecting air quality concentrations in future years). The average of these three DVs is then multiplied by a relative response factor (RRF) to generate an average DV for the future year. If a receptor's average future year DV is greater than or equal to the level of the NAAQS, and the receptor has recent monitored data that violates the NAAQS, that receptor is considered a "nonattainment" receptor at Step 1. To identify maintenance receptors, the EPA's methodology looks to the highest DV of the three DVs used to calculate the 5-year weighted average design value (e.g., in the 2011 example, if 2009-2011 had the highest design value of 2009-2011, 2010-2012, and 2011-2013). The EPA then applies the same relative response factor to that highest design value to generate a projected future maximum design value. Where a receptor's maximum design value exceeds the level of the NAAQS, the EPA has deemed those receptors to be "maintenance" receptors. This methodology was designed to address the D.C. Circuit's holding that the CAA's "interference with maintenance" prong requires states and the EPA to protect areas that may struggle with maintaining the standard in the face of inter-annual variability in ozone-conducive conditions.
      In its modeling, TCEQ adopted an identical approach to the EPA's for identifying nonattainment receptors -- it looked at three sets of DVs over a five-year period and averaged those DVs to generate a base year DV. TCEQ then applied a relative response factor to that base year design value to project a receptor's average design value in the future year. For its maintenance receptors, however, TCEQ used only the most recent design value of the set of three DVs, regardless of whether the most recent design value was highest or lowest, instead of considering variability in conditions over a five-year period, or using the highest DV of the three DVs making up the base year design value. TCEQ's proffered explanation for using the most recent DV to identify maintenance receptors was that the latest DV "takes into consideration . . . any emissions reductions that might have occurred." However, TCEQ in its submission does not explain how this methodology takes into account meteorological variability in identifying those areas that may be meeting the NAAQS or that may be projected to meet the NAAQS but may nevertheless struggle to maintain the NAAQS. 
      TCEQ argued that the 3-year DV used includes some meteorological variability. Unfortunately, the three years of variation that TCEQ accounted for is already built into the structure of the standard. Thus, the TCEQ method gave no consideration to the variability between calculated DVs, which provides a direct indication of the difficulty a receptor will have in maintaining the standard. In other words, to determine whether a receptor will have difficulty maintaining the standard, one must consider the variation in the metric that will be used to determine compliance with the standard. An indication of the variability of a metric cannot be determined by only considering a single estimate of that metric. 
      TCEQ's stated purpose in using the most recent DV was to capture more recent emissions reductions.  TCEQ's methodology, however, limits receptors which could be identified as maintenance receptors, compared to the EPA's methodology largely because it only looks at one design value period rather than selecting the maximum of the three DV periods EPA's methodology considers. Thus, TCEQ's methodology greatly reduces the probability that meteorological conditions which make it difficult to maintain the standard will be considered. As discussed further below, the effects of emissions trends are already captured through other aspects of the methodology to identify receptors. So, in trying to give more weight to emission reductions, by selecting only one design value (2012-2014) for its base year, TCEQ's methodology did not give any consideration to interannual variability in ozone-conducive meteorology as does the EPA's method.
   The EPA's methodology, using the maximum DV which accounts for the variability in ozone concentrations and DVs due to changes in meteorology over the five years of the base year DV period, was designed to identify those areas that might struggle to maintain the NAAQS in particularly ozone conducive conditions. TCEQ claimed that the EPA's method undervalues changes in air quality due to emission reductions and overvalues changes due to variation in meteorology. TCEQ pointed out that emissions nationwide are generally trending downward as a result of federal motor vehicle standards and other technological improvements. The EPA agrees that ozone levels generally trend downward, but there is not a steady decline from year to year in ozone concentrations. Rather, ozone levels tend to vary from year to year with some years showing an increase instead of a decrease mainly due to inter-annual variability in ozone-conducive meteorology. The variation of DVs at individual monitors from year to year can be significant, even where emissions trend downwards. The EPA also assessed a number of monitored DV trends that were provided in TCEQ's SIP submission and previous TCEQ attainment demonstration SIPs indicating that there are at times large annual fluctuations upward from year to year in monitored DVs (sometimes 2-3 ppb increase in one year) that are due to variations in meteorology. As discussed further below, EPA also assessed recent monitoring trends in Texas and in downwind areas identified by EPA or TCEQ's modeling. This is precisely why it is important to consider highly variable meteorology and its influence on DVs -- the issue at the heart of the D.C. Circuit's finding on "interference with maintenance" in North Carolina. Areas that are required under the Act to attain by an attainment date may fail to attain because of a combination of both local emissions, upwind emissions, and ozone conducive meteorology, among other factors. The North Carolina decision made clear that in interpreting the good neighbor provision, upwind state and the EPA obligations to reduce emissions must account for variable conditions that could cause an area that is sometimes attaining the NAAQS to fall out of attainment. See also Wisconsin, 938 F.3d at 327 ("Variations in atmospheric conditions and weather patterns can bring maintenance receptors into nonattainment even without elevated emissions."). 
      In addition, TCEQ claimed that its use of the 2012-2014 DV (i.e., the most recent in the 5-year base period it examined) is more reliable than the EPA's method because the more recent DV accounts for both emission reductions and because there is a shorter interval between the monitored DV and the projected DV. As we note elsewhere, the TCEQ's base year modeled inventory is 2012 emissions and the TCEQ's model projections for 2023 include the expected emission reductions from 2012 thru 2014 as well as to 2023. By just using the 2012-2014 DV data, TCEQ claimed they are giving weight to emission reductions during the final base years where EPA's method does not. The effect of emission reductions, however, is already factored in the modeling analysis since the modeling projection to 2023 is explicitly designed to project the changes in ozone due to emission reductions from the 2012 base year emission levels.  So, in fact, the EPA method does give weight to emission reductions by including the 2012-2014 as one of the 3 DVs considered in EPA's maintenance methodology despite emission reductions in 2012 and 2014 are also included in the model projections from 2012 to 2023. The TCEQ's modeling includes emission reductions that occur in 2013 and 2014 that are also reflected in the monitoring values that TCEQ is using (2012-2014 monitored DVs). Thus,  in essence, TCEQ is double counting the benefit of 2013 and 2014 emission reductions. Furthermore, since TCEQ agrees that the average of the DVs based on 2010-2014 ozone levels are reliable enough to use in the identification of nonattainment receptors, it is unclear how the 2012-2014 period is deemed more reliable for the identification of maintenance receptors since the modeled emissions are still for 2012. We also note, as discussed throughout this action, the EPA has updated its modeling to use a 2016 base year -- that is, a five year period spanning 2014-2018, and applied its methodology for defining maintenance receptors using that five year base period. Using a more recent base period (EPA's 2016v2) provides the most recent design values, shorter period of projection (2016 to 2023 versus a 2011 or 2012 base year), and a more accurate basis for projections of future air quality. We note that the EPA undertook a large collaborative multi-year effort with states (including Texas) and other stakeholders input and review in developing the 2016v2 emission inventories. By virtue of this update, any monitored DV used by the EPA to identify maintenance receptors in this action accounts for more recent emission reductions and provides a shorter interval between base year monitored DV and the projected future analytic year.
      The EPA has reviewed the set of 21 receptors for which Texas had contributions of 0.7 ppb or more in the EPA's 2016 base year modeling analyses, or TCEQ's modeling (2012 base year), and evaluated the results of using TCEQ's alternate maintenance methodology (see Table 1.1 below).  In Table 1.1 below, the EPA compiled the 2023 Average DVs, which is the DV that is used to identify if a receptor may be a nonattainment receptor for the receptors that EPA and TCEQ identified in Illinois, Wisconsin, Colorado, Arizona, and California. TCEQ's modeling uses a base year of 2012 and the EPA's 2016v2 modeling uses a base year of 2016.  EPA utilized TCEQ's modeling file outputs to obtain the data needed to calculate 2023 Average and Maintenance DVs for the receptors in Illinois and Wisconsin. For these 21 receptors, TCEQ's method resulted in 15 of the 21 2023 maintenance DVs predicted to be lower than the 2023 nonattainment DVs from the nonattainment methodology that uses the 5-year center weighted average. Of these 15 receptors, three receptors have 2023 maintenance DVs that are 3 ppb lower, five receptors have 2023 maintenance DVs that are 2 ppb lower, and seven receptors have 2023 maintenance DVs that are 1 ppb lower. In comparison, using the EPA's maintenance methodology results in all 21 2023 maintenance DVs being equal or up to 4 ppb higher than the 2023 nonattainment DVs. Again, the EPA uses the average of the three DVs that contain the base year modeled for the nonattainment methodology and the maximum of these three DVs for the maintenance methodology. Because TCEQ's maintenance methodology of just using the most recent DV (2012-2014 DV) often results in maintenance DVs lower than the 2023 nonattainment DVs methodology results, the EPA finds that the TCEQ methodology is not adequately identifying conditions when a receptor would have more difficulty maintaining the standard. In fact, the TCEQ's method also identified one receptor in their SIP submission as a nonattainment receptor in 2023 that would not have been identified as a maintenance receptor, which further highlights the concern that TCEQ's method did not adequately identify areas that may struggle to maintain the standard. TCEQ did not address whether the three years that comprise the most recent design value (i.e., 2012, 2013, and 2014) had meteorological conditions highly conducive for formation of high ozone concentrations and thus would be an appropriate time period to assess whether area could have difficulty maintaining the standard. The EPA's analysis confirms that this time period is not highly conducive to ozone formation, at least for many receptors. The consequence of TCEQ's maintenance method is that it often results in lower DVs than the nonattainment methodology as demonstrated by our analysis, which indicates that it is often not considering conditions when an area would have difficulty maintaining the standard. It is also unreasonable to have a methodology that would not identify nonattainment receptors also as maintenance receptors. 
   TCEQ also made several additional assertions in support of their conclusion that their method for identifying maintenance receptors was the better reading of the CAA, compared to the EPA's. TCEQ claimed that its approach was more consistent with the CAA's concept of maintenance as areas that were formerly nonattainment and that have since attained and will continue to maintain by accounting for: (1) emissions reductions occurring in the later design values of the base DV period; (2) "commitments regarding contingency measures to address future emission reductions;" and (3) the impact of any maintenance plans that are in place. TCEQ also asserted that the EPA's approach conflates the likelihood of attaining the standard in a future year and the ability of an attainment monitor to maintain that attainment status. Specifically, TCEQ argued that because any remedies devised to address nonattainment monitors would have to apply to maintenance monitors, a practical consequence of the EPA's approach is that it could lead to over-control and that it might require upwind states to consider or implement controls when the downwind state in which the monitor is located does not have any obligations to control local emissions. TCEQ argued that this "conflation" of nonattainment and maintenance results in there being no independent meaning to "maintenance."
   With respect to the first of these assertions from TCEQ, we note that TCEQ's methodology for identifying receptors (like the EPA's) is entirely distinct from ozone designations under the Clean Air Act; neither TCEQ nor the EPA take current or presumed future designations of areas into account, and any implementation requirements like a maintenance plan under CAA section 175A, in identifying receptors. TCEQ's discussion, therefore, of maintenance plan contingency measures or maintenance plans generally is irrelevant and misplaced. None of the areas to which Texas is linked in the EPA 2016v2 modeling has been redesignated to attainment for the 2015 ozone NAAQS, and none of the areas to which Texas is linked in its own modeling has been redesignated to attainment for the 2015 ozone NAAQS. We also fail to see how TCEQ's approach to identifying maintenance receptors differs in any relevant respect from the EPA's approach with regard to the alleged "conflation" of projecting attainment in a future year rather than the ability of an attainment receptor to maintain attainment. Both TCEQ and the EPA identify maintenance receptors based on projections of air quality in a future year to determine whether the receptor will have difficulty attaining or maintaining the standard. TCEQ's arguments about overcontrol based on the application of a uniform remedy to states linked to both nonattainment and maintenance receptors were also not germane; in this case, TCEQ had identified no remedy to apply whatsoever because it had failed to identify that the emissions from Texas cause a problem in the first instance. The D.C. Circuit has already rejected the idea that the application of a uniform control to both nonattainment and maintenance receptors is on its face overcontrol or impermissible under the interstate transport provision. See Wisconsin, 938 F.3d at 327. Based on our evaluation of TCEQ's approach to identify maintenance receptors for 2023, we propose to find the State's approach is inadequate as it does not sufficiently identify maintenance receptors. Further, TCEQ had not explained how its approach meets the statutory requirement to address areas that, even if meeting the NAAQS, may struggle to maintain the standard in years where conditions are conducive to ozone formation. Rather, the TCEQ had created its own approach to identify these areas that they describe as designed to account for the most emission reductions possible -- i.e., the most recent DV of the three under analysis; an approach that likely under-identifies areas that will struggle to maintain the NAAQS and that certainly is not designed to capture potential air quality problems.


Table 1.1 Comparison of Nonattainment methodology (Avg DV) and Maintenance methodology results for receptors identified by EPA and TCEQ in modeling of 2012 and 2016 base case years.
                             Receptor (Monitor ID,
                                County, State)
                          TCEQ 2023 Average DV (ppb)
                        TCEQ 2023 Maintenance DV (ppb)
                     TCEQ DV Delta (Maintenance - Average)
                                     (ppb)
              EPA 2023 Avg DV/Maintenance DV                (ppb)
                       EPA DV Delta (Maintenance - Avg)
                                     (ppb)
                                       





                             80350004, Douglas, CO
                                      73
                                      72
                                      -1
                                     71/72
                                       1
                            80590006, Jefferson, CO
                                      72
                                      73
                                       1
                                     72/73
                                       1
                            80590011, Jefferson, CO
                                      71
                                      71
                                       0
                                     73/74
                                       1
                             80690011, Larimer, CO
                                      72
                                      71
                                      -1
                                       ?
                                       ?
                            80050002, Arapahoe, CO
                                      70*
                                      71
                                       1
                                     68/68
                                       0
                             40038001, Cochise, AZ
                                      71
                                     69**
                                      -2
                                     70/72
                                       2
                           60371201, Los Angeles, CA
                                      80
                                      78
                                      -2
                                     82/85
                                       3
                           60371701, Los Angeles, CA
                                      80
                                      82
                                       2
                                     85/88
                                       3
                           60376012, Los Angeles, CA
                                      87
                                      86
                                      -1
                                     91/93
                                       2
                            60658001, Riverside, CA
                                      88
                                      85
                                      -3
                                     89/90
                                       1
                            60658005, Riverside, CA
                                      84
                                      83
                                      -1
                                     87/90
                                       1
                         60710001, San Bernardino, CA
                                      71
                                      72
                                       1
                                     74/75
                                       1
                         60710306, San Bernardino, CA
                                      76
                                      77
                                       1
                                     76/78
                                       2
                         60711004, San Bernardino, CA
                                      91
                                      90
                                      -1
                                    97/100
                                       3
                         60714001, San Bernardino, CA
                                      82
                                      79
                                      -3
                                     82/83
                                       1
                         60714003, San Bernardino, CA
                                      94
                                      91
                                      -3
                                     95/98
                                       3
                          170310001, Cook County, IL
                                      60
                                      58
                                      -2
                                     69/73
                                       4
                          170310032, Cook County, IL
                                      68
                                      66
                                      -2
                                     69/72
                                       3
                          170314201, Cook County, IL
                                      64
                                      62
                                      -2
                                     69/73
                                       4
                          170317002, Cook County, IL
                                      66
                                      65
                                      -1
                                     70/73
                                       3
                         550590019, Kenosha County, WI
                                      67
                                      66
                                      -1
                                     72/73
                                       1
                         550590025, Kenosha County, WI
                                   No data*
                                   No data*
                                   No data*
                                     69/72
                                       3
                         551010020, Racine County, WI
                                   No data*
                                   No data*
                                   No data*
                                     71/73
                                       2
* Monitors were installed in 2013 and 2014 so they did not have a 3 year DV period (2012-2014) or previous years, so no projections future year DVs are possible with TCEQ's modeling. Kenosha 550590025 was installed and began operating May 13, 2013, so the first three year DV available is 2013-2015. Racine 551010020 was installed in April 14, 2014 so the first three year DV available is 2015-2017
** Calculated from the Relative Response Factor in TCEQ spreadsheet of future 2023 DVs with state contributions and the monitor's 2012-2014 DV (0.983 X 71 ppb, truncation applied).
      EPA's modeling guidance for 8-hour ozone NAAQS attainment demonstrations (1999- 2007)  initially considered using a three year period that included the modeled year (inventory year modeled) as the middle year of the three year period but the EPA ultimately went with 5 consecutive years with the modeled year being the middle year (covering the three DVs that include the base year) to address potential variability in meteorological/ozone conduciveness and emissions. EPA analyzed the variability of using this 5 years center-weighted average method compared with using a 3 year average DV and found that the 5 year center-weighted approach was more stable and had a lower standard deviation. The goal of using 5 years of data as a basis for an attainment demonstration is that such demonstrations be based on an assessment of a range of meteorological conditions rather than a limited set of conditions in one 3-year DV period that may be conducive to higher or lower ozone levels. EPA's use of the highest maximum 3-year DV that includes the base modeled year is intended to provide an indication of conditions that might make it difficult for an area to maintain the ozone NAAQS. As such it accounts for potential meteorological and emissions variability that TCEQ's method does not. Also, TCEQ did not support their claim that there were large reductions in emissions that made the higher DV periods of 2010-2012 or 2011-2013 unreasonable to be used for determining a maintenance receptor.  Furthermore, TCEQ did not provide sufficient data and analysis of the meteorology for the 2010-2014 period to support their claim that 2012-2014 period was a worst-case combination of meteorology compared to the 2010-2012 and 2011-2013 periods. If the future DV projected from this highest value is below the standard, one can be reasonably certain the receptor will not have difficulty maintaining the standard and, as such, upwind states will not interfere with maintenance in downwind states. Because the TCEQ method only looks at one DV and does not account for the variability in DVs due to meteorological conditions, it is less likely to identify maintenance receptors than the EPA method. 
      To further examine TCEQ's assertion that the last DV period including the modeled base year should be used in the maintenance methodology, EPA reviewed a number of ozone design value trends, including several that were included in TCEQ's SIP and in a previous attainment demonstration SIP for Dallas-Fort Worth Nonattainment Area (DFW). EPA also looked at monitoring trends for the most recent 10 years of certified data (2009-2011 DV up to 2018-2020 DV) using ozone monitoring data from EPA's Air Quality System (AQS).  In general, as shown in the figures below, our evaluation of monitor data for Dallas-Fort Worth and Houston-Galveston-Brazoria as well as at at receptors that have been identified in Colorado, Arizona, Southern California, and Midwest Region (Illinois, Michigan, and Wisconsin), the long-term average annual decreases in ozone due to emissions reductions are usually on the order of 0-1 ppb/year at monitors in large metropolitan nonattainment areas. However, variations in monitored 8-hour DVs are not always negative from year to year and sometimes increase by 2 through-4 ppb from one year to the next due to meteorological variations.
      
      For the DFW area, looking at Figures 1.1 and 1.2 (From DFW AD SIP) the long-term DFW DV trend is less than 1 ppb/year in Figure 1.1 and 1.1 ppb/yr in Figure 1.2. For DFW, we started with the 2010-2012 DV as there were some reductions from a previous DFW ozone SIP that had compliance dates in early 2010. In Figure 1.3, the DVs at most of the higher DFW area monitors (2018-2020 DVs greater than 70 ppb) are decreasing at approximately 1 ppb/year for the last 9 years of certified DVs. In evaluating Houston-Galveston-Brazoria trends, TCEQ implemented multiple emission reductions for both NOx and a subset of VOCs that resulted in large decreases of ozone from 2000 to 2007, so EPA is evaluating trends for DVs including 2007 and later as there has not been as significant emission reductions implemented by TCEQ since that time. For the Houston-Galveston-Brazoria Nonattainment area (HGB), looking at Figures 1.4, 1.5, and 1.6 (from HGB AD SIP) the long-term HGB DV trend and DV trends at the monitors with higher ozone DVs from 2009 DV to current DV is less than 1.0 ppb/year. EPA evaluated the trends using the most recent DVs available in Figure 1.6. In Figure 1.7, the DVs at most of the higher HGB area monitors (2018-2020 DVs greater than 70 ppb) are decreasing at 1 ppb/year or less for the last 10 years of certified DVs (2011-2020 DVs). 
      EPA also evaluated DV trends using the most recent DVs available for the last 10 years for receptors identified by EPA's 2011 base year modeling, the EPA's 2016 base year (2016v2) modeling analyses, or TCEQ's modeling (2012 base year), in Colorado, Arizona, Southern California and the Midwest Region including receptors in Illinois, Wisconsin and Michigan. For the Colorado receptors see TCEQ's SIP Figure 3-40 (included as Figure 1.8 below) that show ozone DVs  from 2007-2016 decreased less than 1 ppb/yr on average. EPA also evaluated the trends at the Colorado receptors using the most recent DVs available EPA in Figure 1.9, which shows that the DVs at Colorado receptors have either increased or slightly decreased at less than 0.5 ppb/year for the last 10 years of certified DVs (2011-2020 DVs). For the Arizona receptor see TCEQ's SIP Figure 3-55 (included as Figure 1.10 below) that show ozone DVs from 2007-2016 decreased less than 1 ppb/yr on average but had several ppb increases or decreases from one year to the next. EPA also evaluated the trends at the Arizona receptor using the most recent DVs available in Figure 1.11, which shows that the DVs at Arizona receptor have decreased at less than 1.0 ppb/year on average for the last 10 years of certified DVs (2011-2020 DVs). California receptors (see TCEQ's SIP Figure 3-40 included as Figure 1.12 below) show ozone DVs from 2007-2016 decreased less than 1 ppb/yr on average. EPA also evaluated the trends at the California receptors using the most recent DVs available EPA in Figure 1.13 which show that the DVs at California receptors have either increased or are slightly decreased at less than 0.5 ppb/year on average for the last 10 years of certified DVs (2011-2020 DVs). EPA evaluated the trends at the Chicago Nonattainment Area receptors (IL and WI) using the most recent DVs available EPA in Figure 1.14, which show that the DVs at these receptors have either increased or are slightly decreased at less than 0.7 ppb/year on average for the last 10 years of certified DVs (2011-2020 DVs). EPA also evaluated the trends at the other Wisconsin and Michigan receptors using the most recent DVs available EPA in Figure 1.15, which show that the DVs at Sheboygan Wisconsin and Allegan Michigan receptor have decreased approximately 1 ppb/year on average for the last 10 years of certified DVs (2011-2020 DVs).
      The long-term average annual decreases in ozone due to emissions reductions are usually on the order of 0-1 ppb/year at monitors in large metropolitan nonattainment areas such as Dallas-Fort Worth and Houston-Galveston-Brazoria, and at receptors that have been identified in Colorado, Arizona, Southern California, and Midwest Region (Illinois, Michigan, and Wisconsin) but variations in monitored 8-hour DVs are not always negative and sometimes increase by 2-4 ppb from one year to the next due to meteorological variations.  We also included TCEQ's monitored trends figure for the Arizona receptor that also shows increases of 2 ppb and 3 ppb annually, and from 2009 to 2012 the DV increased from 66 ppb to 73 ppb for a total of 7 ppb increase in 3 years. These larger annual fluctuations upward in DVs demonstrate the amount of changes possible in DV due the meteorological difference between different 3 year design value periods. TCEQ's approach fails to take the much larger influence of meteorological variability into account in their maintenance receptor test.
Figure 1.1 - TCEQ's Figure 1-1: One-Hour and Eight-Hour Ozone Design Values and DFW Population, from DFW 2008 8-hour Attainment Demonstration SIP (July 6, 2016).



Figure 1.2 - TCEQ's Figure 5-1:One-Hour and Eight-Hour Ozone Design Values in the DFW Area from 1997 through 2014 from DFW 2008 8-hour Attainment Demonstration SIP (July 6, 2016).

                                       
        Figure 1.3 DFW Monitor 8-hour Ozone DV (2012-2020 DVs) trends.


Figure 1.4 -  TCEQ's Figure 1-1: Ozone Design Values and Population in HGB Area, from TCEQ's HGB Serious AD SIP Revision for 2008 Ozone (2020).
                                       




Figure 1.5 -  TCEQ's Figure 5-4: Eight-Hour and One-Hour Ozone Design Values in the HGB Area From TCEQ's HGB Serious AD SIP Revision for 2008 Ozone (2020)
                                       


Figure 1.6 - From TCEQ's HGB Serious AD SIP Revision for 2008 Ozone (2020) (Figure 5-5: Eight-Hour Ozone Design Values by Monitor in the HGB Area.)



Figure 1.7 HGB Monitor 8-hour Ozone DV (2011-2020 DVs) trends.

Figure 1.8 TCEQ's SIP Figure 3-40.


                Figure 1.9 EPA Denver NAA receptors DV trends.



                     Figure 1.10 TCEQ's SIP Figure 3-55.


                  Figure 1.11 EPA Arizona receptor DV trend.



                     Figure 1.12 TCEQ's SIP Figure 3-56.


           Figure 1.13 EPA Southern California receptors DV trends.



               Figure 1.14 EPA Chicago NAA receptors DV trends.


       Figure 1.15 EPA Other Wisconsin and Michigan receptors DV trends.




2.	Underestimation Concerns
	As discussed in this section below and in section 3, below, EPA has evaluated the future year model projections for the DFW and HGB areas within Texas and also at downwind areas (Colorado, Arizona, Southern California, Illinois, Wisconsin, and Michigan) that have been identified in EPA's 2011 base case modeling, EPA's 2016v2 modeling, and TCEQ's modeling and compared these 2023 projections with recent DVs (2018-2020) and preliminary DVs (2019-2021). This evaluation (TCEQ's 2023 modeled DVs to recent monitoring) and expected decreases in DVs indicates that the TCEQ's modeling underestimates future ozone levels. When the TCEQ 2023 projected concentrations are compared to 2020 and preliminary 2021 monitor values, it is clear that the TCEQ modeling is projecting an unusual decline in ozone levels without there being an unusual level of emission reductions to support the decline. The EPA compared recent monitoring values and reasonably anticipated decreases (0-1 ppb/year  -  see discussion on ozone trends) in DVs by 2023 both within Texas and in other parts of the country. These underestimations likely result in TCEQ's modeling not adequately identifying nonattainment and/or maintenance receptors in 2023. These underestimations also result in smaller projected contributions from Texas emissions to downwind states.
   In this section, the EPA has evaluated TCEQ's base case (2012) modeling for the DFW and HGB areas. TCEQ's modeling does have some underestimation of higher ozone levels in these two areas in the base case year that could lead to underestimation of Texas's emissions impacts on downwind areas in the future year and also in ozone levels downwind of Texas in the base year and future year. EPA also evaluated TCEQ's model performance for downwind areas and our review indicated some underestimation bias in the base case and general model performance concerns but nothing that was a clear cause of the much lower 2023 DVs that TCEQ's modeling is projecting. 
a.	Underprediction in the base case 2012 period of Texas ozone at DFW/HGB monitors 
      TCEQ used the same 2012 year for the base case year for DFW and HGB ozone attainment demonstration SIPs. TCEQ's 2012 Transport modeling base case has some of the same concerns noted in review of the attainment demonstration SIPs (i.e., that high ozone levels are often underestimated in DFW and HGB areas). TCEQ provides some Model Performance Evaluation data on a website that EPA used to generate the figures below demonstrating underestimation issues at key monitor/receptors in the DFW and HGB area. The EPA has provided time series that indicates the observed Maximum Daily Average 8-hour ozone (MDA8) value in red and modeled value in blue for two monitors in DFW area and one in the HGB area that were identified in EPA's March 2018 memorandum and three additional monitors (Eagle Mountain Lake, Grapevine, and Frisco) in the DFW area, which are some of the monitors that often set the DFW area DV (Figures 2.1 a-f).  These figures also include quantile-quantile (Q-Q) scatter plots, comparing observed to modeled MDA8 and showing the underestimation in the modeled values for the high ozone days.

Figure 2.1(a) Brazoria County  -  Manvel Croix Time Series MDA8 Observed (red) and Modeled (blue) and Scatter Q-Q plots.


Figure 2.1(b) Denton County  -  Denton Airport South Time Series MDA8 Observed (red) and Modeled (blue) and Scatter Q-Q plots.


Figure 2.1(c) Tarrant County  -  FAA Site Time Series MDA8 Observed (red) and Modeled (blue) and Scatter Q-Q plots. 

Figure 2.1(d) Tarrant County  -  Eagle Mountain Lake Time Series MDA8 Observed (red) and Modeled (blue) and Scatter Q-Q plots.

Figure 2.1(e) Tarrant County  -  Grapevine Time Series MDA8 Observed (red) and Modeled (blue) and Scatter Q-Q plots.

Figure 2.1(f) Collin County  -  Frisco Time Series MDA8 Observed (red) and Modeled (blue) and Scatter Q-Q plots.

EPA has also analyzed the data from the MDA8 time series (2012 Observed and 2012 Modeled) for these monitors and evaluated modeled daily MDA8 values for all days above thresholds of 70 ppb, 75 ppb, 80 ppb, and 85 ppb. Table 2.1 includes the mean, minimum, and maximum bias for days over the observed MDA8 threshold (70, 75, 80, and 85 ppb) and the number of days of positive and negative bias over the observed MDA8 threshold. EPA also calculated for each of these observed thresholds, the number of days that modeled negative bias is 0 to less than -5 bias, -5 to -10 ppb, and more than -10 ppb for the days over the observed MDA8 threshold.  

Table 2.1 DFW and HGB monitors that have been identified by EPA as nonattainment/maintenance receptors and additional high DV monitors.
Manvel Croix Brazoria County Texas AQS ID480391004
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >70 ppb
                                    -14.02
                                    -33.83
                                     -2.59
                                       0
                                      13
                                       1
                                       4
                                       8
OBS >75 ppb
                                    -16.20
                                    -33.83
                                     -6.74
                                       0
                                      10
                                       0
                                       2
                                       8
OBS >80 ppb
                                    -16.68
                                    -33.83
                                     -6.74
                                       0
                                       9
                                       0
                                       2
                                       7
OBS >85 ppb
                                    -21.81
                                    -33.83
                                    -12.14
                                       0
                                       5
                                       0
                                       0
                                       5
Denton Airport South Denton County Texas AQS ID 481210034
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >70 ppb
                                     -5.24
                                    -15.98
                                     5.19
                                       5
                                      19
                                       9
                                       4
                                       6
OBS >75 ppb
                                     -9.28
                                    -15.98
                                     -3.04
                                       0
                                      10
                                       3
                                       2
                                       5
OBS >80 ppb
                                    -10.28
                                    -15.98
                                     -5.29
                                       0
                                       5
                                       0
                                       2
                                       3
OBS >85 ppb
                                     -9.70
                                    -12.48
                                     -5.29
                                       0
                                       3
                                       0
                                       1
                                       2
                                       








FAA Tarrant County Texas AQS ID 484392003
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >70 ppb
                                     -5.49
                                    -14.97
                                     0.74
                                       3
                                      10
                                       3
                                       6
                                       1
OBS >75 ppb
                                     -5.63
                                    -14.97
                                     0.53
                                       2
                                       7
                                       2
                                       4
                                       1
OBS >80 ppb
                                     -3.75
                                     -7.66
                                     0.28
                                       1
                                       2
                                       1
                                       1
                                       0
OBS >85 ppb
                                     -3.75
                                     -7.66
                                     0.28
                                       1
                                       2
                                       1
                                       1
                                       0

Table 2.1 (continued) DFW and HGB monitors that have been identified by EPA as nonattainment/maintenance receptors and additional high DV monitors.
Eagle Mountain Lake Tarrant County Texas AQS ID 484390075
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >70 ppb
                                     -4.36
                                    -14.78
                                     5.39
                                       5
                                       7
                                       1
                                       4
                                       2
OBS >75 ppb
                                     -8.41
                                    -14.78
                                     -4.39
                                       0
                                       6
                                       1
                                       4
                                       1
OBS >80 ppb
                                     -8.31
                                    -14.78
                                     -4.39
                                       0
                                       4
                                       1
                                       2
                                       1
OBS >85 ppb
                                     -8.31
                                    -14.78
                                     -4.39
                                       0
                                       4
                                       1
                                       2
                                       1
                                       








Grapevine Tarrant County Texas AQS ID 484393009
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >70 ppb
                                     -7.06
                                    -18.30
                                     3.11
                                       7
                                      21
                                       4
                                       8
                                       9
OBS >75 ppb
                                     -9.47
                                    -18.30
                                     0.82
                                       1
                                      10
                                       0
                                       6
                                       4
OBS >80 ppb
                                    -12.23
                                    -18.30
                                     -6.37
                                       0
                                       6
                                       0
                                       3
                                       3
OBS >85 ppb
                                     -9.77
                                    -17.29
                                     -6.37
                                       0
                                       4
                                       0
                                       3
                                       1
                                       








Frisco Collin County Texas AQS ID 480850005
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >70 ppb
                                     -7.57
                                    -18.27
                                     4.11
                                       4
                                      22
                                       5
                                       8
                                       9
OBS >75 ppb
                                    -10.28
                                    -18.27
                                     1.17
                                       1
                                      13
                                       2
                                       3
                                       8
OBS >80 ppb
                                    -11.69
                                    -18.27
                                     -4.81
                                       0
                                       9
                                       1
                                       3
                                       5
OBS >85 ppb
                                    -15.57
                                    -18.27
                                    -12.52
                                       0
                                       3
                                       0
                                       0
                                       3


      As can be seen in Figure 2.1(a) and Table 2.1, TCEQ's modeling is not replicating the higher ozone monitored values on many days at the HGB Manvel Croix monitor in the time series and scatter plot. Table 2-1 shows that the model is underestimating ozone by 2.59-33 ppb on all days with monitored values above 70 ppb and a mean bias of -14.02 ppb for all days using the 70 ppb threshold. On the higher ozone days with a minimum observed MDA8 threshold of 75 ppb or 80 ppb, the bias is worse with the mean bias of greater than -16 ppb on all days and a mean bias of -21 ppb on days using the 85 ppb threshold. We have looked at some other monitors in the HGB area that are typically the highest ozone monitors in HGB and TCEQ's model performance varies but there is a general underprediction bias at several of the monitors.
      As can be seen in Figure 2.1(b) and Table 2.1, TCEQ's modeling is not replicating the higher ozone monitored values on many days at the DFW area Denton Airport South monitor in the time series and in the scatter plot. Table 2-1 shows that the model is underestimating ozone by 0-15.98 ppb on most days with monitored values above 70 ppb and a mean  bias of -5.24 ppb for all days using the 70 ppb threshold. On the higher ozone days with a minimum observed MDA8 threshold of 75 ppb or 80 ppb the bias is worse with the mean bias of -9 ppb or worse on all days. 
      As can be seen in Figure 2.1(c) and Table 2.1, TCEQ's modeling is not replicating the higher ozone monitored values on many days at the DFW area FAA Site monitor in the time series and in the scatter plot. Table 2-1 shows that the model is underestimating ozone by 0-14.97 ppb on most days with monitored values above 70 ppb and a mean bias of -5.49 ppb for all days using the 70 ppb threshold. 
      As can be seen in Figure 2.1(d) and Table 2.1, TCEQ's modeling is not replicating the higher ozone monitored values on many days at the DFW area Eagle Mountain Lake monitor in the time series and in the scatter plot. Table 2-1 shows that the model is underestimating ozone by 0-14.78 ppb on most days with monitored values above 70 ppb and a mean bias of -4.36 ppb for all days using the 70 ppb threshold. On the higher ozone days with a minimum observed MDA8 threshold of 75 ppb, 80 ppb, or 85 ppb the bias is worse with an underestimation of 4.39 to 14.78 ppb and mean bias of -8.3 ppb or worse on all days (-8.31 ppb and -8.41 ppb depending on which minimum cutoff). 
      As can be seen in Figure 2.1(e) and Table 2.1, TCEQ's modeling is not replicating the higher ozone monitored values on many days at the DFW area Grapevine monitor in the time series and in the scatter plot. Table 2-1 shows that the model underestimates ozone by 0-18.3 ppb on most days with monitored values above 70 ppb and a mean bias of -7.06 ppb for all days using the 70 ppb threshold. On the higher ozone days with a minimum observed MDA8 threshold of 75 ppb, 80 ppb, or 85 ppb the underestimation bias is worse. Specifically, the mean bias is -9.4 ppb or worse on all days (-9.47 ppb using 75 ppb threshold, -12.23 ppb using 80 ppb threshold, and -9.77 ppb using the 85 ppb threshold).
      As can be seen in Figure 2.1(f) and Table 2.1, TCEQ's modeling is not replicating the higher ozone monitored values on many days at the DFW area Frisco monitor in the time series and in the scatter plot. Table 2-1 shows that the model is underestimating ozone by 0-18.27 ppb on most days with monitored values > 70 ppb and a mean bias of -7.57 ppb for all days using the 70 ppb threshold. On the higher ozone days with a minimum observed MDA8 of 75 ppb, 80 ppb, or 85 ppb the underestimation bias is worse. For those thresholds,  the mean bias is -10.28 ppb, -11.69 ppb, and -15.57 ppb respectively. 
      Overall, TCEQ's 2012 base case modeling is not replicating the higher monitored ozone levels at a number of monitors in the DFW and HGB area. This indicates that TCEQ's modeling is not replicating the amount of ozone generated by the emissions from the largest source categories in these two large urban areas. This raises a concern that TCEQ's 2012 modeling underestimates ozone levels generated by Texas's emissions which results in the underestimation of both downwind ozone levels in other states and the amount of Texas's contribution to downwind states and their potential receptors. This underestimation of ozone levels from Texas's emissions in the base year can also lead to underestimations of Texas projected ozone generation in the future year modeling, underestimation of downwind ozone levels outside of Texas, and also underestimation of Texas's impact on downwind receptors. This underestimation that exists in TCEQ's modeling also raises a concern in how it impacts model projections from that the Oklahoma Department of Environmental Quality (ODEQ) relied upon in its 2015 8-hour Ozone Transport SIP to analyze Oklahoma's potential contribution to downwind ozone at receptors in DFW and HGB areas of Texas.
      To further evaluate ozone production in the model for the Texas areas, the EPA also evaluated monitors typically upwind of DFW (Corsicana monitor in Navarro County, Texas {AQS ID 483491051}and Italy monitor in Ellis County, Texas {AQS ID 481391044}) and HGB (a monitor in Jefferson County, Texas {AQS ID 482450022}and a Livingston monitor in Polk County Texas {AQS ID 483739991}). EPA evaluated the time series and scatter plots for these monitors available on TCEQ's website. EPA also generated the same information included in Table 2-1 for these monitors in Table 2-2 below. Since these are typically background monitors, the EPA used a lower threshold of  >= 60 ppb MDA8 to analyze model performance. Overall the mean bias at these four monitors that represent upwind/background for the DFW and HGB areas has a slight negative mean bias for days with monitored MDA8s >=60 ppb, but the underestimation is smaller with mean bias ranging from -1.0 ppb to -3.2 ppb. This further indicates that underestimation at the higher ozone values in DFW and HGB discussed above in the 2012 base case modeling is more of a result of the model not producing enough ozone in these nonattainment areas from local emission sources on the high ozone monitored days in the HGB and DFW. As discussed below, we also note Figures 2.2 and 2.3 below (reproduced from TCEQ's SIP Figure 3-26 and 3-27) indicate that most monitors in Texas have an underprediction bias and a high Root Mean Square Error (RMSE), further raising concern with the accuracy of Texas's modeling.

               Table 2.2. DFW and HGB Upwind/background Monitors
Corsicana TX AQS ID 483491051
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >60 ppb
                                     -3.2
                                     -14.0
                                      5.2
                                       4
                                      19
                                      12
                                       5
                                       2
                                       








Italy TX AQS ID 481410029
 
 
 
 
 
 
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >60 ppb
                                     -1.0
                                     -10.5
                                     12.6
                                      10
                                      16
                                      12
                                       3
                                       1
                                       








Jefferson County TX AQS ID 482450022
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >60 ppb
                                     -2.2
                                     -28.6
                                     10.8
                                       8
                                      10
                                       6
                                       1
                                       3
Livingston TX AQS ID 483739991
 
                  Modeled Value  vs Observed Value MDA8 (ppb)
                      Number of Days above OBS Threshold
                Number of Days Modeled MDA8 is negative biased
 



                                                            OBS Threshold Value
                                     MEAN
                                      MIN
                                      MAX
                                 Positive Bias
                                 Negative Bias
                               0 to > -5 ppb
                                 -5 to -10 ppb
                              More than (-10) ppb
OBS >60 ppb
                                     -1.6
                                     -7.8
                                      6.2
                                       3
                                       4
                                       1
                                       3
                                       0

b.	Underprediction in the TCEQ's base case 2012 period at downwind areas in Midwest and West.
      TCEQ included some Model Performance Evaluation (MPE) analysis in the main SIP documents that indicates model performance varied across the U.S. but had underestimation at key areas in the Midwest and west including the Chicago area, southeast Wisconsin, Denver, Arizona, and Southern California areas.  These areas include sites that were identified in EPA's modeling as nonattainment/maintenance receptors. We have included 2 figures from TCEQ's SIP that show Mean Bias and Root Mean Square Error (RMSE) for each monitor in the domain. Included as Figure 2.2 is TCEQ's Figure 3-26 in their SIP submission, which provides Mean Bias for the May through September 2012 Episode at all the EPA Air Quality System (AQS) monitoring sites for days with observed Maximum Daily Average Eight-Hour (MDA8) ozone concentration >= 60 ppb. Figure 2.2 shows that the mean biases for days with MDA8 ozone concentrations >= 60 ppb differ by region, but at a number of monitors in the Chicago, southeastern Wisconsin, Denver, Arizona, and southern California areas TCEQ's modeling has a negative mean biases of more than 5 ppb.  TCEQ's modeling having a mean bias underestimation of more than 5 ppb in areas that EPA's modeling identified as having nonattainment/maintenance receptors is concerning because the underestimation in TCEQ's modeling in the base and future years may result in their modeling not identifying receptors as nonattainment/maintenance receptors. 
      Figure 2.3 (TCEQ's Figure 3-27 in their SIP submission) provides RMSE for May through September 2012 Episode at AQS Monitoring Sites for each of the days with observed MDA8 > 60 ppb. The RMSE at most monitors in the modeling domain were in the range of 6 to 12 ppb. However, many monitors in the Midwest including Chicago area and southeastern Wisconsin and monitors in Arizona and California have worse model performance with higher deviations with RMSE in the 8 to 14 ppb range. 
      
    Figure 2.2 TCEQ's SIP Figure 3-26 showing Mean Bias at each monitor.


Figure 2.3 TCEQ' SIP Figure 3-27 showing Root Mean Square Error (RMSE) at each monitor.


TCEQ's model performance evaluation for their 2012 base case modeling is included in their SIP submission as "Appendix C  -  Photochemical Model Performance Evaluation for the Transport State Implementation Plan Revision for the 2015 Eight-Hour Ozone National Ambient Air Quality Standard".  EPA evaluated information in Appendix C and available on TCEQ's websites in our review of TCEQ's modeling.
      TCEQ's modeling underestimates observed values at key monitor/receptors in Cook County, IL, Wisconsin, and Michigan. This underprediction bias is especially troublesome at the monitors/receptors located in Kenosha County WI, Sheboygan County WI, and Allegan County, MI as many of these monitors typically have the highest DVs in the multi-state area around Chicago. The underprediction at these monitors/receptors in TCEQ's modeling likely results in some of these monitors not being identified as potential nonattainment/maintenance receptors. Both EPA's initial modeling and  most recent modeling has consistently identified nonattainment/maintenance receptors in the Chicago area and downwind areas in Wisconsin and Michigan with Texas emissions having impacts >=0.7 ppb.  TCEQ's modeling, however, is underestimating the ozone at these nonattainment/maintenance receptors in their 2012 base case modeling it does have similar contributions from Texas's emissions (>= 0.7 ppb).  If TCEQ's modeling had less underprediction issues and actually identified nonattainment and/or maintenance receptors in the Chicago Nonattainment Area or downwind areas in Wisconsin and Michigan, TCEQ would have identified linkages to Texas as EPA's modeling has consistently identified. 

      TCEQ's modeling also underestimates observed values at key monitor/receptors in Colorado. Even with these underprediction concerns, TCEQ's modeling analysis still identified 4 nonattainment/maintenance receptors in Colorado including the Jefferson County (AQS ID 80590011), which the EPA also identified as a receptor with Texas linkage in modeling results included in our March 2018 memorandum. If TCEQ's modeling had less underprediction bias, there may have been more monitors/receptors identified with Texas emissions linkage in Colorado and other downwind areas. The ability of TCEQ's modeling to replicate the highest days both in Texas and at downwind areas are important to having an adequate modeling analysis to both assess potential nonattainment/maintenance receptors and linkages to upwind states. In addition to the underprediction at these monitors/receptors that may result in some monitors not being identified as potential nonattainment/maintenance receptors, underprediction issues at monitors in Texas in TCEQ's modeling results in an underestimation of Texas's impacts at Colorado monitors. Both EPA's modeling included in the March 2018 memorandum and TCEQ's modeling identified nonattainment/maintenance receptors in Jefferson County, CO.  We also note that TCEQ's analysis also identified one additional receptor in Jefferson County, CO, one monitor in Larimer County, CO, and one receptor in Douglas County, CO with linkage impacts greater than 0.7 ppb from Texas emissions despite the multiple underestimation issues with TCEQ's modeling. 
      TCEQ's modeling also underestimates observed values at key monitor/receptors in Arizona and California that TCEQ identified as nonattainment and/or maintenance receptors with linkages to Texas.. Despite this underprediction of high ozone values within Texas and at these downwind receptors, TCEQ did identify Texas emissions as having impacts of more than 0.7 ppb, and thus, linked to some nonattainment/maintenance receptors in Arizona and California.
c.	Underestimation of Future year projections at DFW/HGB receptors

      TCEQ's modeling includes projections of future year 2023 Design Values for 2015 8-hour ozone levels for monitors within the large metropolitan areas of DFW and HGB, both of which continue to monitor ozone nonattainment and have been designated as 2015 8-hour ozone nonattainment areas. To support attainment demonstration planning, TCEQ has in the past performed photochemical modeling for both of these HGB and DFW nonattainment areas. TCEQ also included in these previous attainment demonstration SIPs that long term ozone DV trend analysis for both the HGB and DFW areas indicate that the average annual decrease in ozone DVs was approximately 1 to 1.2 ppb/year. EPA's analysis of the most recent 9 years of DVs in DFW and 10 years of DVs in HGB indicate a similar trend of 1 ppb/year or less. Using these DV trends from TCEQ's previous 2008 Ozone Attainment Demonstration SIP submittals indicate that emissions decreases between 2020 and 2023 would be expected to result in a drop in ozone concentrations of approximately 3-4 ppb if no variability in meteorology is considered and no large emission reduction programs are implemented in Texas or on a national level. In evaluating TCEQ's model projections for 2023 in their 2015 Ozone Transport SIP, this information provided a helpful comparison that indicates a good approximation of the difference in 8-hour ozone DVs at monitors/receptors in DFW and HGB from 2020 to 2023 should be approximately 3-4 ppb at most for the higher DV monitors.  That is to say, without significant unusual reductions, not just those expected to from fleet turnover, 3-4 ppb during this 3 year period is the most that should be expected.  To assess if TCEQ's ozone transport modeling is potentially underestimating future year 2023 modeled DVs in DFW and HGB areas, EPA  compared 2018-2020 monitored DVs at many typically higher ozone monitors in the DFW and HGB area with TCEQ's projected 2023 DVs in Table 2.3. While not as exact as developing new modeling of emission changes from 2020 to 2023 to project 2023 DVs, using the general 3-4 ppb provides a ballpark estimate to evaluate whether TCEQ's modeling might be underpredicting 2023 future DVs. As can be seen in Table 2.3, there are many monitors in DFW and HGB that would need a drop in ozone DVs over the next 3 years of  >5 ppb (values of -5 ppb or more in the far right column) to reach TCEQ's 2023 modeled projected levels. The average decrease needed to meet TCEQ's modeled DV projections is 4.97 ppb, which is more than 3-4 ppb. EPA also evaluated just the 2023 DVs for monitors with 2018-2020 DVs of 74 ppb or greater.  For these monitors, a 2023 monitored DV of 70 ppb or less is needed to monitor attainment, implying a drop of at least 3-4 ppb. For this subset of monitors with 2018-2020 DVs of 74 ppb or more, the average amount of decrease in DVs needed to match TCEQ's 2023 projection is 7.56 ppb. In other words, for those monitors with the highest monitored DVs for 2018-2020, TCEQ's modeling projected average reductions in DVs over the three year period of 7.56 ppb, much larger than the 3-4 ppb that would be typical for these areas. Both in TCEQ's past SIP modeling and EPA's modeling experience, drops in 8-hour ozone DVs on average of 7.56 ppb in three years do not typically occur unless there is a large change in emissions or large change in meteorological conduciveness for ozone generation. TCEQ did not identify any large emission reductions to be implemented in the 2021-2023 timeframe. This information supports our finding that TCEQ's modeling is underestimating future ozone levels in the two nonattainment areas in Texas that make up a large proportion of the total ozone and emissions of ozone pre-cursors that transport to downwind areas. This underestimation of future year ozone levels from Texas emissions can cause both an underestimation of ozone in downwind areas (discussed further below)  and an underestimation of Texas's impacts on downwind State's ozone nonattainment and maintenance receptors. 

Table 2.3 TCEQ Projected Future 2023 DVs Compared to 2018-2020 Monitored DVs.
State
County
AQS ID
                    TCEQ 2023 Future Design Value-DV (ppb)
                          EPA Monitored Data from AQS
                              2018-2020 DV (ppb)
         Delta in 3 years 2018-2020 Monitor DV - TCEQ Future DV (ppb)
Texas
Brazoria
                                                                      480391004
                                                                             78
                                                                             73
                                                                              5
Texas
Brazoria
                                                                      480391016
                                                                             60
                                                                             65
                                                                             -5
Texas
Collin
                                                                      480850005
                                                                             66
                                                                             75
                                                                             -9
Texas
Dallas
                                                                      481130069
                                                                             64
                                                                             69
                                                                             -5
Texas
Dallas
                                                                      481130075
                                                                             63
                                                                             74
                                                                            -11
Texas
Dallas
                                                                      481130087
                                                                             61
                                                                             69
                                                                             -8
Texas
Denton
                                                                      481210034
                                                                             68
                                                                             72
                                                                             -4
Texas
Denton
                                                                      481211032
                                                                             66
                                                                             72
                                                                             -6
Texas
Ellis
                                                                      481390016
                                                                             60
                                                                             64
                                                                             -4
Texas
Ellis
                                                                      481391044
                                                                             57
                                                                             63
                                                                             -6
Texas
Galveston
                                                                      481671034
                                                                             67
                                                                             74
                                                                             -7
Texas
Harris
                                                                      482010024
                                                                             68
                                                                             79
                                                                            -11
Texas
Harris
                                                                      482010026
                                                                             66
                                                                             69
                                                                             -3
Texas
Harris
                                                                      482010029
                                                                             67
                                                                             73
                                                                             -6
Texas
Harris
                                                                      482010046
                                                                             66
                                                                             64
                                                                              2
Texas
Harris
                                                                      482010047
                                                                             69
                                                                             72
                                                                             -3
Texas
Harris
                                                                      482010051
                                                                             68
                                                                             70
                                                                             -2
Texas
Harris
                                                                      482010055
                                                                             68
                                                                             76
                                                                             -8
Texas
Harris
                                                                      482010062
                                                                             72
                                                                             67
                                                                              5
Texas
Harris
                                                                      482010066
                                                                             68
                                                                             69
                                                                             -1
Texas
Harris
                                                                      482010416
                                                                             72
                                                                             73
                                                                             -1
Texas
Harris
                                                                      482011015
                                                                             64
                                                                             67
                                                                             -3
Texas
Harris
                                                                      482011034
                                                                             71
                                                                             73
                                                                             -2
Texas
Harris
                                                                      482011035
                                                                             68
                                                                             70
                                                                             -2
Texas
Harris
                                                                      482011039
                                                                             74
                                                                             78
                                                                             -4
Texas
Harris
                                                                      482011050
                                                                             68
                                                                             70
                                                                             -2
Texas
Johnson
                                                                      482510003
                                                                             62
                                                                             73
                                                                            -11
Texas
Kaufman
                                                                      482570005
                                                                             58
                                                                             64
                                                                             -6
Texas
Montgomery
                                                                      483390078
                                                                             64
                                                                             74
                                                                            -10
Texas
Tarrant
                                                                      484390075
                                                                             65
                                                                             75
                                                                            -10
Texas
Tarrant
                                                                      484391002
                                                                             63
                                                                             72
                                                                             -9
Texas
Tarrant
                                                                      484392003
                                                                             66
                                                                             73
                                                                             -7
Texas
Tarrant
                                                                      484393009
                                                                             68
                                                                             76
                                                                             -8
Texas
Tarrant
                                                                      484393011
                                                                             62
                                                                             69
                                                                             -7
 
 
 
 
                                    Avg ALL
                                                                          -4.97
                                                                               

 
 
                                                 Avg if Current DV >= 74 ppb
                                                                          -7.56
      We also note that ODEQ relied on TCEQ's future year 2023 modeled DVs at three nonattainment/maintenance receptors in their SIP submission.  Oklahoma was linked to these three receptors in previous modeling results included in the EPA's March 2018 memorandum.  These monitors included 2 monitors in DFW area (Tarrant County AQS ID 484392003 and Denton County AQS ID 481210034) and one monitor in the HGB area (Brazoria County AQS ID 480391004). Included as Table 2.4 is information from the Oklahoma 2015 Ozone Transport SIP that utilized information from the EPA's March 2018 memorandum on Transport Modeling Results. 
Table 2.4 Oklahoma's Transport SIP Information Referring to Linkages to Texas Receptors. 
                                   Receptor
                                   (Site ID,
                                County, State)
                             2023 Average DV (ppb)
                             2023 Maximum DV (ppb)
                             Oklahoma Contribution
                                     (ppb)
                    Oklahoma's Step 1 and 2 Determination
                             481210034, Denton, TX
                                     69.7
                                     72.0
                                     1.23
             Maintenance receptor identified for further analysis.
                            484392003, Tarrant, TX
                                     72.5
                                     74.8
                                     1.71
            Nonattainment receptor identified for further analysis.
                            480391004, Brazoria, TX
                                     74.0
                                     74.9
                                     0.90
Nonattainment receptor with contribution less than 1 ppb; no further analysis.

      From Table 2.3 we note that the two receptors in DFW area are currently monitoring (2018-2020 DV) nonattainment with a 72 ppb at the Denton County maintenance receptor and 73 ppb at the Tarrant County nonattainment receptor. We note that the Nonattainment receptor in HGB (Brazoria County AQS ID 480391004) is currently monitoring a 73 ppb (2018-2020 DV) but was projected to be 78 ppb in TCEQ's 2023 modeled projections. We note that in the HGB area the maximum monitored values can vacillate between monitors depending upon the variations in local meteorology from one year to the next. Based on most recent DVs at these three monitors it is possible that they may reach attainment by 2023, however even if they attain, it is likely that they would be close to the NAAQS, and therefore, still be considered maintenance receptors. A drop of 3 ppb (1 ppb/year) would result in DVs of 70 ppb at two monitors and 69 ppb at the Denton monitor. As discussed above, TCEQ's 2023 model projections for these three monitors has underestimation concerns for the two DFW monitors and raises concerns with the adequacy of relying on TCEQ's modeling to conclude that the two DFW monitors are not at least maintenance receptors in 2023 for which Oklahoma is linked.

d.	Underestimation in TCEQ's 2023 Future year projections at downwind areas 
      As discussed above, EPA has concerns that TCEQ's 2012 base case modeling is underestimating ozone levels in downwind areas in Illinois, Wisconsin, Michigan, Colorado, and California.  These underpredictions in the base case also result in TCEQ's modeling underestimating 2023 Ozone DVs at downwind monitors/receptors that TCEQ may be linked to. As discussed above, absent variability in meteorology and any large local/regional/national implementation of emission reductions, the drop in monitored Ozone DVs from 2020 to 2023 would be expected to be 3-4 ppb at large nonattainment metropolitans like DFW and HGB.  While not as exact as developing new modeling of emission changes from 2020 to 2023 to project 2023 DVs, using the general 3-4 ppb provides a ballpark estimate to evaluate if TCEQ's modeling might be underpredicting 2023 future DVs at similar size nonattainment areas in the  Chicago, IL/Wisconsin/Michigan area; the Denver, CO area; Cochise County, AZ receptor; and Southern California areas. We note that in our evaluation of most recent DV trends in these areas that the annual average change from 2010 to 2020 DVs ranged from increasing to decreasing at most approximately 1 ppb/year, so use of 3-4 ppb is still an optimistic assessment of how much the DVs might drop between 2020 and 2023 but the trends data for these areas indicates the change will likely be less than a 3-4 ppb decrease for many monitors. To assess if TCEQ's modeling is potentially underestimating future year 2023 modeled DVs in these downwind areas, the EPA has compared 2018-2020 monitored DVs at many typically higher ozone monitors in these areas with TCEQ's projected 2023 DVs in Table 2.5, Table 2.6 and Table 2.7.

 Table 2.5 TCEQ Projected Future 2023 DVs Compared to 2018-2020 Monitored DVs 
                              (IL, WI, & MI).
Chicago and Downwind areas in Wisconsin and Michigan
                                       
State
County
AQS ID
                    TCEQ 2023 Future Design Value-DV (ppb)
                           EPA Monitored 2018-2020 DV
                                     (ppb)
             Delta in 3 years 2018-2020 Monitor DV - TCEQ 2023 DV
                                     (ppb)
Illinois
Cook
                                                                      170310001
                                      60
                                      75
                                      -15
Illinois
Cook
                                                                      170310032
                                      68
                                      74
                                      -6
Illinois
Cook
                                                                      170310076
                                      61
                                      69
                                      -8
Illinois
Cook
                                                                      170311003
                                      60
                                      73
                                      -13
Illinois
Cook
                                                                      170311601
                                      61
                                      71
                                      -10
Illinois
Cook
                                                                      170314002
                                      61
                                      71
                                      -10
Illinois
Cook
                                                                      170314007
                                      57
                                      71
                                      -14
Illinois
Cook
                                                                      170314201
                                      64
                                      77
                                      -13
Illinois
Cook
                                                                      170317002
                                      66
                                      75
                                      -9
Michigan
Allegan
                                                                      260050003
                                      71
                                      73
                                      -2
Wisconsin
Door
                                                                      550290004
                                      64
                                      72
                                      -8
Wisconsin
Kenosha
                                                                      550590019
                                      67
                                      74
                                      -7
Wisconsin
Manitowoc
                                                                      550710007
                                      65
                                      70
                                      -5
Wisconsin
Milwaukee
                                                                      550790010
                                      58
                                      62
                                      -4
Wisconsin
Milwaukee
                                                                      550790085
                                      66
                                      70
                                      -4
Wisconsin
Sheboygan
                                                                      551170006
                                      70
                                      75
                                      -5
                                       


                                       
                                    Avg ALL
                                     -8.31
                                       


Avg if Current DV >= 74 ppb
                                     -9.17


Table 2.6 TCEQ Projected Future 2023 DVs Compared to 2018-2020 Monitored DVs (CO).
Denver and downwind areas in Colorado
                                       
                                       
                                       
State
County
AQS ID
                    TCEQ 2023 Future Design Value-DV (ppb)
                           EPA Monitored 2018-2020 DV
                                     (ppb)
             Delta in 3 years 2018-2020 Monitor DV - TCEQ 2023 DV
                                     (ppb)
Colorado
Adams
                                                                       80013001
                                      65
                                      69
                                      -4
Colorado
Arapahoe
                                                                       80050002
                                      70
                                      77
                                      -7
Colorado
Arapahoe
                                                                       80050006
                                      66
                                      71
                                      -5
Colorado
Boulder
                                                                       80130011
                                      67
                                      74
                                      -7
Colorado
Denver
                                                                       80310002
                                      57
                                      70
                                      -13
Colorado
Douglas
                                                                       80350004
                                      73
                                      81
                                      -8
Colorado
Jefferson
                                                                       80590005
                                      68
                                      71
                                      -3
Colorado
Jefferson
                                                                       80590006
                                      72
                                      79
                                      -7
Colorado
Jefferson
                                                                       80590011
                                      71
                                      80
                                      -9
Colorado
Larimer
                                                                       80690007
                                      69
                                      70
                                      -1
Colorado
Larimer
                                                                       80690011
                                      72
                                      75
                                      -3
Colorado
Larimer
                                                                       80691004
                                      65
                                      67
                                      -2
Colorado
Weld
                                                                       81230009
                                      70
                                      70
                                       0
 
 
 
                                       
                                    Avg ALL
                                     -5.31
 
 
 
Avg if Current DV >= 74 ppb
                                     -6.83

 Table 2.7 TCEQ Projected Future 2023 DVs Compared to 2018-2020 Monitored DVs 
                                 (AZ & CA)
                     Arizona and Southern California areas
State
County
AQS ID
                    TCEQ 2023 Future Design Value-DV (ppb)
                           EPA Monitored 2018-2020 DV
              Delta in 3 years 2018-2020 Monitor  -  TCEQ 2023 DV
                                     (ppb)
Arizona
Yuma
                                                                       40278011
                                      72
                                      68
                                       4
California
Los Angeles
                                                                       60370002
                                      74
                                      97
                                      -23
California
Los Angeles
                                                                       60370016
                                      87
                                      107
                                      -20
California
Los Angeles
                                                                       60370113
                                      59
                                      70
                                      -11
California
Los Angeles
                                                                       60371103
                                      63
                                      76
                                      -13
California
Los Angeles
                                                                       60371201
                                      80
                                      92
                                      -12
California
Los Angeles
                                                                       60371302
                                      52
                                      64
                                      -12
California
Los Angeles
                                                                       60371602
                                      70
                                      78
                                      -8
California
Los Angeles
                                                                       60371701
                                      80
                                      88
                                      -8
California
Los Angeles
                                                                       60372005
                                      72
                                      93
                                      -21
California
Los Angeles
                                                                       60375005
                                      57
                                      62
                                      -5
California
Los Angeles
                                                                       60376012
                                      87
                                      101
                                      -14
California
Riverside
                                                                       60650012
                                      85
                                      99
                                      -14
California
Riverside
                                                                       60650016
                                      63
                                      78
                                      -15
California
Riverside
                                                                       60651016
                                      87
                                      99
                                      -12
California
Riverside
                                                                       60652002
                                      75
                                      84
                                      -9
California
Riverside
                                                                       60655001
                                      78
                                      88
                                      -10
California
Riverside
                                                                       60656001
                                      78
                                      94
                                      -16
California
Riverside
                                                                       60658001
                                      88
                                      96
                                      -8
California
Riverside
                                                                       60658005
                                      84
                                      98
                                      -14
California
Riverside
                                                                       60659001
                                      73
                                      87
                                      -14
California
San Bernardino
                                                                       60710001
                                      71
                                      81
                                      -10
California
San Bernardino
                                                                       60710005
                                      96
                                      90
                                       6
California
San Bernardino
                                                                       60710012
                                      83
                                      90
                                      -7
California
San Bernardino
                                                                       60710306
                                      76
                                      83
                                      -7
California
San Bernardino
                                                                       60711004
                                      91
                                      106
                                      -15
California
San Bernardino
                                                                       60712002
                                      94
                                      102
                                      -8
California
San Bernardino
                                                                       60714001
                                      82
                                      87
                                      -5
California
San Bernardino
                                                                       60714003
                                      94
                                      114
                                      -20
California
San Bernardino
                                                                       60719002
                                      79
                                      86
                                      -7
California
San Bernardino
                                                                       60719004
                                      88
                                      110
                                      -22
 
 
 
                                 Avg ALL CA 
                                    -11.80
 
 
 
Avg Current DV >= 74 ppb
                                    -11.64


      As can be seen in Table 2.5, there are many monitors in Illinois, Wisconsin, and Michigan that would need a drop in ozone DVs over the next 3 years of  >= 5 ppb (values of -5 ppb or more in the far right column) to reach TCEQ's 2023 modeled projected levels. The average decrease to meet TCEQ's modeled DV projections is 8.31 ppb, which is more than the 0-4 ppb that would be typical for these areas. EPA also separately evaluated  the TCEQ modeled 2023 DVs for monitors with 2018-2020 DVs of 74 ppb or greater.  For these monitors, a 2023 monitored DV of  <=70 ppb  would be needed to reach attainment, implying a drop of at least 3 to 4 ppb.  However even if these monitors show attainment  in 2023, it is likely that their DVs would be close enough to the NAAQS to still be considered maintenance receptors. For this subset of monitors with 2018-2020 DVs of 74 ppb or more, the average amount of decrease in DVs needed to match TCEQ's 2023 projection is 9.17 ppb. In EPA's modeling experience, drops in 8-hour ozone DVs on average of 9.17 ppb in three years rarely occur unless there is a very large change in emissions or large change in meteorological conduciveness for ozone generation. TCEQ did not identify any large emission reductions to be implemented in the 2021-2023 timeframe. Based on this information TCEQ's modeling is underestimating future ozone levels at monitors/receptors in Illinois, Wisconsin, and Michigan and these underestimations likely prevented TCEQ from identifying nonattainment/maintenance receptors. Furthermore, as discussed above, TCEQ's modeled underestimation of future year ozone levels in Texas and production of ozone from Texas emissions can cause an underestimation of Texas's impacts on downwind State's ozone nonattainment and maintenance receptors had they been identified by Texas as receptors. EPA's modeling results included in the March 2018 memorandum and EPA's 2016v2 modeling both identified nonattainment/maintenance receptors in the Chicago area of Illinois and downwind areas in Wisconsin and Michigan that also had linkages to emissions from Texas (Texas' contribution was greater than 0.7 ppb). 
      As can be seen in Table 2.6, there are many monitors in the Denver area and downwind of Denver that would need a drop in ozone DVs the next 3 years of > 5 ppb (values of -5 ppb or more in the far right column) to reach TCEQ's 2023 modeled projected levels. The average decrease to meet TCEQ's modeled DV projections is 5.31 ppb which is more than the 0-3 ppb that would be typical for this area.  EPA also evaluated just the 2023 DVs for monitors with 2018-2020 DVs of > 74 ppb.  For these monitors, a 2023 monitored DV of <= 70 ppb would be needed to monitor attainment which would imply a drop of at least 3-4 ppb.  However, even if the monitors monitor attainment in 2023, it is likely that they would be close to the NAAQS, and therefore, still be considered maintenance receptors. For this subset of monitors with 2018-2020 DVs of >74 ppb, the average amount of decrease in DVs needed to match TCEQ's 2023 projection is 6.83 ppb. In EPA's modeling experience drops in 8-hour ozone DVs on average of 6.83 ppb in three years do not typically occur unless there is a large change in emissions or large change in meteorological conduciveness for ozone generation. Based on this information, it does appear that TCEQ's modeling is underestimating future ozone levels at monitors/receptors in Denver and surrounding downwind areas of Colorado and these underestimations may have prevented TCEQ from identifying nonattainment/maintenance receptors. Furthermore, TCEQ's modeled underestimation of future year ozone levels in Texas and production of ozone from Texas emissions can also result in an underestimation of Texas's impacts on downwind ozone nonattainment and maintenance receptors in Colorado.  Even with these underestimation concerns, TCEQ identified several nonattainment/maintenance receptors in Colorado with linkages to Texas based on contributions of emissions from Texas equal to or greater than 0.7 ppb. EPA's modeling results included in the March 2018 memorandum also identified nonattainment/maintenance receptors in the greater Denver area of Colorado (Jefferson County) that also had linkages to emissions from Texas. (Texas' contribution was greater than 0.7 ppb).
      As can be seen in Table 2.7, there are many monitors in the Southern California area that would need a drop in ozone DVs the next 3 years of more than 5 ppb (values of -5 ppb or more in the far right column) to reach TCEQ's 2023 modeled projected levels. The average decrease to meet TCEQ's modeled DV projections is 11.80 ppb, which is more than the 0-3 ppb that would be typical for these areas of Southern California. EPA also evaluated a subset of these monitors, just the 2023 DVs for monitors with 2018-2020 DVs of 74 ppb or greater.  For these monitors, a 2023 monitored DV of <= 70 ppb would be needed to monitor attainment which would imply a drop of at least 3-4 ppb. However even if they monitor attainment in 2023, it is likely that they would be close to the NAAQS, and therefore, still be considered maintenance receptors. For this subset of monitors with 2018-2020 DVs of 74 ppb or more, the average amount of decrease in DVs needed to match TCEQ's 2023 projection is 11.64 ppb. In EPA's modeling experience drops in 8-hour ozone DVs on average of 11.6 ppb in three years rarely occur unless there is a large change in emissions or large change in meteorological conduciveness for ozone generation. Based on this information, TCEQ's modeling is underestimating future ozone levels at monitors/receptors in Southern California area and these underestimations may have prevented TCEQ from identifying nonattainment/maintenance receptors. Further, as discussed above, TCEQ's modeled underestimation of future year ozone levels in Texas and production of ozone from Texas emissions can also result in an underestimation of Texas's impacts on downwind ozone nonattainment and maintenance receptors in California.  Even with these underestimation concerns, TCEQ identified several nonattainment/maintenance receptors in California with linkages to Texas with contributions <= 0.7 ppb. 
e.	Summary
      TCEQ's 2012 base case modeling underestimates ozone levels in Texas at high ozone monitors in the DFW and HGB areas and also at some of the high ozone monitors in areas downwind of Texas including the Chicago, IL area and downwind receptors in Wisconsin and Michigan, the greater Denver area of Colorado, and in Southern California. TCEQ's 2023 future year projections seem to be underestimating ozone levels in Texas at high ozone monitors in DFW and HGB and also underestimating downwind ozone levels at monitors in areas downwind of Texas including the Chicago area and downwind receptors in Wisconsin and Michigan, greater Denver area of Colorado, and in Southern California, as shown above by comparing 2018-2020 monitored DVs to the projected 2023 DVs and examining the amount of ozone reductions that would be necessary to meet those projections. These underestimations likely prevented TCEQ from identifying nonattainment/maintenance receptors in the Chicago area and downwind receptors in Wisconsin and Michigan.  Furthermore, as discussed above, TCEQ's modeled underestimation of future year ozone levels in Texas and production of ozone from Texas emissions can cause an underestimation of Texas's impacts on downwind State's ozone nonattainment and maintenance receptors had they been identified by TCEQ as receptors.  Despite these underprediction/underestimation biases, TCEQ's modeling did identify a number of receptors in Colorado and California that Texas emissions were linked (contribution equal or greater than 0.70 ppb). 

3.	Comparison of TCEQ and EPA 2023 future modeling and identified maintenance and nonattainment receptors 
   For the EPA's 2016 base year modeling, the EPA undertook a large collaborative multi-year effort with states (including Texas) and other stakeholder input in developing the 2016 emission inventories including 2016v2, so that the EPA's modeling would be based on the best data available. Using a 2016 base year also provides a more recent platform that shortens the number of years to project emission changes, reducing uncertainties in the 2023 projection compared to TCEQ's projection from a 2012 base to 2023 or the EPA's earlier 2011 base year modeling. Use of a more recent 2016 base year also allows for the use of monitored DVs from a more recent period. The combination of these and other issues result in less model uncertainty compared to TCEQ's 2012 base year modeling and has provided a better estimate of 2023 ozone levels; and therefore, we believe a more reliable tool for predicting which areas of the country will be nonattainment or have difficulty maintaining the standard as well as assessing contributions from upwind states.
   The EPA's modeling using both 2011 and 2016 base year periods identified that Texas was linked to nonattainment and/or maintenance receptors in 2023 in the Midwest Region (Illinois, Wisconsin, and Michigan), while TCEQ's modeling using a 2012 base year indicated only linkages to western receptors. As discussed above, the TCEQ's modeling is underestimating projected ozone levels in the Midwest Region for 2023. If TCEQ's 2023 modeled DVs were closer to recent observed monitoring data and anticipated 2023 monitored DVs, TCEQ would likely have also identified nonattainment and/or maintenance receptors in the Midwest Region. 
      

      EPA's base case model performance is acceptable and the base case model performance analysis is included in the Air Quality Modeling TSD. As discussed in Air Quality Modeling TSD, the EPA's base case modeling model performance demonstrates the scientific credibility of EPA's 2016v2 modeling platform and provides confidence in the ability of the modeling platform to provide a reasonable projection of expected future year ozone concentrations and contributions.  In Tables 3.1 and 3.2, the EPA is evaluating TCEQ's identification of potential receptors using their 2023 model projections, including TCEQ's methodology for identifying maintenance receptors.  As discussed above, the EPA has significant concerns with the methodology TCEQ used to identify these maintenance receptors. In Tables 3.1 and 3.2, the EPA presents TCEQ's 2023 modeling projections and EPA's 2023 modeling projections (nonattainment and maintenance methodologies) and also preliminary 2019-2021 monitored DVs (subject to QA/QC by states and EPA). 
In Table 3.1 EPA shows the TCEQ's list of potential receptors identified by TCEQ that are nonattainment and/or maintenance receptors with Texas linkages (Texas contributions of >= 0.7 ppb) in their SIP including the 2023 average DV (nonattainment receptor if >70 ppb), 2023 Maintenance DV using TCEQ's maintenance methodology (uses only the 2012-2014 base DV).  For these receptors, EPA also includes the EPA's 2023 modeling projections for both average DV and maintenance DV (using EPA's methodology) and the most recent monitor data available. 

Table 3.1 Projected 2023 Nonattainment and Maintenance Receptors Identified by TCEQ Modeling Using a 2012 Base Year
                                  Receptor, 
                                 (Monitor ID,
                                County, State)
                          TCEQ 2023 Average DV (ppb)
                        TCEQ 2023 Maintenance DV (ppb)*
                           Texas Contribution (ppb)
                        EPA 2023 Avg DV/Maintenance DV
                 2018-2020 DV/Preliminary 2019-2021 DV** (ppb)
                                       


                                       


                                   80350004,
                                  Douglas, CO
                                      73
                                      72
                                     1.42
                                     71/72
                                     81/83
                                  80590006, 
                                 Jefferson, CO
                                      72
                                      73
                                     1.26
                                     72/73
                                     79/81
                                  80590011, 
                                 Jefferson, CO
                                      71
                                      71
                                     1.26
                                     73/74
                                     80/83
                                  80690011, 
                                  Larimer, CO
                                      72
                                      71
                                     1.22
                                     64/65
                                     75/77
                                  80050002, 
                                 Arapahoe, CO
                                     70***
                                      71
                                     1.15
                                     68/68
                                     77/80
                                  40038001, 
                                  Cochise, AZ
                                      71
                                    69****
                                     1.06
                                     70/72
                                     66/66
                                  60371201, 
                                Los Angeles, CA
                                      80
                                      78
                                     0.76
                                     82/85
                                     92/87
                                  60371701, 
                                Los Angeles, CA
                                      80
                                      82
                                     0.72
                                     85/88
                                     88/90
                                  60376012, 
                                Los Angeles, CA
                                      87
                                      86
                                      0.9
                                     91/93
                                    101/101
                                  60658001, 
                                 Riverside, CA
                                      88
                                      85
                                     0.73
                                     89/90
                                     96/95
                                  60658005, 
                                 Riverside, CA
                                      84
                                      83
                                     0.71
                                     87/90
                                     98/97
                                  60710001, 
                              San Bernardino, CA
                                      71
                                      72
                                     0.84
                                     74/75
                                     81/77
                                  60710306, 
                              San Bernardino, CA
                                      76
                                      77
                                     0.81
                                     76/78
                                     83/83
                                   60711004,
                               San Bernardino, CA
                                      91
                                      90
                                     0.88
                                    97/100
                                    106/103
                                  60714001, 
                              San Bernardino, CA
                                      82
                                      79
                                     0.86
                                     82/83
                                     87/87
                                  60714003, 
                              San Bernardino, CA
                                      94
                                      91
                                     0.74
                                     95/98
                                    114/114
* Uses TCEQ's Maintenance receptor methodology that EPA has concerns (See section on maintenance receptor methodology discussion). 
** 2021 monitoring data is preliminary and still has to undergo Quality Assurance/Quality Control analysis and be certified by the State of Texas, submitted to the EPA, and reviewed and concurred on by EPA.
*** From TCEQ spreadsheet of future 2023 DVs with state contributions.
**** TCEQ did not provide calculations for this receptor so EPA calculated from the Relative Response Factor in TCEQ spreadsheet of future 2023 DVs with state contributions and the monitor's 2012-2014 DV (0.983 X 71 ppb, truncation applied).

      We note that this comparison combines TCEQ's 2012 base case year modeling results with EPA's 2016 base case year modeling results. As a result, we would expect some differences due to the differing meteorology of these two base years and the observed DVs in the base years that are used in both the nonattainment and maintenance receptor identification are for different periods. However, despite these differences, we can examine the potential accuracy of future year 2023 estimates by TCEQ and EPA compared to the most recent monitoring values since the 2023 DV is only 3 years away from the 2018-2020 DV and 2 years away from the preliminary 2019-2021 DVs. Meteorology can vary, so this analysis assumes that meteorology will be similar for the 2018-2020, 2019-2021, and 2021-2023 periods since future meteorology is unknown. Since EPA's 2016v2 modeling a utilizes a different base year meteorology than TCEQ, EPA's modeling results are different and did not find Texas linked to the receptors in Colorado, Arizona and Southern California with >= 0.7 ppb.
      
Colorado
      In general EPA and TCEQ's 2023 model projections for the nonattainment receptor analysis provides generally similar results for the Colorado receptors identified by TCEQ with some variability of 1-2 ppb for some monitors and both seem to underestimate the expected DV based on recent monitoring data. It is interesting to note that TCEQ's maintenance receptor methodology resulted in a 1 ppb lower Maintenance DV than the Nonattainment DV for the Douglas County and Larimer County, Colorado receptors. EPA and TCEQ's Maintenance DVs at Colorado receptors are similar in values despite the difference in base years and methodologies with a couple of monitors varying by 3 ppb (EPA projects 3 ppb higher at one monitor and 3 ppb lower at another monitor).

Arizona
      In general EPA and TCEQ's 2023 model projections for the nonattainment receptor analysis provides generally similar results for the Arizona receptor identified by TCEQ with some variability of 1 ppb between EPA and TCEQ's results and both seem to overestimate the expected DV based on recent monitoring data. It is interesting to note that TCEQ's maintenance receptor methodology resulted in a 2 ppb lower Maintenance DV than the Nonattainment DV for the Cochise County Arizona receptor indicating that it is nonattainment receptor but the value is lower than TCEQ's maintenance receptor methodology yields. EPA and TCEQ's Maintenance DVs at the Cochise receptor vary by 3 ppb with EPA's Maintenance DV being 3 ppb higher.

Southern California
      TCEQ identified 10 receptors in Southern California. In general EPA and TCEQ's 2023 model projections for the nonattainment receptor analysis provides differing results with EPA's projected DVs being higher than TCEQ's for many of the receptors. EPA projected higher nonattainment DVs for five of the ten receptors with values ranging from 1 to 6 ppb higher than TCEQ's. Both TCEQ and EPA underestimate the 2023 DV based on recent monitoring, but EPA's projections are closer to recent monitoring values, therefore EPA's modeled projections seems more accurate than TCEQ's. It is interesting to note that TCEQ's maintenance receptor methodology resulted in a seven of the ten receptors having maintenance DVs lower than the nonattainment DVs (three were 1 ppb lower, one was 2 ppb lower and three were 3 ppb lower) All ten receptors were identified by EPA as potential receptors. 

      In Table 3.2, the EPA shows the receptors that Texas was linked as identified by EPA in EPA's 2016v2 2023 Modeling including the 2023 average DV (nonattainment receptor if  >70 ppb) and 2023 Maintenance DV using TCEQ's maintenance methodology (uses only the 2012-2014 base DV).   


Table 3.2 Projected 2023 Nonattainment/Maintenance Receptors Identified by EPA Using EPA's 2016 base year modeling
                       Receptor (Site ID, County, State)
                    Nonattainment/ Maintenance (EPA 2016v2)
                     EPA: 2023 Average DV/Maximum DV (ppb)
                  TCEQ: 2023 Average DV/ Maintenance DV (ppb)
                 2018-2020 DV/Preliminary 2019-2021 DV** (ppb)
                                     EPA: 
                                       
                           Texas Contribution (ppb)
                                    TCEQ: 
                                       
                           Texas Contribution (ppb)







                                       
                          170310001, Cook County, IL
                                  Maintenance
                                   69.6/73.4
                                     60/58
                                     75/71
                                     0.86
                                      1.6

                          170310032, Cook County, IL
                                  Maintenance
                                   69.8/72.4
                                     68/66
                                     74/75
                                     1.46
                                     1.31

                          170314201, Cook County, IL
                                  Maintenance
                                   69.9/73.4
                                     64/62
                                     77/74
                                     1.15
                                     1.25

                          170317002, Cook County, IL
                                  Maintenance
                                   70.1/73.0
                                     66/65
                                     75/73
                                     1.58
                                     1.22

                         550590019, Kenosha County, WI
                                 Nonattainment
                                   72.8/73.7
                                     67/66
                                     74/74
                                     1.72
                                     1.44

                         550590025, Kenosha County, WI
                                  Maintenance
                                   69.2/72.3
                                   No data*
                                     74/72
                                     1.81
                                   No data*

                         551010020, Racine County, WI
                                 Nonattainment
                                   71.3/73.2
                                   No data*
                                     73/73
                                     1.34
                                   No data*

*  Kenosha AQS ID 550590025 was installed and began operating May 13, 2013, so the first three year DV available is 2013-2015. Racine AQS ID 551010020 was installed in April 14, 2014 so the first three year DV available is 2015-2017. TCEQ's modeling used monitored DV data for 2010-2012, 2011-2013, and 2012-2014 to project to the future year. Since these monitors do not have valid DVs for these periods, TCEQ's modeling can't be used to project 2023 values and identify if they would be nonattainment or maintenance receptors. 
** 2021 monitoring data is preliminary and still has to undergo Quality Assurance/Quality Control analysis and be certified by the State of Texas, submitted to the EPA, and reviewed and concurred on by EPA.
      The underprediction in the Midwest is a significant concern, especially in light of the fact that TCEQ did not identify any receptors in the Chicago area and downwind of Chicago areas of Wisconsin and Michigan while the EPA did identify receptors in these areas.
      
Cook County, IL 
      In Table 3.2, EPA identified four receptors as maintenance receptors in Cook County, IL. TCEQ's 2023 nonattainment DVs are 9-15 ppb lower than recent monitoring data and TCEQ's modeling underprediction issues discussed previously are a concern. In TCEQ's modeling they project 2023 nonattainment DVs of 60 to 68 ppb, where EPA's 2023 nonattainment DVs range from 69 to 70 ppb for the four receptors. EPA's projected 2023 nonattainment DVs are much higher than TCEQ's projected 2023 nonattainment  DVs and closer to the recently monitored data. EPA identified all four receptors as maintenance receptors. TCEQ's maintenance DVs using only the 2012-2014 monitored DVs projected 2023 maintenance DVs that were 1 -2 ppb lower than the nonattainment DVs. As discussed in the maintenance receptor methodology discussion above, TCEQ's methodology does not protect areas that may struggle with maintaining the standard in the face of inter-annual variability in ozone-conducive conditions when the nonattainment receptor method would be more stringent than the maintenance receptor method. It is worth noting that if TCEQ's modeling had identified any of these four receptors as maintenance receptors, TCEQ's modeling estimates contributions from Texas sources ranging from 1.22 ppb to 1.6 ppb and therefore Texas would have been linked to these receptors for further analysis of potential emission reductions.
      
Wisconsin
      In Table 3.2, EPA identified two receptors as nonattainment receptors and one as a maintenance receptor in Wisconsin. As footnoted in Table 3.2, TCEQ modeled a 2012 base case so TCEQ's modeling can only be used for monitors that had a one or more valid DVs in the 2010-2014 period. Only the Kenosha County AQS ID 550590019 monitor had a valid DV during this time, as the other two monitors were not installed until 2013-2014. EPA's modeling uses a 2016 base case, so EPA had valid DVs for all three monitors in Wisconsin to project 2023 DVs. TCEQ's projected nonattainment DV is 5 ppb lower than EPA's 2023 nonattainment DV and 7 ppb lower than recent monitored data. EPA's 2023 nonattainment DVs are 2 ppb, 5 ppb, and 2 ppb lower than recent monitoring data so EPA's modeling does not appear to be underestimating or overestimating 2023 nonattainment method DVs. Again TCEQ's maintenance methodologies prove to be less stringent than the nonattainment methodology and resulted in the maintenance methodology DV being 1 ppb lower than the nonattainment methodology DV. It is worth noting that if TCEQ's modeling had identified the one Wisconsin receptor that they had valid base DVs as nonattainment or maintenance, TCEQ's modeling indicated that contributions from Texas sources was 1.44 ppb and therefore Texas would have been linked to these receptors for further analysis of potential emission reductions.
      Overall, TCEQ's modeling underestimation issues resulted in no receptors being identified in the Illinois, Wisconsin, Michigan areas in the TCEQ analysis. While the TCEQ modeling projects lower overall ozone levels for these areas in 2023, TCEQ's modeling does tend to corroborate the amount of projected impact that emissions from Texas may be contributing to 5 of the 7 nonattainment and maintenance receptors identified by EPA's most recent air quality modeling, EPA MP2016v2 modeling, as having a Texas contribution greater than or equal to one percent of the 2015 ozone NAAQS (0.7 ppb). EPA's modeling had less of an underprediction bias for these monitors and is more reliable for projecting Future year ozone levels and contributions than TCEQ's modeling. EPA did find that Texas is linked to nonattainment and maintenance receptors in Cook County Illinois, and Wisconsin Counties (Kenosha and Racine). 
      

4. 	Other Potential Modeling Concerns
   a.	TCEQ did not use latest EGU Emissions available at the time they proposed the SIP that incorporated reductions from CSAPR Update/ latest ERTAC projections in their modeling. It is unclear if this made any substantial changes in the modeling analysis without updating the EGU emissions and redoing the modeling. 
   b.	TCEQ used an alternate methodology for calculating contributions in determining contributions from Texas emissions at each monitor Texas identified as a receptor in Colorado, Arizona, and Southern California. TCEQ did not use the direct modeled impacts from Texas source apportionment for each day and average for the 5 to 10 days used to generate the Relative Response Factor (RRF) that is used for calculating the future year projected DVs. TCEQ calculated the average of the modeled contribution for the top 10 MDA8 modeled days. TCEQ then adjusted the modeled average impacts (for the 10 days used for the RRF) by multiplying by a ratio. Specifically, TCEQ calculated a ratio by using the 2012 Base Design Value and dividing by the average of the RRF's 5-10 days modeled MDA8 values and then adjusted the average contribution amount. This adjustment made small changes in the Texas contribution amounts, but these adjustments did not change the conclusion that Texas contributions were above 0.7 ppb at nonattainment/maintenance receptors in Colorado, Arizona, and Southern California.
   c.	In generating their future year boundary conditions TCEQ used estimated 2023 emissions for the Representative Concentration Pathway (RCP) scenario 8.5 developed by the Intergovernmental Panel on Climate Change (IPCC). This includes some estimated changes based on climate modeling. EPA typically has not seen adjustment for climate changes included in SIP modeling; Without a modeling sensitivity analysis, the actual impact can not be quantified, but is expected to be a relatively smaller amount of the total modeled concentrations  in TCEQ's transport SIP modeling, and would not be expected to significantly change the total modeled concentrations.
5.	TCEQ Other Factor Analysis (Weight of Evidence)
   TCEQ stated that the Texas contribution to a receptor should be deemed "significant" only if there is a persistent and consistent pattern of contribution on several days with elevated ozone. To try and assess persistent and consistent patterns, TCEQ did some additional analysis of several different factors such as DV trends, number of elevated ozone days, back trajectory analyses on elevated ozone days, modeled concentrations on future expected elevated ozone days, total interstate contributions at tagged monitors, consideration of conceptual model of downwind area, and responsiveness of ozone to emissions from Texas. Based on their assessment, TCEQ concluded that emissions from Texas do not contribute significantly to nonattainment or interfere with maintenance of the 2015 ozone NAAQS at any downwind monitors.
   TCEQ Summary:
   For Colorado receptors, TCEQ provided analysis of: Design value trends, number of monitored elevated ozone days, back trajectory analysis on elevated ozone days, the modeled contributions on expected elevated ozone days, total interstate contribution, and the responsiveness of ozone formation at the tagged Colorado monitors to Texas NOX emissions. TCEQ evaluated these factors in a general Weight of Evidence approach and concluded that Texas emissions do not contribute significantly to nonattainment or interfere with maintenance of the 2015 ozone NAAQS at the tagged monitors in Colorado.
   For California receptors, TCEQ provided analysis of: design value trends, back trajectory analysis on elevated ozone days, average Texas modeled contributions on projected future elevated ozone days, and collective interstate contributions to future design values, TCEQ evaluated these factors in a general Weight of Evidence approach and concluded that Texas emissions do not contribute significantly to nonattainment or interfere with maintenance of the 2015 ozone NAAQS at the tagged monitors in California.
   While these additional factors can be informative to understand some of the complexities of ozone formation in downwind areas, how ozone is changing in those areas, areas that may contribute to downwind areas, modeled contributions from upwind areas, and potential responsiveness to upwind changes in emissions, most of these do not provide the quantitative assessment that chemical transport modeling with source contribution analyses provides. Chemical transport modeling is the most sophisticated tool available to project future year ozone levels, contributions from upwind states and frequency of the impacts of upwind states. 
   EPA has reviewed these analyses that TCEQ provided in a Weight of Evidence approach for receptors in Colorado and California and found significant concerns with some of the analyses and some analyses were not conclusive or didn't provide much information that counters the results of the photochemical modeling.  EPA's assessment of the totality of TCEQ's Weight of Evidence factors for Colorado and California receptors is that they do not provide substantial evidence that refutes the photochemical modeling analysis results including the contribution analysis using EPA's contribution methodology. EPA does not concur with TCEQ's conclusion that Texas emissions do not contribute significantly to nonattainment or interfere with maintenance of the 2015 ozone NAAQS at the tagged monitors in California.  EPA has included some comments, concerns, and limitations with using these factors that Texas used to negate the findings of the chemical transport model below. 
   a.	TCEQ indicated that there are problems with DV trends
   TCEQ provided analyses of recent DV trends at the 4 Colorado monitors that they indicated as nonattainment/maintenance receptors. TCEQ summarized that 2013 was a high monitored ozone year but that the most recent 2016 DVs at most monitors show attainment and long-term ozone decreases are modest for the time evaluated (See Figure 1.8 above that is TCEQ's SIP Figure 3-40).
   TCEQ also provided analyses of recent DV trends at the tagged California monitors along with the other monitors located in the Los Angeles CSA. These trends are displayed in TCEQ's SIP Figure 3-56: Eight-Hour Ozone Design Values for Monitors in the Los Angeles Area that is included above (See Figure 1.12 above that is TCEQ's SIP Figure 3-56). TCEQ colored the tagged California monitors and used gray coloring for the other monitors in the Los Angeles CSA.. The 10 monitors tagged had eight-hour ozone design values ranging from 101 ppb to 80 ppb in 2016. TCEQ indicated that 8-hour ozone DVs values in the area have decreased overall for the past 10 years and the 10 monitors tagged for further review have observed eight-hour ozone design value decreases from 2007 through 2016 ranging from 13% at Reseda (AQS ID: 60371201) to 5% at Victorville-Park Avenue (AQS ID: 60710306).
   The most recent monitoring data and trends are useful to help identify if the monitors are still near or above the 8-hour ozone NAAQS. The historical trend data is also informative as to whether the modeled year of 2012 was a normal year or lower/higher than normal for monitored ozone levels. The Colorado data for example supports that 2012 was probably more ozone conducive than several other years, but the AQS ID 80690011 (Larimer County, Colorado) monitor 2014-2016 DV is similar to the 2012 monitor DV despite four more years of fleet turnover in the Denver area. In looking at monitoring trends and historical ozone data, 2012 did seem to favor more transport of Texas emissions to western areas. Overall, the DV trends show that ozone levels are reducing at a  slow pace.  Based on the data for the AQS ID 80690011 (Larimer County, Colorado) monitor and the ten California area monitors identified by TCEQ as receptors (See receptors in Table 3.1 above), the long-term trends do not clearly show that the receptors in California or Colorado are expected to be well below the NAAQS, and thus, not be potential nonattainment or maintenance receptors in 2023. Furthermore, the photochemical modeling includes changes in emissions throughout the modeling domain to project 2023 modeling values based on changes from the base period that is more technically sophisticated than ozone monitored trends. The modeling trends analysis for California and Colorado receptors do not provide evidence to that refutes the photochemical modeling analysis results.
   b.	TCEQ evaluated the number of elevated ozone days (over 70 ppb) annually for monitors in the Denver and Southern California areas.
    For the greater Denver area Colorado, TCEQ provided number of days over 70 ppb trends for 2007-2016 indicating decreases in days over the long-term and that 2012 was the highest year at the five monitors evaluated. However, in 2016 3 of 5 monitors had more than 10 days of monitored exceedances and one monitor had more exceedances in 2016 than it had in 2014 and 2015. TCEQ also provided a similar analysis of trends in California for 2012 thru 2016, which shows a general decrease in monitored ozone days above 70 ppb from 2012 for most of the 10 monitors tagged by TCEQ's modeling, but most monitors still have 30-95 monitored exceedance days  (MDA8 > 70 ppb) in 2016. This data supports that the number of ozone exceedance days are improving but neither the Colorado or California analysis of number of exceedance days annually provide any evidence to refute the photochemical modeling analysis results.
   
   c.	TCEQ Back Trajectory Analysis
   TCEQ provided multi-year back trajectory analyses (2012-2016) for all the monitored ozone exceedance days for the five monitors in Colorado and 10 monitors in Southern California using National Oceanic and Atmospheric Administration (NOAA) HYbrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) . TCEQ performed 72 hour back trajectories for both Colorado and California monitors. Given the large distance from Texas to Colorado and Texas to California, 120 hour back trajectories should have been considered. Most of the Colorado back trajectories only reached central or northern Texas so longer back trajectories should have been completed (See TCEQ SIP submission, Figure 3-42). TCEQ used start heights of 500 m AGL, 1000 m AGL, and 1500 m AGL. The EPA would also recommend running a 100 m AGL start height as well since both areas have complex terrain nearby or in the potential pathways that could result in different back trajectory paths due to the difference in meteorology between 100 m AGL and 500 m AGL. Another concern is that TCEQ used the 1[st] hour of the 8-hour exceedance as the start time instead of mid 8-hour or highest 1-hour number as the start time. The EPA typically uses a mid-8-hour or later start time, or runs multiple trajectories with varying start times as the transport patterns often change throughout the monitored 8-hour exceedance. Not running multiple hours for the start time, or at least the highest 1-hour or the middle of the 8-hour period when monitored 1-hour values are typically higher than the 1[st] hour of the 8-hour period, results in some uncertainty that TCEQ's trajectories fully represent transport that occurred during a monitored 8-hr ozone exceedance. 
   TCEQ also screened the back trajectories out if they touched down at some point and also screened out any trajectories that were not below the Planetary Boundary Layer (PBL) over Texas.  TCEQ also screened out back trajectories if the back trajectory start height at the point above the receptor was above the PBL. The EPA has multiple concerns with the screening out of trajectories that TCEQ performed. 
   Before we explain our specific concerns with the screening out of back trajectories, it is important to understand what HYSPLIT back trajectories represent and their limitations. HYSPLIT back trajectory analyses use archived meteorological modeling that includes actual observed data (surface, upper air, airplane data, etc.) and modeled meteorological fields to estimate the most likely route of an air parcel transported to a receptor at a specified time. The method essentially follows a parcel of air backward in hourly steps for a specified length of time. HYSPLIT estimates the central path in both the vertical and horizontal planes. The HYSPLIT central path represents the centerline with the understanding that there are areas on each side horizontally and vertically that also contribute to the end point at the monitor. The horizontal and vertical areas from the centerline grow wider the further back in time the trajectory goes. Therefore, a HYSPLIT centerline does not have to pass directly over emissions sources or emission source areas, but merely relatively near emission source areas. Nor does the HYPSLIT centerline have to be below the PBL over Texas as the vertical spread area on the centerline for distances of 300 to 600 miles or more would most likely be below the PBL even if the centerline was well above the PBL. Likewise, even if a centerline is indicated as touching down, that adds some uncertainty but does not void the back trajectory. Given the distance from Texas to Colorado receptors and Texas to California receptors, it is unclear how many back trajectories TCEQ inappropriately screened out when the centerline was close to Texas such that the horizontal spread area could encompass areas of Texas but did not directly cross any parts of Texas.  . In addition, Figure 3-42 in TCEQ's SIP submission indicates that many of the 72-hour trajectories that pass over Texas end before they have fully traversed Texas. Some of these trajectories may have been screened out because they were not below the PBL, but may have been below the mix layer in Texas and thus retained for the analysis if  longer 120 hour trajectories had been performed.  Starting the back trajectories with only the 1[st] hour of the 8-hour ozone exceedance means that the trajectory is started when the PBL is usually the lowest.  With a later start hour (middle of the 8-hour, multiple start hours, etc.) the PBL would be higher in most cases. Thus the TCEQ only using the 1[st] hour may have resulted in back trajectories not being performed or screened out when a later start time would have resulted in a back trajectory that passed TCEQ's screening based on start height and PBL. Regardless, the EPA does not agree with screening out of trajectories as TCEQ has done in their analysis. The multiple ways that TCEQ inappropriately screened out trajectories, along with the several other issues mentioned above, limits the information that can be gleaned from TCEQ's HYSPLIT and Endpoint trajectory analysis for receptors in Colorado and California.
   The EPA relies on back trajectory analysis as a corollary analysis along with observation-based meteorological wind fields at multiple heights to examine the general plausibility of the photochemical model "linkages." Since the back trajectory calculations do not account for any air pollution formation, dispersion, transformation, or removal processes as influenced by emissions, chemistry, deposition, etc., the trajectories cannot be used to develop quantitative contributions. Therefore, back trajectories cannot be used to quantitatively evaluate the magnitude of the existing photochemical contributions from upwind states to downwind receptors.
   For Colorado, TCEQ proffered that their trajectory analysis of transport from Texas to Colorado indicates that emissions from Texas are unlikely to affect ozone concentrations in the mixing layer over Colorado on elevated ozone days. TCEQ asserted that filtering the back trajectories by only looking at trajectories during elevated ozone episodes (which EPA agrees is acceptable), that start within Colorado's mixing layer, that do not hit the surface, and that have endpoints within Texas' mixing layer is an attempt to find a clear case where emissions in Texas would affect the ozone in Colorado. TCEQ's conclusion that those filters showed that 6% of elevated ozone days in Colorado had trajectories that reached the mixing layer in Texas and that 66% of days during the 2007-2016 period evaluated where trajectories reached the Texas mixing layer occurred in 2011 and 2012 is flawed by the way TCEQ performed the HYSPLIT back trajectories and by the inappropriate screening out of back trajectories. For these reasons, EPA does not concur with TCEQ's conclusion that there are many years where no trajectories reach Texas from Colorado and that in the years where TCEQ determined that no trajectories reached Texas, the tagged monitors still observed a high number of elevated ozone days and fourth-highest eight-hour ozone concentrations above 70 ppb. EPA also disagrees with TCEQ's conclusion that Texas may not be upwind in 2015 and 2016 during any elevated ozone days at any of the five sites shown in Figure 3-44 of TCEQ's SIP submission. EPA notes that endpoint analysis and percentage of endpoints generally provides limited value but due to the concerns with the overall back trajectory analysis, the EPA finds little value in the endpoint analysis that TCEQ provided.
   TCEQ did indicate that there were also many more elevated eight-hour ozone days observed in 2012 compared to other years and that this may indicate that there were some unusual meteorological patterns that occurred in 2012 that resulted in more severe ozone season. TCEQ noted that overall, trends in the fourth-highest eight-hour ozone concentrations have only slightly decreased at the Colorado receptors that TCEQ identified. EPA concludes that TCEQ's HYSPLIT back trajectory analysis for the Colorado receptors is flawed and does not provide any evidence that Texas should not be linked to these Colorado receptors, as supported by the TCEQ photochemical modeling analysis. We note that based on EPA's analysis using a 2016 base year, Texas is not linked to these receptors likely due to differences in 2016 meteorology compared with the 2012 meteorology that TCEQ used. 
   For California, TCEQ performed a similar HYSPLIT back trajectory analysis for ozone exceedance days at the 10 monitors that TCEQ identified in Southern California as Texas being potentially linked. TCEQ performed the HYPSLITs the same way that they did for the Colorado receptors including the inappropriate screening of trajectories, and other issues discussed above in review of TCEQ's back trajectory and endpoint analysis for ozone exceedances for the Colorado receptors. 
   For the reasons mentioned above in the Colorado back trajectory and endpoint analyses, the EPA concludes that TCEQ's HYSPLIT back trajectory and endpoint analyses for the California receptors is flawed and does not provide any evidence that Texas should not be linked to these California receptors. 
   
   d.	Texas contribution on high ozone days modeled greater than 70 ppb.
   TCEQ evaluated contributions from Texas on projected future year elevated ozone days in their chemical transport modeling. TCEQ evaluated the subset of 2023 days with modeled MDA8 greater than 70 ppb, and TCEQ calculated the average modeled Texas contributions for this subset of days. Table 3-14 in TCEQ's SIP submission is included below as Table 5.1. TCEQ also provided this same analysis for the receptors identified in Southern California and included the results in in its SIP submission in Table 3-17, which is included below as Table 5.2.
Table 5.1 TCEQ evaluation of Texas contribution to 2023 modeled days with MDA8 values greater than 70 ppb at receptors in Colorado.
   
TCEQ proffered in assessing this information that for the Colorado monitors, the expected average Texas contribution on projected future elevated ozone days is a small percentage of the projected average MDA8 on these days. As a result, TCEQ argued that these impacts are not significant, since the average contribution is less than one ppb for all the monitors on a relative few number of days (2-11 days), especially when considering the uncertainties associated with model predictions. 

Table 5.2 TCEQ evaluation of Texas contribution to 2023 modeled days with MDA8 values greater than 70 ppb at receptors in Southern California.
                                         
For California monitors, TCEQ indicated that the calculated average Texas contribution on projected future elevated ozone days (2023 modeled days with MDA8 greater than 70 ppb) is less than 1% of the projected average MDA8 at all of the monitors on these days. The EPA has concern that this is not tied to evaluating contribution on the days that are used in calculating and determining the 2023 DVs used in the nonattainment and maintenance tests. The EPA uses the same days for the test to also determine contributions from upwind states to have a clear relationship that the contributions are tied to days that determined the receptor is a nonattainment or maintenance receptor. TCEQ's alternate contribution analysis does not have this relationship and TCEQ's approach could result in averaging in days that were modeled above the standard but are not necessarily the days driving the future DV levels that indicate there are nonattainment and/or maintenance receptors. The EPA's approach uses the days that were highest in the base year in the calculation of the future year DV and nonattainment and/or maintenance receptor identification and this process of selection of days to be used is the same process used in Attainment Demonstration SIPs for determining whether modeling projects that an area will attain the NAAQS. TCEQ's method has the potential to decouple these days by just relying on the future year modeling values regardless if those were the days that resulted in the future year DV being projected above the NAAQS. The EPA's approach targets the days driving the ability to attain the NAAQS in the future, whereas the inclusion of all days could average in many days with lesser ozone levels that are not as consequential to the area attaining as targeting the highest base case modeled days. The EPA's methodology is designed to address CAA requirements of not interfering with other State's ability to attain and maintain attainment, not address all attainment demonstration requirements of a downwind nonattainment area in another state and uses the same 5-10 days used for determining if the area will have problems attaining or maintain attainment due to upwind contributions to local ozone levels from another State.
   For Colorado, TCEQ's modeling only has 2 days modeled above 70 ppb at the Fort Collins-West monitor in 2023 which does not seem reasonable considering the 2018-2020 DV is 77 and the preliminary 2019-2021 DV is 80 ppb, so this analysis is likely impacted by the underestimation issues. The EPA has concerns that TCEQ's modeling has underprediction both with base case and future ozone levels at DFW and HGB monitors (indicating underestimation of Texas emissions/ozone levels that are transported to downwind areas), concerns that there is also underestimation in the base case of higher monitored ozone levels at these Colorado and California monitors, and also an underestimation of the future year 2023 model projections at these Colorado and California monitors (see discussions in Section 2 of this document).
   
   EPA uses at least 5 days and up to 10 days to calculate the average contributions and EPA does use values lower than 70 ppb in order to obtain 10 days for the calculation of average contribution from upwind states because of potential uncertainties in using the modeled values as a hard cut point (in TCEQ's case >70 ppb was used regardless of how many days this included in the calculations). We note that for three of the five Colorado receptors TCEQ's alternate contribution analysis does indicate an average contribution of more than 1% (0.7 ppb) using 9, 10, or 11 days in the average contribution calculation. 
   In addition to the concerns raise above, TCEQ's method also included many more days in the California analysis than EPA's method. Adding in all days modeled above 70 ppb greatly  increases the number of days used in the contribution averages from 10 days to 50-60 days for a number of receptors. Adding in many more days that may be more locally driven and/or not have as high monitored/modeled ozone levels could result in averaging in days with lower contributions from Texas sources and that are not critical to demonstrating attainment as the days that are used in the projection of Future year DVs in the nonattainment and maintenance tests. 
   TCEQ's alternate contribution method analysis for California and Colorado receptors does not provide substantial evidence that refutes the photochemical modeling analysis results including the contribution analysis using EPA's contribution methodology.
   e	Collective Interstate Contribution to Future DV
      TCEQ provided an analysis of collective interstate contribution to the 2023 DV for the five Colorado and ten California receptors. The collective interstate contribution at tagged Colorado receptors ranges from 9.32% to 10.27% (See Table 5.3 below). The collective interstate contribution at tagged California receptors ranges from 3.2% to 4.58% (See Table 5.4 below). TCEQ argues that these are small percentages (Colorado and California) and not as high as the collective interstate contribution percentages the EPA calculated for monitors in Eastern States, which ranged from 17% to 67%. TCEQ also notes that a significant portion of the tagged Colorado monitors' 2023 modeled DVs is due to background emissions (sum of contributions from to biogenic, fires, and boundary conditions). For the California receptors TCEQ argues that these percentages are small compared to Intra-State contribution.
      As an initial matter, the EPA is not solely relying on TCEQ's findings of linkages to Colorado and California but is also relying on its own findings of linkages to areas in the Midwest Region. As such, TCEQ's analysis of collective contributions to Colorado and California does not provide justification for not addressing downwind impacts. Nonetheless, EPA has found in the past that certain California receptors are so heavily impacted by local emissions, and total upwind contribution is so low, that those receptors may not be considered to be affected by interstate ozone transport. See 81 FR 15200 (Mar. 22, 2016). However, this is a narrow circumstance that does not apply in the vast majority of cases and has never been applied outside of California. EPA has previously found, for instance, that receptors in Colorado are heavily impacted by upwind-state contribution. See 82 FR 9155 (Feb. 3, 2017); 81 FR 71991 (Oct. 19, 2016). EPA need not draw any conclusions here regarding whether the California sites TCEQ identified should or should not be considered receptors for ozone-transport purposes. EPA affirms, contrary to TCEQ's suggestion, that the Colorado receptors TCEQ analyzed are impacted by upwind state contributions. However, the EPA's finding that Texas is linked to receptors in other states is based on still other linkages found in EPA's modeling to receptors in other states, which are clearly impacted by the collective contribution of multiple upwind states, including Texas. Under CAA section 110(a)(2)(D)(i)(I), downwind states are not obligated to reduce emissions on their own to resolve nonattainment or maintenance problems. Rather, states are obligated to eliminate their own significant contribution or interference with the ability of other states to attain or maintain the NAAQS. 
   Table 5.3  -  from TCEQ's SIP Table 3-15: Collective Interstate Contribution to Future Design Value at Tagged Colorado Monitors
   
   
   Table 5.4  -  from TCEQ's SIP Table 3-18: Collective Interstate Contribution to Future Design Value at Tagged California Monitors
   
   
   f.	TCEQ also performed Direct Decoupled Method (DDM) modeling for receptors in Colorado. 
   DDM provides a first derivative of the changes in ozone (linear relationship where the DDM value is the slope of the line for changes in ozone) from changes in NOx emissions in this case. 
   
                 Figure 5.1 TCEQ DDM results for select monitors
   
   
   
   TCEQ indicated that Rocky Flats (AQS ID: 80590006) shows a small (approximately 2 ppb) responsiveness to Texas NOx emissions in mid to late July. However, when ozone at the monitor is responsive to Texas NOx emissions, elevated ozone was not modeled except for a minor response on one day, July 23.
   TCEQ summarized their DDM modeling analysis indicating that there is some minor responsiveness to Texas' NOX emissions on two high ozone days in the third week of July, but similar to other Colorado monitors studied, the degree of responsiveness is much smaller than the responsiveness to Colorado NOX and "Other NOX" emissions.
   We note that TCEQ plotted modeled ozone levels in 2023, but did not provide information or analysis indicating if the 2012 base case modeling was replicating the 2012 monitored ozone or if the base case modeling had underestimation/overestimation issues on specific days/hours that Texas modeled non-zero and larger DDM values. The DDM modeling does show some response to Texas NOx emissions but from the scale it is hard to discern the level of response.  The response to Texas NOX emissions appears to be in the 0-2 ppb range in general with some values in the 0.2-2 ppb range for modeled values over 60 ppb. Since the TCEQ modeling has underprediction and underestimation issues, a lower future year modeled threshold is appropriate to consider. The DDM modeling does show much more responsiveness to Colorado NOx group and Other NOx group (which represents many other states), but the responsiveness to Texas NOx group is not zero for some of the time including some days that modeled over 60 ppb. Overall, the DDM modeling is inconclusive and does not provide evidence that NOx reductions in Texas would not provide some benefit at receptors in Colorado.
   
   g.	TCEQ briefly discussed conceptual model for ozone in Southern California indicating the topography and climate of the area's severe air pollution problem is a consequence of the combination of emissions from the nation's second largest urban area and meteorological conditions that are adverse to the dispersion of those emissions. TCEQ asserted that the unique features of Southern California result in the Southern California basin area make it a relative isolated area. We note that TCEQ's modeling already takes into account the unique meteorological, topographical, and amount of local emissions in Southern California and in the rest of the Continental U.S. (CONUS) including Texas emissions. TCEQ's assessment of the Southern California conceptual model does not provide substantial evidence that refutes the photochemical modeling analysis results.
   Overall, these additional analyses performed by TCEQ do not provide sufficient evidence to refute the modeling results that TCEQ's modeling indicates downwind nonattainment and/or maintenance receptors in Colorado and Southern California are impacted by Texas emissions and Texas' contribution is 0.7 ppb or greater. In fact, the monitored ozone design value trends provide evidence that future year modeled ozone levels are underestimated by TCEQ's modeling and there are likely more receptors that should have been identified with additional potential linkages. Although Texas asserted that its additional air quality factor analysis is a permissible way to interpret which contributions are "significant" because that analysis examines whether there was a "persistent and consistent pattern of contribution on several days with elevated ozone" we find that such pattern is already established by a modeled linkage at Step 2.
   In addition, EPA 2016v2 modeling using 2016 base year meteorology indicates linkages from Texas to receptors in the Midwest Region but does not indicate impacts from Texas emissions on the Colorado and other western receptors identified by TCEQ. With a different base period such as TCEQ's 2012 base period meteorology and the EPA's 2016 base period meteorology, it is not uncommon that the potential downwind nonattainment or maintenance receptors could change. These differing results about receptors and linkages can be affected by the varying meteorology from year to year and the selection of different base years, but we do not think the differing results mean that the modeling or the EPA methodology for identifying receptors or linkages is inherently unreliable. Rather, these separate modeling runs indicated (1) that there were receptors that would struggle with nonattainment or maintenance in the future, and (2) that Texas was linked to some set of these receptors, even if the receptors and linkages differed from one another in their specifics (e.g., a different set of receptors were identified to have nonattainment or maintenance problems, or Texas was linked to different receptors in one modeling run versus another). We think this common result indicates that Texas's emissions were substantial enough to generate linkages at Steps 1 and 2 to some set of downwind receptors, under varying assumptions and meteorological conditions, even if the precise set of linkages changed between modeling runs.
   In sum, the EPA's more recent and robust 2016 base year modeling platform indicates that Texas is linked to several receptors in the Midwest Region as does the EPA's earlier 2011 base year modeling. TCEQ's 2012 base case modeling showed linkages to states in the west. As discussed, the EPA does not find the additional weight of evidence evaluations conducted by TCEQ provide compelling reasons to discount the impacts indicated in Colorado and California by the TCEQ modeling. In fact, we think TCEQ's modeling likely underestimates these issues. 

6.	EPA Summary
   EPA has reviewed TCEQ's modeling, alternate maintenance receptor methodology, recent monitoring data, underestimation concerns with TCEQ's modeling, and TCEQ's Weight of Evidence analyses (other factors analyses). Overall, we find the TCEQ's alternate maintenance receptor methodology to be flawed and not acceptable because it is not protective of receptors that may have difficulty maintaining the standard in the future. We did identify underprediction biases in TCEQ's modeling analysis including future year projections that are likely underestimating future year ozone DVs and likely have resulted in TCEQ's modeling analysis not identifying nonattainment or maintenance receptors in Illinois, Wisconsin, Michigan whereas the EPA's modeling using both 2011 and 2016 base years did find nonattainment and/or maintenance receptors with linkages to emissions from Texas.
   Photochemical modeling simulations for ozone interstate transport assessment is relied upon by the EPA to simulate the formation and fate of oxidant precursors, primary and secondary particulate matter concentrations, and deposition over regional and urban spatial scales. Photochemical modeling is the most sophisticated tool available to estimate future ozone levels and contributions to those modeled future ozone levels. Consideration of the different processes that affect primary and secondary pollutants at the regional scale in different locations is fundamental to understanding and assessing the effects of emissions on air quality concentrations. For the 2015 ozone NAAQS interstate transport analysis, the EPA performed nationwide, state-level ozone source apportionment modeling using CAMx to quantify the contribution of NOx and VOC emissions from all sources in each state to project 2023 ozone concentrations at ozone monitoring sites. Detailed information for the EPA's modeling may be found in the Air Quality Modeling TSD in Docket No EPA-HQ-OAR-2021-0663.
   We recognize that the results of the EPA (2011 and 2016 base year) and TCEQ (2012 base year) modeling indicated different receptors and linkages at Steps 1 and 2 of the framework. These differing results regarding receptors and linkages can be affected by the varying meteorology from year to year and base year DVs used to project future year DVs, but we do not think the differing results mean that the general modeling approach for identifying receptors or linkages is inherently unreliable.  Rather, these separate modeling runs indicated that: (1) There were receptors that would struggle with nonattainment or maintenance in the future, and (2) That Texas was linked to some set of these receptors, even if the receptors and linkages differed from one another in their specifics (e.g., a different set of receptors were identified to have nonattainment or maintenance problems, or Texas was linked to different receptors in one modeling run versus another). We think this common result indicates that Texas's emissions were substantial enough to generate linkages at Steps 1 and 2 to some downwind receptors, under varying assumptions and meteorological conditions, even if the precise set of linkages changed between modeling runs.
   We propose to find that TCEQ's maintenance methodology is not sufficiently supported or protective but we do note that the maintenance receptors in Colorado and California that were identified by TCEQ would also be identified using EPA's maintenance methodology if applied to TCEQ's modeling results using the 2012 meteorology. Thus, the EPA agrees that the receptors that TCEQ identified as maintenance receptors are maintenance receptors in TCEQ's modeling analysis using the 2012 base year.  While EPA is concerned that TCEQ's modeling has some underestimation that resulted in not identifying receptors in the Midwest, TCEQ's modeling still identified receptors and identified linkages to receptors in Colorado and Southern California that EPA considers valid based on the 2012 base year modeling performed by TCEQ. EPA's primary concern is the underestimation biases may have resulted in less receptors than if the modeling did not have the biases. TCEQ's model performance issues may have prevented TCEQ from identifying receptors in Illinois, Wisconsin and Michigan for which Texas would be linked. We acknowledge that TCEQ's modeling an alternate meteorology year (2012) that had different ozone conduciveness in separate areas of the country may also have had some impact on the 2023 projected ozone. We note that 2012 does appear to have been a higher ozone conduciveness for the Denver area than average and also the transport patterns in the 2012 ozone season were more conducive for transport from Texas to the west than in other years. Additionally, these meteorology patterns also indicate that 2012 may have had less transport from Texas to typical downwind areas in the Midwest and eastern half of the continental United States. Regardless of the meteorology and transport to downwind area differences that the different base year modeling provides, having multiple base years modeled actually provides a more robust analysis for review. These three separate modeling runs all indicated 1) that there were receptors that would struggle with nonattainment or maintenance in the future, and 2) that Texas was linked to some set of these receptors in each of the three base years modeled, even if the receptors and linkages differed from one another in their specifics (e.g., a different set of receptors were identified to have nonattainment or maintenance problems, or Texas was linked to different receptors in one modeling run versus another). We think this common result indicates that Texas' emissions were substantial enough to generate linkages at steps 1 and 2 to some set of downwind receptors, under varying assumptions and meteorological conditions, even if the precise set of linkages changed between modeling runs.  
   
