May 1, 2014


TO: DOCKET EPA-R09-OAR-2014-0185

FROM:  Carol Bohnenkamp

SUBJECT:  Review of the Modeling for the Attainment Demonstration for the Proposed Rulemaking Action on the South Coast 2012 AQMP for the One Hour Ozone Standard

 

A. CAA and Regulatory Requirements for 1-Hour Ozone Air Quality Modeling and EPA Guidance 

For any ozone nonattainment area classified as serious or above, section 182(c)(2)(A) of the CAA specifically requires the State to submit a modeled attainment demonstration based on a photochemical grid modeling evaluation or any other analytical method determined by the Administrator to be at least as effective as photochemical modeling. The following guidance documents describe the criteria that an air quality model and its application should meet to qualify for use in an ozone attainment demonstration. 

   * Appendix W to 40 CFR part 51, as interpreted in EPA guidance. See, e.g., Guideline for Regulatory Application of the Urban Airshed Model, EPA-450/4-91-013 (July 1991); 
   * "Guidance on Use of Modeled Results to Demonstrate Attainment of the Ozone NAAQS," EPA-454/B-95-007 (June 1996); 
   * "Guidance for the 1-hour Ozone Nonattainment Areas that Rely on Weight-of-Evidence for Attainment Demonstrations, Mid-Course Review Guidance"(March 28, 2002); 
   * "Guidance for Improving Weight-of-Evidence Through Identification of Additional Emission Reduction Not Modeled (November 1999);
   * "Guidance on the Use of Models and Other Analyses for Air Quality Goals in Attainment Demonstrations for Ozone, PM2.5, and Regional Haze," April 2007. 

These documents describe the components of the attainment demonstration, explain how the modeling and other analyses should be conducted, and provide overall guidance on the technical analyses for attainment demonstrations.

As with any predictive tool, inherent uncertainties are associated with photochemical grid modeling. EPA's guidance recognizes these limitations and provides recommended approaches for considering other analytical evidence to help assess whether attainment of the NAAQS is likely. This process is called a weight of evidence (WOE) analysis. EPA's modeling guidance (updated in 1996, 1999, and 2002) discusses various WOE analyses. The guidance recommends that all attainment demonstrations include supplemental analyses beyond the recommended modeling. These supplemental analyses provide additional information such as data analyses, and emissions and air quality trends, which would help strengthen the conclusion based on the photochemical grid modeling. 

B. 1-Hour Attainment Demonstration Modeling and Weight of Evidence in the South Coast 2012 AQMP

1. Modeling Approaches for the SCAQMD Attainment Demonstration

a.  Photochemical Grid Model. 

The model selected for the Final 2012 AQMP 1 - hour ozone attainment demonstration was developed using the U.S. EPA supported Community Multiscale Air Quality (CMAQ) (version 4.7) air quality modeling platform with Statewide Air Pollution Research Center-99 (SAPRC99) chemistry, and the Weather Research and Forecasting Model (WRF) (version 3.3) meteorological fields. The modeling system (including the photochemical model, meteorological inputs, and chemical mechanism) is consistent with the previous advice of outside peer reviewers. CMAQ is a state-of-the-art air quality model that can simulate ozone and PM2.5 concentrations together in a "one-atmosphere" approach for attainment demonstrations. 

Table 1: Summary of Final 2012 AQMP 1-hour Ozone Model Selection and Modeling Protocol Final 2012 AQMP 1-Hour Ozone Modeling Protocol

Ozone

Dispersion Platform:  CMAQ
Chemistry:  SAPRC99

Domain/ Coordinates

Expanded SCOS97
Meteorology, Emissions and Model application:  Lambert Conformal
Grid:  4 Km X 4 Km

Ozone: 18 layers

Emissions Inventories
   *    2008 Base year
   *    Day-Specific Emissions
   *    Shipping emissions split into 2 layers
   *    EMFAC2011
         o 3- modules
         o Modified DTIM
   *    Adjustments to fugitive PM2.5 Paved road U.S. EPA with CA modifications
   *    Day-Specific Biogenic emissions
   *    Revised Mexican emissions profile
   
Meteorology
   *    WRF initialized with NCEP data with FDDA

Air Quality Model Performance
      *    Assess model performance based on both 1-hour statistics: Normalized  gross bias
      Normalized gross error
      Peak prediction accuracy
   *    60 ppb threshold (both indices)
   *    49 Cell averaging

2008 Base Year Simulations
      June  -  August 2008
      92 days of simulations evaluated
      Peak Episode 6/18-6/21
Future Year Projections --  Deterministic Approach /Tiered RRF Approach
   *    2022


b. Episode Selection. 
The attainment demonstration modeling focuses on 92 days of ozone air quality observed during June through August of the base year 2008. Overall, the 92 day period provides a robust description of the 2008 ozone meteorological season. During this period, there were 16 days with measured daily 1-hour maximum ozone concentrations of 125 ppb or higher at the Crestline monitoring site which is the monitoring site with the maximum number of exceedances.  Seven well defined multi-day ozone episodes occurred in the Basin. When assessed for a normalized meteorological ozone episode potential using a regression based weighting covering 30-years of data (1998-2010), the June 18-22, 2008 period was ranked in the 99th percentile. This episode contained the top four daily Basin ozone maximum concentrations for 2008 and was selected as the focus of the attainment demonstration.

c. Model Performance. 
Model performance was evaluated in three zones in the South Coast Basin: the San Fernando Valley; the eastern San Gabriel, Riverside and San Bernardino Valleys; and Los Angeles and Orange County.  



Figure 1: Model Performance Zones in the South Coast Basin: 

      3). the San Fernando Valley; 
      4). the eastern San Gabriel, Riverside and San Bernardino Valleys; and 
      5). Los Angeles and Orange County.  

Normalized Gross Bias, Normalized Gross Error, and Peak Prediction Accuracy were determined for each area.  

Normalized Gross Bias <  +   15 Paired in space and time 
Normalized Gross Error <   35 Paired in space (+2 grid cells) and time 
Peak Prediction Accuracy < +   20 Unpaired in space and time

Although not a requirement for determining acceptable model performance, the performance statistics were compared to the EPA performance goals presented in guidance documents.


Table 2: June 18-22, 2008 Base Year 1-Hour Average Ozone Performance
  (Bold type indicates meeting statistical performance criteria).



 Date
                                               Zone 3
                                                  
                                    Observed
                                      (ppb)
                                   Predicted
                                      (ppb)
                                    Unpaired
                                      Peak
                                      Ratio
                                   Normalized
                                   Bias* (ppb)
                                   Normalized
                                   Error* (ppb)
 618
                                        87
                                        93
   1.07
                                        -17
                                         25
 619
                                        95
                                       109
   1.15
                                         4
                                         18
 620
                                       111
                                        99
   0.89
                                        -10
                                         19
 621
                                       122
                                       107
   0.87
                                        -19
                                         20
 622
                                       123
                                        92
   0.75
                                        -29
                                         29

 Date
                                               Zone 4
                                                  
                                    Observed
                                      (ppb)
                                   Predicted
                                      (ppb)
                                    Unpaired
                                      Peak
                                      Ratio
                                   Normalized
                                   Bias* (ppb)
                                   Normalized
                                   Error* (ppb)
 618
                                       162
                                       123
   0.76
                                        -17
                                         20
 619
                                       152
                                       136
   0.90
                                         -1
                                         18
 620
                                       176
                                       129
   0.73
                                        -12
                                         16
 621
                                       156
                                       150
   0.96
                                         -1
                                         18
 622
                                       123
                                       134
   1.09
                                         10
                                         21

 Date
                                               Zone 5
                                                  
                                    Observed
                                      (ppb)
                                   Predicted
                                      (ppb)
                                    Unpaired
                                      Peak
                                      Ratio
                                   Normalized
                                   Bias* (ppb)
                                   Normalized
                                   Error* (ppb)
 618
                                       118
                                       107
   0.91
                                         0
                                         22
 619
                                       110
                                       111
   1.01
                                         11
                                         15
 620
                                       114
                                       106
   0.93
                                         0
                                         13
 621
                                       107
                                       115
   1.07
                                         4
                                         12
 622
                                       107
                                       121
   1.13
                                         13
                                         19


 Date
                                               Zone 3
                                                  
                                    Observed
                                      (ppb)
                                   Predicted
                                      (ppb)
                                    Unpaired
                                      Peak
                                      Ratio
                                   Normalized
                                   Bias* (ppb)
                                   Normalized
                                   Error* (ppb)
 618
                                        87
                                        93
   1.07
                                        -17
                                         25
 619
                                        95
                                       109
   1.15
                                         4
                                         18
 620
                                       111
                                        99
   0.89
                                        -10
                                         19
 621
                                       122
                                       107
   0.87
                                        -19
                                         20
 622
                                       123
                                        92
   0.75
                                        -29
                                         29

 Date
                                               Zone 4
                                                  
                                    Observed
                                      (ppb)
                                   Predicted
                                      (ppb)
                                    Unpaired
                                      Peak
                                      Ratio
                                   Normalized
                                   Bias* (ppb)
                                   Normalized
                                   Error* (ppb)
 618
                                       162
                                       123
   0.76
                                        -17
                                         20
 619
                                       152
                                       136
   0.90
                                         -1
                                         18
 620
                                       176
                                       129
   0.73
                                        -12
                                         16
 621
                                       156
                                       150
   0.96
                                         -1
                                         18
 622
                                       123
                                       134
   1.09
                                         10
                                         21

 Date
                                               Zone 5
                                                  
                                    Observed
                                      (ppb)
                                   Predicted
                                      (ppb)
                                    Unpaired
                                      Peak
                                      Ratio
                                   Normalized
                                   Bias* (ppb)
                                   Normalized
                                   Error* (ppb)
 618
                                       118
                                       107
   0.91
                                         0
                                         22
 619
                                       110
                                       111
   1.01
                                         11
                                         15
 620
                                       114
                                       106
   0.93
                                         0
                                         13
 621
                                       107
                                       115
   1.07
                                         4
                                         12
 622
                                       107
                                       121
   1.13
                                         13
                                         19


*Normalized bias and normalized error calculated for hours where observations > 60 ppb















The performance goals for Normalized Gross Error and Peak Prediction Accuracy were met in the eastern San Gabriel, Riverside and San Bernardino Valleys. The statistic for bias (Normalized Gross Bias) tends to be negative, indicating that the model tends to slightly under-predict ozone. Both June 18th and June 20th failed to meet the model acceptance criteria for the unpaired peak analysis. As a result, the attainment demonstration focused on the June 19th and 21st, days with observed peak concentrations that closely matched the design values. Based on their analysis, South Coast concludes and EPA agrees that model performance is acceptable for this application.


2. Results of SCAQMD Modeling

Photochemical model simulations were conducted for the base year 2008 emissions, and future-year 2022 baseline and controlled emissions. The ozone attainment demonstration was based on the deterministic approach outlined in the "Guidance on Use of Modeled Results to Demonstrate Attainment of the Ozone NAAQS," EPA-454/B-95-007 (June 1996.) In addition, the weight of evidence analysis uses the model in a relative sense, using the relative response factor (RRF) technique described in the "Guidance on the Use of Models and Other Analyses for Air Quality Goals in Attainment Demonstrations for Ozone, PM2.5, and Regional Haze," April 2007. 

a. Modeled Attainment 

The model predicted a maximum 1-hour ozone concentration for 2022 of 125 ppb (0.125 ppm) on June 19th at Pasadena. All other predicted concentrations during the five day episode are projected to be below a concentration of 0.12 ppm. The results of the attainment demonstration for 2022 indicate that, allowing for one day per year above the standard, the federal 1-hour ozone standard would be attained by 2022 at all monitors with the controlled emissions inventory. The attainment targets (410 tpd VOC and 150 tpd NOx) are based on both short-term and long-term control measures. With controls in place, it is expected that all stations in the South Coast ozone nonattainment area will meet the federal 1 -hour ozone standard in calendar year 2022.
      
b. Weight of Evidence Analysis

Relative reduction factor

The weight of evidence analysis for the ozone attainment demonstration relies on the use of site- specific relative response factors (RRFs) being applied to the 2008 weighted design values. The RRFs are determined from the future year controlled and the 2008 base year simulations. The results of the relative response factor analysis supports the deterministic attainment demonstration and the level of emission reductions needed for attainment. 

The second element of the weight of evidence discussion utilizes the tiered RRF approach to determine station specific future year design concentrations based on base year 2008 emissions and 2022 controlled emissions.  The proposed methodology tiers the concentration threshold for accepting a simulation station day based on three criteria for evaluation: (1) the base year daily maximum concentration absolute prediction error (calculated for a station per episode day) must be 20 percent or less; (2) the observed station concentration must be within 25 percent of the design value; and (3) a minimum of four station specific days simulated must meet the error at the set concentration threshold for the RRF to be calculated. The 20 percent error criteria is the same level used in the 8-hour ozone analysis and the four day minimum was iteratively determined to provide a measure of robustness to the RRF calculation. Basically, the four day criteria represented a balance between an analysis based on a higher error criteria with potentially more days included at a higher concentration threshold vs. a limited set of better simulated station days with lower prediction error. Table VII-5-9 lists the impacted stations and the threshold concentration used for the RRF calculation. Also listed in Table VII-5-9 are the base year average percentage prediction bias and error for those days included in the future year projection. Overall, the base year tendency is towards under prediction. 


Air Quality Data Trends
      
The most recent ambient air quality data that we have reviewed indicate that the area is on track to attain the 1-hour ozone standard by 2022.  The 1-hour ozone design value has decreased from 23.4 expected exceedence days 2000-2002 to 5.5 expected exceedence days in 2010-2012. The peak 1-hour concentration has decreased from 0.169 ppb in 2002 to 0.147 ppb in 2012.  The peak levels have decreased at all monitors throughout the South Coast Basin.  



Figure 3:  South Coast Air Basin One hour Ozone Trends from 1990 to 2013  -  at Maximum Monitoring Site.
      


Figure 4:  South Coast Air Basin One hour Ozone Trends from 1990 to 2013  -  at All Monitors.



c. EPA's Evaluation and Proposed Conclusion on the Modeling Demonstration

Our evaluation of the air quality modeling analyses and supporting information provided in the South Coast 2012 AQMP addressing 1-hour ozone indicate that the South Coast area will attain the 1-hour ozone standard by its attainment year of 2022. The appropriate episode days that represent days conducive to ozone formation in the South Coast Air Basin were selected.  The modeling platform, including photochemical model, domain, number of vertical layers, and initial and boundary conditions most appropriate for the South Coast Air Basin, and other factors are appropriate.  Finally, the performance of the selected modeling platform on each of the modeled episode days meets the performance criteria specified in EPA guidance. The modeling predicts attainment at all of the twenty ozone monitoring stations.


In addition to the attainment demonstration provided in the South Coast 2012AQMP, we have considered supplemental technical information, including ambient air quality monitoring data, which was not available at the time the attainment modeling was performed by SCAQMD. The modeling shows significant reductions in ozone from the base period. 

References:

Appendix VII, Air Quality Management Plan, 1-Hour Ozone Attainment Demonstration, February 2013, South Coast Air Quality Management District. 

Guideline for Regulatory Application of the Urban Airshed Model, EPA-450/4-91-013 (July 1991).

Guidance on Use of Modeled Results to Demonstrate Attainment of the Ozone NAAQS, EPA-454/B-95-007 (June 1996).

Guidance for the 1-hour Ozone Nonattainment Areas that Rely on Weight-of-Evidence for Attainment Demonstrations, Mid-Course Review Guidance (March 28, 2002).

Guidance for Improving Weight-of-Evidence Through Identification of Additional Emission Reduction Not Modeled (November 99).


