Air Quality Modeling Technical Support Document: 2016 CAMx PM2.5 Model Evaluation to Support of EGU Benefits Assessments Office of Air Quality Planning and Standards United States Environmental Protection Agency September 2022 ------- Contents I. Introduction 3 II. Methodology 3 III. Results 6 Summary of Findings 6 Detailed Description of Model Performance Statistics and Graphics 7 Sulfate: 7 Nitrate: 13 Organic Carbon: 19 Elemental Carbon: 26 Soil 32 IV. Use of 2016fj PM Modeling as Base Year for Estimating Future EGU Benefits 38 2 ------- I. Introduction An operational model evaluation was conducted for the 2016 base year CAMx v7.10 simulation performed for the 12 km U.S. modeling domain. CAMx model configurations and inputs are described in US EPA (2022a) and in Appendix J of US EPA (2022b). This modeling is being used by EPA to support PM2.5 benefits assessments for multiple EGU rulemakings. The purpose of this evaluation is to examine the ability of the 2016 air quality modeling platform to represent the magnitude and spatial and temporal variability of measured (i.e., observed) concentration in the context of its use as the base-year from which future year EGU PM2.5 benefits can be projected. In this context, we evaluated the model's representation of 2016 spatial and temporal patterns of the following PM2.5 component species: organic carbon (OC), elemental carbon (EC), sulfate (S04), nitrate (N03) and crustal material (soil). The evaluation presented here is based on model simulations using the 2016v2 emissions platform (i.e., scenario name 2016fj) (US EPA, 2022c). II. Methodology The model evaluation for PM2.5 focuses on comparisons of daily (24-hr average) concentrations of PM2.5 component species to the corresponding observed data at CSN and IMPROVE monitoring sites in the EPA Air Quality System (AQS). The locations of the CSN and IMPROVE monitoring sites in this network are shown in Figure 1. CSN monitoring sites are more often located in urban and suburban areas while IMPROVE monitoring sites are often located in rural areas. Therefore, concentrations at CSN sites are higher, on average, than concentrations in nearby IMPROVE sites. CSN sites provide more information on the model performance in the more densely populated locations while IMPROVE sites provide more information on the model performance in pristine locations and class I areas. 3 ------- Figure 1. Location of PM monitoring sites that include speciated measurements from CSN, IMPROVE, NCORE and Other networks as of 2021. This evaluation includes statistical measures and graphical displays of model performance based upon model-predicted versus observed concentrations. The evaluation focusses on model predicted and observed PM2.5 component species concentrations that were paired in space and time. Model performance statistics were calculated for several spatial scales and temporal periods. Statistics were calculated for individual monitoring sites and in aggregate for monitoring sites within each of nine climate regions of the 12 km U.S. modeling domain. The regions include the Northeast, Ohio Valley, (Upper) Midwest, Southeast, South, Southwest, Northern Rockies, Northwest and West1,2, which are 1 The nine climate regions are defined by States where: Northeast includes CT, DE, ME, MA, MD, NH, NJ, NY, PA, Rl, and VT; Ohio Valley includes IL, IN, KY, MO, OH, TN, and WV; Upper Midwest includes IA, Ml, MN, and Wl; Southeast includes AL, FL, GA, NC, SC, and VA; South includes AR, KS, LA, MS, OK, and TX; Southwest includes AZ, CO, NM, and UT; Northern Rockies includes MT, NE, ND, SD, WY; Northwest includes ID, OR, and WA; and West includes CA and NV. 2 Note most monitoring sites in the West region are located in California (see Figure 2), therefore the statistics for the West region will be mostly representative of model performance in California ozone. 4 ------- defined based upon the states contained within the National Oceanic and Atmospheric Administration (NOAA) climate regions (Figure 2)3 as defined in Karl and Koss (1984). U.S. Climate Regions Figure 2. NOAA climate regions (source: http://www.ncdc.noaa.gov/monitoring-references/maps/us- clirnate-regions.phpffreferences) Seasonal model performance statistics were created for monitoring locations within each climate region. Seasons are defined as follows: Winter includes December, January and February; Spring includes March, April, and May; Summer includes June, July and August; Fall includes September, October and November. Statistics were created using data on all days with valid observed data during this period. The aggregate statistics by season and climate region are presented in Tables 1-10. For this evaluation we have selected the mean bias, mean error, normalized mean bias, normalized mean error and correlation to characterize model performance. These statistics are consistent with the recommendations in Simon et al. (2012) and EPA's photochemical modeling guidance (U.S. EPA, 2018). Mean bias (MB) is the average of the difference (predicted - observed) divided by the total number of replicates (n). Mean bias is given in units of jig/m3 and is defined as: MB = -£,(P — 0) , where P = predicted and O = observed concentrations Mean error (ME) calculates the absolute value of the difference (predicted - observed) divided by the total number of replicates (n). Mean error is given in units of jig/m3 and is defined as: 3 NOAA, National Centers for Environmental Information scientists have identified nine climatically consistent regions within the contiguous U.S., http://www.ncdc.noaa.gov/monitoring-references/maps/us-climate- regions.php. 5 ------- ME=±J%\P-0\ n Normalized mean bias (NMB) is the average the difference (predicted - observed) over the sum of observed values. NMB is a useful model performance indicator because it avoids over inflating the observed range of values, especially at low concentrations. Normalized mean bias is given in percentage units and is defined as: NMB = ^g=p*100 Normalized mean error (NME) is the absolute value of the difference (predicted - observed) over the sum of observed values. Normalized mean error is given in percentage units and is defined as: NME=^=^*100 Correlation is a measure of how well the model captures spatial and temporal variations in the observed concentrations as is calculated as: mpi - n x (ot - o) cor = h(Pi-n2zm-d)2 In addition to the above performance statistics presented in Tables 1-10, we prepared several graphical presentations of model performance for MDA8 ozone. These graphical presentations include: (1) maps that show the observed and modeled PM component species concentrations at individual monitoring sites; (2) maps that show PM component species mean bias at individual monitoring sites; (3) bar and whisker plots that show the distribution of the predicted and observed PM2.5 component species concentrations by month for the US as a whole. III. Results Summary of Findings The PM2.5 component species model performance statistics by season and climate region are provided in Tables 1-10. Maps and boxplot figures also provide additional information on spatial and temporal patterns of observed and modeled PM2.5 component species and associated model biases. As indicated by the information in the tables and figures, the model generally captures the observed spatial and temporal patterns of sulfate but overestimates the magnitude of concentrations at CSN and IMPROVE sites in most regions and season by 0.1-0.6 ng/m3 depending on the region season with the exception of small model underestimates noted in summer in the Southeast (IMPROVE only), South, Southwest and West (CSN only) regions. Observed nitrate concentrations are highest during winter in the Midwestern US and in the San Juaquin Valley, CA and the Salt Lake Valley, UT. The model generally captures these spatial patterns but overestimates the magnitude of wintertime nitrate in the Southeastern US by over 1 ng/m3 at some 6 ------- sites, underestimates the nitrate in the San Juaquin Valley by 0.5-1 ng/m3, and mostly misses the elevated nitrate observed near Salt Lake City. In addition, the model overestimates nitrate in the Eastern US by 0-0.6 ng/m3 in seasons when observed concentrations are low. The observations also show elevated sulfate in the Los Angeles areas which is also predicted by the model but is underestimated. The CAMx modeling generally captures the spatial and temporal patterns of organic carbon which are the result of a myriad of source category and atmosphere formation mechanisms. The model underpredicts the magnitude of wintertime episodes in California and Utah but overestimates the concentrations in Washington and the Eastern US. Similarly, the organic carbon concentrations in the Southeast and along the Atlantic coast tend to be overpredicted in spring, summer and fall while predictions of organic carbon in the Western US are mixed during these seasons. Nationally, the organic carbon NMB was 40% in winter, 34% in spring, 14% in summer and 17% in fall at CSN monitoring locations and 36% in winter, -26% in spring, -5% in summer and 9% in fall at IMPROVE monitoring locations. The highest elemental carbon concentrations are generally observed in winter and fall when mixing of local pollution is most limited. At most monitors observed elemental carbon concentrations are less than 0.5 ng/m3 but concentrations of 1-2 ng/m3 in winter and fall are observed in the San Joaquin Valley and in certain urban areas. CAMx predictions of elemental carbon concentrations generally follow the same spatial and temporal patterns as observations. Model predictions of seasonal elemental carbon concentrations fall within ±20% of observations in most regions and seasons at both CSN and IMPROVE sites. CAMx model predictions generally overpredict soil concentrations over much of the US in all seasons by ± 0.2-0.8 at CSN sites and ± 0.1-0.4 ng/m3 at IMPROVE sites for most regions and seasons except in the Southwest most likely because windblown dust emissions are not included in the simulation. Underpredictions of soil in the summer across the South, Southwest, and West range from -0.5 to -0.9 Hg/m3. Below we describe in more detail the results shown in these figures and tables for sulfate, nitrate, organic carbon, elemental carbon, and soil. Detailed Description of Model Performance Statistics and Graphics Sulfate: Spatial patterns of observed and modeled 2016 sulfate concentrations vary seasonally (Figures 3 and 4). Observed and modeled sulfate concentrations are generally higher in the US Midwest and South compared to the Western US and the Northeast. Observed seasonally averaged concentrations at monitoring sites in the Midwest and South range from 1-3 ng/m3 depending on location and season, while observed seasonally averaged concentrations in Northeast and most of the Western US are generally less than 1 ng/m3. In Southern California, summertime sulfate observations also reach levels of 2-3 ng/m3 similar to the higher observed values in the Ohio Valley region. While the modeled concentrations tend to be somewhat higher than observed values, the model depicts these same spatial and seasonal patterns. The spatial extent of the modeled elevated summertime sulfate above 1 ng/m3 in the Western US covers the entire West Coast from Washington state down to Southern California 7 ------- while the observations only register summertime concentrations above 1 ng/m3 at sites in the southern half of California. Overall, Figure 5 shows a consistent mean bias of about 0.1-0.5 ng/m3 at most sites across seasons with the exception of model underpredictions across the southern half of the US during summer. When bias is expressed as a percent of the observed concentrations, the sulfate overestimates at CSN monitors are generally less than 50% in most regions and seasons except the Northeast and Upper Midwest during fall, the Northern Rockies and Plains during fall, the southwest during spring and the Northwest during all seasons. The sulfate overestimates at IMPROVE monitors are generally less than 50% in most regions and season except the Northeast during fall, the west and southwest during winter and spring, and the Northern Rockies and Plains and the Northwest during all seasons. The overestimates on a percentage basis are especially pronounced in the Northwest, given the low observed concentrations. Figure 6 shows the magnitude of 25th to 75th percentile modeled and observed sulfate values at CSN and IMPROVE monitors by month. Observed sulfate concentrations peak in July with mean values just above 1 ng/m3 at the more urban CSN monitors and around 0.6 ng/m3 at the more rural IMPROVE monitor. Modeled sulfate concentrations also peak in July at CSN monitors although the seasonal pattern is not as pronounced in the model as in the observations. This results in a smaller overpredictions of median sulfate concentrations across sites in July (around 0.2 ng/m3) than in other months with the largest overpredictions occuring in the fall. At IMPROVE monitors the model sulfate concentrations peak in spring rather than the observed mid-summer peak leading to an overall median of bias across monitors/days of around 0.3 ng/m3 in the spring with a somewhat smaller bias in July of around 0.1 Hg/m3. Tables 1 and 2 further break down the sulfate model performance statistics by season and region. In addition to the biases already discussed, the tables provide correlation which show how well the model captures spatial and temporal variation. The correlations are generally greater than 0.5 for sulfate at CSN sites except in the Northeast and Ohio River Valley during winter, the Southeast and South during summer, the Southwest during winter, summer and fall, the Northwest during winter and summer and the West during winter spring and summer. Correlations are generally greater than 0.5 for sulfate at IMPROVE sites except in the Southeast during spring, the Northern Rockies and Plains during summer, the Southwest during winter, summer and fall, the Northwest during summer and the West during winter spring and summer. 8 ------- >3 2.8 2.6 2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Figure 3. Observed sulfate concentrations (pg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). Figure 4. Modeled sulfate concentrations (pg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). 9 ------- • IMPROVE * CSN .FROZE * CSN Figure 5. CAMx sulfate mean bias (ng/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). CSN, CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2, S04, 20150101 to 20161231, Stale=None ' CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2 2016_01 2016_03 2016_05 2016_07 2016J Months IMPROVE, CAMx_2016fLv710_CB6r5_NH3RscaleO_12US2, S04, 20150101 to 20161231, State=None 14 ' • IMPROVE - - a CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months Figure 6: Boxplots of observed and modeled sulfate concentrations (|ig/m3) by month at CSN (right) 10 ------- and IMPROVE (left) monitoring sites. Lines indicate median concentrations across monitors in each month. Boxes delineate the 25th and 75th percentile ranges. Table 1: sulfate model performance at CSN sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 747 1.04 1.27 0.23 21.74 0.54 51.70 0.21 spring 800 0.92 1.32 0.34 36.78 0.45 48.92 0.68 Northeast summer 813 1.16 1.49 0.20 17.60 0.39 33.88 0.77 fall 762 0.87 1.35 0.51 58.79 0.59 67.47 0.65 Annual 3122 1.00 1.36 0.32 31.92 0.49 49.02 0.56 winter 326 1.00 1.26 0.28 27.53 0.44 43.74 0.68 Upper Midwest spring 354 0.91 1.32 0.38 42.00 0.48 52.03 0.66 summer 314 0.99 1.32 0.33 33.15 0.46 46.36 0.81 fall 310 0.73 1.29 0.49 67.53 0.54 73.07 0.75 Annual 1304 0.91 1.30 0.37 40.58 0.48 52.29 0.73 winter 547 1.35 1.46 0.10 7.29 0.53 39.38 0.48 Ohio River Valley spring 562 1.18 1.51 0.27 22.71 0.49 41.75 0.50 summer 554 1.63 1.85 0.32 19.30 0.61 37.50 0.65 fall 541 1.24 1.64 0.40 32.36 0.57 46.00 0.65 Annual 2204 1.35 1.62 0.27 20.04 0.55 40.83 0.60 winter 513 0.92 1.32 0.42 45.66 0.53 57.17 0.59 spring 551 1.12 1.42 0.30 26.79 0.47 42.37 0.56 Southeast summer 524 1.12 1.21 0.09 8.05 0.44 39.52 0.42 fall 506 0.97 1.39 0.40 41.15 0.48 49.50 0.70 Annual 2094 1.03 1.34 0.30 29.07 0.48 46.44 0.55 winter 327 1.08 1.47 0.32 29.69 0.54 49.80 0.64 spring 351 1.45 1.46 -0.04 -2.43 0.64 44.34 0.69 South summer 336 1.55 1.27 -0.30 -19.39 0.65 42.01 0.41 fall 331 1.40 1.57 0.23 16.22 0.58 41.57 0.60 Annual 1345 1.37 1.44 0.05 3.61 0.60 44.03 0.56 winter 143 0.51 0.65 0.22 43.42 0.37 73.30 0.65 Northern spring 151 0.54 0.75 0.27 49.71 0.35 64.91 0.61 Rockies summer 153 0.54 0.66 0.16 29.65 0.28 52.44 0.72 and Plains fall 139 0.47 0.68 0.28 60.35 0.33 71.07 0.82 Annual 586 0.52 0.69 0.23 45.01 0.33 64.86 0.69 winter 247 0.57 0.58 0.05 8.88 0.45 79.32 0.29 Southwest spring 255 0.43 0.75 0.36 82.92 0.37 86.85 0.54 summer 250 0.79 0.57 -0.21 -27.23 0.35 44.24 0.24 11 ------- fall 260 0.55 0.62 0.10 18.15 0.27 48.56 0.31 Annual 1012 0.58 0.63 0.07 12.83 0.36 61.58 0.19 Northwest winter 157 0.29 0.59 0.30 104.04 0.35 122.80 0.29 spring 161 0.40 0.83 0.47 116.07 0.48 117.82 0.65 summer 166 0.54 1.09 0.60 112.00 0.62 115.42 0.47 fall 161 0.36 0.76 0.47 129.93 0.49 136.39 0.57 Annual 645 0.40 0.82 0.46 115.66 0.49 122.04 0.54 West winter 341 0.48 0.73 0.27 55.60 0.42 86.24 0.30 spring 352 0.84 1.03 0.23 27.60 0.50 60.11 0.47 summer 349 1.45 1.27 -0.11 -7.48 0.62 42.86 0.30 fall 332 0.83 0.96 0.15 18.13 0.38 46.13 0.57 Annual 1374 0.90 1.00 0.13 14.94 0.48 53.44 0.46 National winter 3348 0.92 1.19 0.24 26.42 0.49 53.20 0.51 spring 3537 0.96 1.28 0.28 29.37 0.48 49.83 0.62 summer 3459 1.20 1.34 0.12 10.22 0.49 41.15 0.59 fall 3342 0.91 1.29 0.37 40.14 0.50 54.93 0.68 Annual 13686 1.00 1.28 0.25 25.32 0.49 49.11 0.59 Table 2: sulfate model performance at IMPROVE sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 431 0.73 0.92 0.19 25.53 0.34 46.42 0.50 spring 477 0.76 1.00 0.22 28.51 0.30 39.07 0.72 Northeast summer 486 0.76 1.06 0.26 34.77 0.35 46.36 0.83 fall 456 0.62 1.01 0.36 58.49 0.41 66.44 0.73 Annual 1850 0.72 1.00 0.26 35.90 0.35 48.65 0.73 winter 200 0.76 0.94 0.12 16.33 0.29 37.74 0.74 Upper Midwest spring 208 0.76 1.02 0.17 22.00 0.31 40.87 0.60 summer 210 0.68 0.94 0.16 23.67 0.28 41.34 0.85 fall 215 0.63 0.99 0.27 42.52 0.34 53.62 0.84 Annual 833 0.71 0.97 0.18 25.68 0.31 43.12 0.77 winter 220 1.10 1.25 0.14 12.73 0.39 35.61 0.68 Ohio River Valley spring 244 1.17 1.22 0.06 5.54 0.33 28.07 0.64 summer 239 1.49 1.61 0.14 9.43 0.57 38.60 0.67 fall 227 1.31 1.50 0.20 15.64 0.39 29.87 0.81 Annual 930 1.27 1.40 0.14 10.72 0.42 33.24 0.70 winter 342 0.95 1.18 0.21 22.21 0.41 43.21 0.57 Southeast spring 379 1.24 1.27 0.06 5.00 0.41 32.72 0.40 summer 394 1.21 1.05 -0.10 -8.34 0.44 35.98 0.57 12 ------- fall 366 1.04 1.18 0.20 19.77 0.35 33.32 0.73 Annual 1481 1.12 1.17 0.09 7.93 0.40 35.87 0.54 South winter 240 0.78 1.00 0.25 32.83 0.40 51.46 0.63 spring 273 0.96 1.03 0.06 6.61 0.34 35.46 0.69 summer 252 1.44 1.05 -0.37 -25.89 0.58 40.09 0.56 fall 264 1.12 1.29 0.16 14.43 0.42 37.97 0.69 Annual 1029 1.08 1.09 0.03 2.42 0.43 40.34 0.60 Northern Rockies and Plains winter 542 0.32 0.56 0.24 74.42 0.29 90.27 0.75 spring 573 0.38 0.64 0.26 68.47 0.28 74.35 0.74 summer 603 0.36 0.54 0.18 50.72 0.25 69.38 0.42 fall 574 0.34 0.57 0.22 65.37 0.27 80.33 0.67 Annual 2292 0.35 0.58 0.22 64.19 0.27 77.90 0.68 Southwest winter 910 0.25 0.48 0.24 97.42 0.29 115.01 0.37 spring 991 0.38 0.69 0.30 78.94 0.33 85.25 0.54 summer 985 0.65 0.46 -0.19 -28.77 0.30 45.72 0.36 fall 962 0.47 0.52 0.06 12.32 0.24 52.67 0.36 Annual 3848 0.44 0.54 0.10 23.20 0.29 65.71 0.30 Northwest winter 427 0.15 0.37 0.23 154.69 0.24 164.51 0.60 spring 505 0.31 0.68 0.37 121.66 0.37 121.85 0.68 summer 519 0.34 0.82 0.48 139.43 0.49 141.27 0.43 fall 499 0.24 0.59 0.35 144.72 0.36 149.10 0.62 Annual 1950 0.27 0.62 0.36 137.21 0.37 140.12 0.62 West winter 565 0.21 0.50 0.29 138.53 0.33 156.35 0.38 spring 608 0.49 0.78 0.30 61.86 0.36 73.92 0.46 summer 603 0.71 0.83 0.11 15.30 0.37 51.55 0.29 fall 576 0.46 0.67 0.20 43.81 0.29 62.41 0.52 Annual 2352 0.47 0.70 0.23 47.80 0.34 71.39 0.47 National winter 3877 0.47 0.71 0.23 48.85 0.32 68.35 0.72 spring 4258 0.61 0.86 0.24 39.24 0.33 55.14 0.70 summer 4291 0.74 0.82 0.07 9.56 0.37 50.49 0.62 fall 4139 0.58 0.81 0.21 35.59 0.32 54.82 0.77 Annual 16565 0.60 0.80 0.18 30.66 0.34 55.98 0.69 Nitrate: Observed nitrate concentrations have distinct seasonal and regional patterns shown in Figures 7. Nitrate concentrations are low (e.g. less than 1 ng/m3) at most locations throughout most of the year. In the Eastern US, the exceptions are the Midwest during the winter when nitrate concentrations are in the range of 2-5 ng/m3 and along the mid-Atlantic coast where the range is 1.5-2.5 ng/m3. In the Western US there are several locations with elevated observed nitrate concentrations during winter with concentrations above 5 ng/m3 in Salt Lake City, UT and in the San Juaquin Valley, CA. In southern California near Los Angeles, nitrate concentrations in the range of 3-4 ng/m3 are observed year-round. The model also generally predicts low nitrate concentrations in most locations and seasons with 13 ------- localized elevated nitrate during winter in the Midwest (2-3.5 ng/m3) and along mid-Atlantic coast (1.5- 3.5 ng/m3). The model also shows moderately elevated nitrate concentrations of 1.5-2.5 ng/m3 in the Great Lakes region in the spring in in the Ohio Valley region in the fall. In the Western US model- predicted elevated winter nitrate only reached around 1 ng/m3 in Salt Lake City, UT, 3.5 ng/m3 in San Juaquin Valley, CA and 1.5-2 ng/m3 in southern California. Moderately elevated nitrate in the range of 1- 2 ng/m3 in California were modeled in spring, summer, and fall but were not as high is monitored levels in these locations. Figure 9 shows a mix of over- and under-predictions at different monitoring sites and in different seasons. Across all sites there is a modest underprediction of nitrate in the winter at CSN and IMPROVE sites (-6% and -11% respectively). This is driven by wintertime underpredictions in all regions except for the Northeast and Southeast where the model overpredicts nitrate concentrations. In the summer, nitrate is overpredicted in the Ohio Valley and Upper Midwest regions, underpredicted in the West, Southwest and along the East Coast and relatively unbiased (within ± 0.1 ng/m3) throughout most of the rest of the country leading to overall summertime normalized mean biases across all CSN sites of 10% and across all IMPROVE sites of -26%. Figure 10 shows the distribution of modeled monthly nitrate concentrations at CSN and IMPROVE monitors closely track the overall temporal patterns of the observed concentrations at both CSN and IMPROVE monitors. Observed nitrate concentrations peak in December and January with median values between 1-1.5 ng/m3 at the more urban CSN monitors and around 0.2 ng/m3 at the more rural IMPROVE monitors. The observed nitrate concentrations are lowest during summer months of June-September with median concentrations around 0.2 ng/m3 at CSN monitors and around 0.1 at IMPROVE monitors. Modeled nitrate concentrations generally follow the same seasonal pattern as observed concentrations but are slightly higher in most months at CSN sites and higher in the winter but lower in the summer at IMPROVE sites. Nitrate correlations shown in Tables 3 and 4 above 0.5 in most regions in the winter, spring, and fall seasons when nitrate concentrations are highest. The exceptions are somewhat lower correlations at CSN sites during winter and spring in the Ohio Valley and the Southeast, during winter in the Southwest, and during winter, spring and fall in the Northwest. At IMPROVE sites, the exceptions are lower correlations during winter in the Ohio Valley, during winter and spring in the Southeast, during winter and spring in the Southwest, and during winter and fall in the Northwest. During summer when observed concentrations were low, correlations are also in most regions (0.07-0.56 at CSN sites and 0.19-0.53 at IMPROVE sites). 14 ------- >5 4,e 4.6 4.4 4.2 4 3.8 3.6 3.4 3.2 3 2.8 2.6 2.4 2.2 2 1.8 1.6 1.4 1.2 0.8 0.6 0.4 0.2 0 Figure 7. Observed nitrate concentrations (|ig/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). > IMPROVr * CSM Figure 8. Modeled nitrate concentrations (pg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). 15 ------- • IMPROVE * CSK PROVE * CSN >1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0,7 -0.8 -0.9 > IMPROVE » CSN • IMPROVE » CSN Figure 9. CAMx nitrate mean bias (|jg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). CSN, CAMx_2016fi_v710_CB6r5_NH3Rscale0_12US2, N03, 20150101 to 20161231, State=None IMPROVE, CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2, N03, 20150101 to 20161231, State=None E 2 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months —• IMPROVE - * CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2 1 1 1 1 1 I 1 1 1 1 1 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months Figure 10: Boxplots of observed and modeled nitrate concentrations (jig/m3) by month at CSN (right) and IMPROVE (left) monitoring sites. Lines indicate median concentrations across monitors in each month. Boxes delineate the 25th and 75th percentile ranges. 16 ------- Table 3: nitrate model performance at CSN sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 747 1.69 1.95 0.34 19.87 0.93 55.09 0.66 spring 800 0.86 1.05 0.19 22.45 0.59 68.91 0.67 Northeast summer 813 0.32 0.40 0.05 15.88 0.20 63.34 0.50 fall 762 0.63 1.21 0.58 92.82 0.69 110.07 0.65 Annual 3122 0.86 1.14 0.29 33.14 0.60 69.21 0.70 winter 326 2.59 2.32 -0.27 -10.39 1.15 44.27 0.71 Upper Midwest spring 354 1.07 1.41 0.33 30.50 0.78 72.32 0.58 summer 314 0.32 0.56 0.21 64.75 0.36 110.43 0.37 fall 310 0.75 1.28 0.42 56.15 0.59 78.24 0.77 Annual 1304 1.19 1.40 0.17 14.40 0.72 60.49 0.73 winter 547 2.38 2.18 -0.14 -5.90 1.27 53.27 0.42 Ohio River Valley spring 562 0.88 1.10 0.27 31.20 0.68 77.84 0.37 summer 554 0.36 0.60 0.36 99.24 0.47 131.08 0.24 fall 541 0.79 1.16 0.51 63.88 0.72 90.69 0.58 Annual 2204 1.10 1.25 0.25 22.63 0.78 71.28 0.54 winter 573 0.61 1.20 0.71 117.20 0.80 131.40 0.46 spring 643 0.34 0.49 0.18 52.59 0.28 83.49 0.28 Southeast summer 610 0.20 0.24 0.05 24.52 0.12 61.35 0.26 fall 560 0.30 0.62 0.34 112.57 0.38 127.89 0.62 Annual 2386 0.36 0.63 0.31 86.63 0.39 108.52 0.57 winter 327 0.83 1.15 0.33 40.37 0.68 82.88 0.51 spring 351 0.33 0.50 0.16 50.23 0.29 87.02 0.50 South summer 336 0.25 0.28 0.03 12.72 0.19 74.06 0.17 fall 331 0.31 0.54 0.23 75.04 0.32 103.85 0.55 Annual 1345 0.43 0.62 0.19 44.39 0.37 86.12 0.59 winter 143 1.18 0.72 -0.16 -13.74 0.64 54.27 0.67 Northern spring 151 0.49 0.48 0.15 29.44 0.35 71.40 0.73 Rockies summer 153 0.16 0.23 0.08 48.39 0.14 83.58 0.52 and Plains fall 139 0.31 0.45 0.22 69.64 0.33 103.68 0.59 Annual 586 0.53 0.47 0.07 13.25 0.36 67.64 0.69 winter 247 2.54 0.80 -1.73 -68.16 1.87 73.37 0.49 spring 255 0.44 0.33 -0.09 -19.70 0.24 54.75 0.56 Southwest summer 250 0.27 0.16 -0.10 -37.61 0.17 63.06 0.07 fall 260 0.54 0.30 -0.22 -41.09 0.36 65.40 0.53 Annual 1012 0.94 0.39 -0.53 -56.16 0.65 69.22 0.58 Northwest winter 157 1.20 0.97 -0.28 -23.55 0.92 77.11 0.39 17 ------- spring 161 0.41 0.65 0.43 104.89 0.49 119.89 0.46 summer 166 0.27 0.33 0.14 50.66 0.21 78.19 0.46 fall 161 0.51 0.67 0.29 57.84 0.53 104.49 0.29 Annual 645 0.59 0.66 0.15 24.95 0.53 90.52 0.31 winter 341 3.28 1.80 -1.36 -41.45 1.96 59.90 0.60 spring 352 1.57 1.00 -0.43 -27.08 0.83 52.69 0.64 West summer 349 1.25 0.56 -0.64 -51.37 0.81 64.64 0.56 fall 332 1.96 1.01 -0.83 -42.45 1.24 63.15 0.65 Annual 1374 2.01 1.08 -0.81 -40.38 1.20 59.98 0.64 winter 3408 1.80 1.64 -0.10 -5.79 1.12 62.16 0.50 spring 3629 0.74 0.86 0.14 19.07 0.52 70.36 0.58 National summer 3545 0.38 0.40 0.04 9.70 0.30 78.75 0.38 fall 3396 0.69 0.91 0.25 36.85 0.60 88.18 0.52 Annual 13978 0.89 0.95 0.08 9.18 0.63 70.56 0.57 Table 4: nitrate model performance at IMPROVE sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 431 0.52 0.77 0.19 36.27 0.39 73.49 0.62 spring 477 0.32 0.41 0.06 18.88 0.21 64.34 0.67 Northeast summer 486 0.15 0.16 0.00 0.31 0.10 63.93 0.50 fall 456 0.25 0.47 0.18 71.22 0.27 109.48 0.55 Annual 1850 0.31 0.45 0.10 33.74 0.24 76.88 0.65 winter 200 1.43 1.28 -0.36 -25.22 0.69 48.36 0.71 Upper Midwest spring 208 0.58 0.74 0.04 6.90 0.40 70.04 0.57 summer 210 0.12 0.32 0.19 159.89 0.21 176.11 0.53 fall 215 0.38 0.70 0.22 57.34 0.41 108.80 0.53 Annual 833 0.62 0.76 0.03 4.40 0.43 69.23 0.66 winter 220 1.34 1.14 -0.21 -16.04 0.84 62.82 0.45 Ohio River Valley spring 244 0.52 0.54 0.03 5.40 0.35 66.31 0.53 summer 239 0.19 0.30 0.12 60.95 0.19 97.26 0.39 fall 227 0.49 0.53 0.06 11.18 0.35 70.89 0.53 Annual 930 0.62 0.63 0.00 0.06 0.42 67.89 0.58 winter 342 0.49 0.66 0.13 26.76 0.35 70.85 0.49 spring 379 0.34 0.34 0.01 2.98 0.21 62.32 0.36 Southeast summer 394 0.19 0.17 -0.01 -5.63 0.13 69.29 0.19 fall 366 0.29 0.35 0.07 23.86 0.22 76.96 0.51 Annual 1481 0.32 0.38 0.05 14.75 0.22 69.69 0.52 South winter 240 0.89 0.81 -0.03 -3.56 0.60 66.85 0.50 spring 273 0.34 0.35 0.01 3.57 0.21 61.88 0.53 18 ------- summer 252 0.22 0.15 -0.06 -29.56 0.15 68.67 0.14 fall 264 0.25 0.31 0.06 23.30 0.18 69.61 0.59 Annual 1029 0.42 0.40 0.00 -1.08 0.28 66.44 0.60 Northern Rockies and Plains winter 542 0.39 0.26 -0.14 -36.75 0.27 69.28 0.62 spring 573 0.16 0.22 0.05 33.53 0.13 81.15 0.56 summer 603 0.08 0.08 0.01 7.11 0.04 57.58 0.29 fall 574 0.11 0.16 0.05 46.09 0.10 94.55 0.56 Annual 2292 0.18 0.18 -0.01 -3.60 0.13 74.39 0.58 Southwest winter 910 0.27 0.18 -0.09 -34.26 0.19 70.66 0.48 spring 991 0.18 0.17 0.00 0.11 0.09 51.23 0.39 summer 985 0.15 0.05 -0.10 -65.15 0.10 67.78 0.32 fall 962 0.12 0.08 -0.05 -38.39 0.07 56.41 0.53 Annual 3848 0.18 0.12 -0.06 -33.13 0.11 62.71 0.48 Northwest winter 427 0.32 0.24 -0.07 -23.17 0.31 97.40 0.37 spring 505 0.15 0.26 0.11 73.67 0.15 99.05 0.54 summer 519 0.14 0.10 -0.03 -24.31 0.09 69.05 0.47 fall 499 0.16 0.21 0.04 27.18 0.16 100.69 0.41 Annual 1950 0.19 0.20 0.01 7.95 0.17 92.95 0.36 West winter 565 0.47 0.41 -0.04 -8.10 0.31 65.79 0.78 spring 608 0.38 0.41 0.03 8.98 0.23 60.33 0.77 summer 603 0.32 0.11 -0.21 -64.87 0.24 72.89 0.36 fall 576 0.41 0.26 -0.15 -36.03 0.26 63.70 0.84 Annual 2352 0.39 0.30 -0.09 -22.80 0.26 65.37 0.76 National winter 3877 0.53 0.50 -0.06 -10.56 0.36 67.36 0.65 spring 4258 0.28 0.33 0.04 13.33 0.18 65.54 0.63 summer 4291 0.17 0.13 -0.04 -25.85 0.13 73.96 0.24 fall 4139 0.24 0.27 0.02 10.22 0.19 79.50 0.64 Annual 16565 0.30 0.31 -0.01 -2.98 0.21 70.29 0.65 Organic Carbon: Observed organic carbon concentrations are shown in Figure 11. The spatial and temporal patterns of organic carbon reflect its varied sources and formation mechanisms including primary emissions from wildfires in the summer and woodsmoke in the winter along with secondary formation from biogenic precursors which are prevalent in the Southeastern US and from anthropogenic precursors such as vehicles and cooking emissions in urban areas. Organic carbon is highest in California, in the Southeastern US and along the mid-Atlantic coast. In addition, there are a few organic carbon hotspots in western mountain valleys in Oregon, Washington, Idaho, Utah, and Montana during winter due to woodsmoke emissions and in Idaho and Montana during summer due to wildfires. Elevated organic carbon in the southeastern US is present year-round but peaks during the fall with concentrations reaching above 5 ng/m3 at some monitoring locations. The spatial and seasonal patterns of organic carbon predicted by CAMx (Figure 12) are similar to observed patterns although the model underpredicts the wintertime concentrations in California, and Utah but overestimates the 19 ------- concentrations in Washington and the Eastern US. The organic carbon concentrations in the Southeast and along the Atlantic coast tend to be overpredicted in spring, summer, and fall while in the Western US there is no consistency in terms of model performance with a mix of underprediction and overprediction at various monitoring sites during these seasons. As shown in Tables 5 and 6, CAMx organic carbon estimates were within ±30% of monitored values for the majority of region/season combinations. CAMx OC concentrations were more often overpredicted than underpredicted. Overpredictions were most notable in the Northwest (fall, spring, and summer) and in the eastern US during winter. Underpredictions occurred more frequently in the Western half of the US and in the Southeast during fall. Nationally, the organic carbon NMB was 40% in winter, 34% in spring, 14% in summer and 17% in fall at CSN monitoring locations and 36% in winter, -26% in spring, -5% in summer and 9% in fall at IMPROVE monitoring locations. Monthly 25th-75th percentile concentrations of observed and modeled organic carbon at monitor locations are shown in Figure 14. At the more urban CSN monitors, observed and modeled concentrations are highest during winter months when colder temperatures lead to less dispersion of local pollution, with a peak in November in both the model and observations. Conversely, organic carbon concentrations peak during summer at the more rural IMRPOVE monitor locations due to secondary formation in the atmosphere and as seen in both the monitor values and the model predictions. Correlations between CAMx modeled OC and observed OC (Tables 5 and 6) were higher at CSN monitoring sites than at IMPROVE monitoring sites. Correlation at CSN monitoring sites was above 0.5 in all regions and seasons except in the Upper Midwest in spring/summer, the Ohio Valley in summer, the Northern Rockies and Plains in winter/summer/fall, the Northwest in winter/spring/fall and the Southwest in all seasons. Correlations at IMPROVE sites tended to be somewhat lower with the best performance in the Northeast, South and West regions. 20 ------- >5 4.8 4.6 4.4 4.2 3.8 3.6 3.4 3.2 3 2.8 2.6 2.4 22 2 1.8 1.6 1.4 1.2 0.8 0.6 0.4 0.2 Figure 11. Observed organic carbon concentrations (pg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). Figure 12. Modeled organic carbon concentrations (pg/m3) at CSN (triangles) and IMPROVE (circles) 21 ------- monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). PC MB [MQnS) lor mn CAM« 201611 V7I0 CB»S NH3Ric«I»0 l2US21wO»cw*f toFitonaryi • IMPROVE » CSN • IMPROVE * CSN Figure 13. CAMx organic carbon mean bias (pg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). 22 ------- CSN, CAMx_2016fj_v710_CB6r5_NH3RscaleO_12US2, OC, 20150101 to 20161231, State=None IMPROVE, CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2, OC, 20150101 to 20161231, State=None 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months Months Figure 14: Boxplots of observed and modeled organic carbon concentrations (ng/m3) by month at CSN (right) and IMPROVE (left) monitoring sites. Lines indicate median concentrations across monitors in each month. Boxes delineate the 25th and 75th percentile ranges. Table 5: OC model performance at CSN sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 751 1.79 2.92 1.33 74.01 1.54 85.73 0.65 spring 815 1.57 2.33 0.79 50.01 0.96 61.36 0.61 Northeast summer 819 1.95 2.33 0.36 18.47 0.71 36.31 0.58 fall 805 1.85 2.64 0.85 46.10 1.09 58.61 0.65 Annual 3190 1.79 2.55 0.82 45.83 1.06 59.38 0.62 winter 334 1.13 2.54 1.38 122.01 1.41 124.92 0.54 Upper Midwest spring 347 1.47 1.90 0.53 36.00 0.96 65.25 0.41 summer 332 1.61 1.74 0.19 11.97 0.58 35.94 0.48 fall 338 1.50 2.03 0.54 35.61 0.77 51.44 0.68 Annual 1351 1.43 2.05 0.66 46.06 0.93 65.17 0.43 Ohio River Valley winter 535 1.62 2.51 0.87 53.61 1.09 67.46 0.56 spring 571 1.57 2.12 0.40 25.33 0.74 47.21 0.60 summer 532 1.85 2.08 0.17 9.27 0.58 31.39 0.47 23 ------- fall 535 2.44 2.62 0.09 3.67 0.85 34.75 0.75 Annual 2173 1.86 2.33 0.38 20.51 0.81 43.69 0.64 Southeast winter 436 2.00 2.57 0.72 36.12 1.05 52.32 0.66 spring 478 2.01 2.34 0.51 25.18 0.78 38.81 0.75 summer 445 1.90 2.50 0.64 33.73 0.84 44.01 0.71 fall 430 2.85 2.80 -0.14 -4.94 1.02 35.86 0.67 Annual 1789 2.18 2.55 0.44 20.03 0.92 42.04 0.60 South winter 272 1.98 2.35 0.47 23.74 1.16 58.52 0.59 spring 297 1.45 1.86 0.35 23.77 0.74 50.58 0.60 summer 251 1.50 1.99 0.41 26.97 0.89 58.99 0.58 fall 238 2.11 2.50 0.37 17.58 0.99 47.14 0.62 Annual 1058 1.75 2.17 0.40 22.74 0.94 53.67 0.60 Northern Rockies and Plains winter 141 0.95 0.85 -0.04 -4.25 0.82 86.15 0.12 spring 145 0.87 0.81 -0.07 -7.56 0.43 49.80 0.55 summer 161 1.45 1.13 -0.52 -35.98 0.69 47.26 0.41 fall 146 1.01 0.95 -0.27 -26.44 0.49 47.90 0.25 Annual 593 1.08 0.94 -0.23 -21.56 0.61 56.05 0.27 Southwest winter 228 2.53 2.30 0.06 2.33 1.32 52.22 0.35 spring 254 1.06 1.13 0.28 26.84 0.54 51.30 0.42 summer 237 1.41 1.15 -0.13 -9.26 0.50 35.79 0.45 fall 240 1.64 1.47 0.08 4.93 0.76 46.70 0.45 Annual 959 1.64 1.50 0.08 4.71 0.77 47.20 0.46 Northwest winter 140 2.46 3.82 1.29 52.31 2.19 88.67 0.39 spring 150 1.41 2.38 1.41 100.38 1.50 106.57 0.46 summer 158 1.49 2.42 1.39 93.33 1.46 97.75 0.66 fall 155 1.95 3.04 1.53 78.20 1.87 95.87 0.47 Annual 603 1.82 2.92 1.41 77.59 1.75 96.07 0.46 West winter 286 3.66 3.12 -0.35 -9.48 1.63 44.39 0.51 spring 294 1.54 1.75 0.22 14.03 0.60 38.82 0.61 summer 290 2.47 2.19 -0.37 -15.17 0.89 36.15 0.52 fall 277 2.82 2.48 -0.06 -2.00 1.07 37.94 0.57 Annual 1147 2.61 2.38 -0.14 -5.33 1.04 39.90 0.59 National winter 3123 1.96 2.61 0.79 40.17 1.34 68.29 0.52 spring 3351 1.53 2.01 0.52 33.74 0.82 53.22 0.58 summer 3225 1.82 2.08 0.26 14.25 0.74 40.89 0.54 fall 3164 2.10 2.42 0.36 17.21 0.98 46.67 0.59 Annual 12863 1.85 2.27 0.48 25.96 0.96 52.22 0.55 24 ------- Table 6: OC model performance at IMPROVE sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 429 429 0.75 1.50 0.85 113.11 0.87 115.77 spring 478 478 0.75 1.18 0.45 60.07 0.52 69.64 Northeast summer 482 482 1.20 1.41 0.22 18.33 0.45 37.23 fall 459 459 0.92 1.41 0.51 56.03 0.63 68.56 Annual 1848 1848 0.91 1.37 0.50 54.90 0.61 67.09 winter 228 228 0.60 1.14 0.59 99.46 0.62 103.53 Upper Midwest spring 239 239 0.90 1.14 0.24 26.56 0.63 69.87 summer 237 237 1.18 1.08 -0.10 -8.18 0.39 32.84 fall 245 245 0.89 1.03 0.14 15.72 0.36 40.30 Annual 949 949 0.90 1.10 0.21 23.99 0.50 55.47 winter 217 217 1.00 1.79 0.92 92.64 1.10 109.94 Ohio River Valley spring 242 242 1.11 1.79 0.71 63.42 0.93 83.67 summer 242 242 1.34 1.61 0.27 20.45 0.49 36.79 fall 232 232 1.80 2.04 0.19 10.41 0.81 44.92 Annual 933 933 1.32 1.81 0.52 39.18 0.83 62.74 winter 398 398 1.18 1.58 0.51 42.82 0.89 74.95 spring 447 447 6.23 1.82 -4.38 -70.38 5.52 88.65 Southeast summer 455 455 1.49 1.55 0.14 9.18 0.71 47.72 fall 423 423 1.95 1.80 -0.08 -4.35 0.83 42.63 Annual 1723 1723 2.76 1.69 -1.01 -36.38 2.03 73.47 winter 240 240 0.86 1.22 0.44 51.32 0.63 73.14 spring 273 273 1.06 1.29 0.23 21.40 0.70 65.77 South summer 250 250 1.16 1.09 -0.02 -1.43 0.57 49.21 fall 264 264 1.17 1.18 0.00 0.24 0.50 42.52 Annual 1027 1027 1.07 1.19 0.16 15.01 0.60 56.21 winter 565 565 0.30 0.34 0.03 11.16 0.20 66.79 Northern spring 603 603 0.61 0.54 -0.12 -19.21 0.37 60.44 Rockies summer 631 631 1.22 1.04 -0.15 -12.03 0.71 58.68 and Plains fall 602 602 0.62 0.48 -0.13 -21.54 0.35 56.40 Annual 2401 2401 0.70 0.60 -0.09 -13.40 0.41 59.37 winter 910 910 0.65 0.45 -0.17 -26.93 0.37 57.14 spring 994 994 0.44 0.46 0.02 5.28 0.23 52.83 Southwest summer 979 979 0.87 0.64 -0.22 -25.81 0.48 54.60 fall 964 964 0.63 0.54 -0.08 -11.98 0.34 54.98 Annual 3847 3847 0.64 0.52 -0.11 -17.25 0.35 54.98 winter 447 447 0.35 0.59 0.24 67.88 0.41 117.31 Northwest spring 513 513 0.52 0.75 0.22 42.60 0.38 71.62 summer 519 519 1.26 1.42 0.17 13.16 0.90 70.95 25 ------- fall 500 500 0.74 1.32 0.58 77.61 0.85 114.31 Annual 1979 1979 0.74 1.02 0.30 41.02 0.64 87.18 West winter 562 562 0.61 0.52 -0.07 -11.27 0.33 55.14 spring 605 605 0.61 0.59 -0.02 -3.13 0.27 44.44 summer 611 611 1.71 1.29 -0.43 -24.90 0.92 53.81 fall 576 576 1.07 1.01 -0.08 -7.78 0.49 45.58 Annual 2354 2354 1.01 0.85 -0.15 -15.10 0.51 50.41 National winter 3996 3996 0.65 0.86 0.23 35.94 0.52 79.68 spring 4394 4394 1.22 0.91 -0.32 -26.11 0.93 76.39 summer 4406 4406 1.24 1.16 -0.06 -5.09 0.64 51.84 fall 4265 4265 0.98 1.07 0.09 8.83 0.54 55.25 Annual 17061 17061 1.03 1.00 -0.02 -2.12 0.66 64.25 Elemental Carbon: Spatial and temporal patterns of observed elemental carbon concentrations are more heterogenous than sulfate, nitrate, or organic carbon with localized hotspots rather than regional patterns. As shown in Figure 15, the highest elemental carbon concentrations are generally observed in winter and fall when mixing of local pollution is minimized. At most monitors elemental carbon concentrations are less than 0.5 ng/m3 but concentrations of 1-2 ng/m3 in winter and fall are observed in the San Joaquin Valley and in certain urban areas such as Los Angeles, Atlanta, Denver, Pittsburgh and along the Northeast corridor from Philadelphia to New York City. CAMx predictions of elemental carbon concentrations shown in Figure 16 generally follow the same spatial and seasonal patterns as the corresponding observations. Model over and under predictions of seasonal elemental carbon concentrations shown in Figure 17 are with ± 0.2 ng/m3 at most monitoring sites with a few isolated locations with larger biases. As shown in Tables 7 and 8, those elemental carbon biases correspond to normalize mean bias values within ±20% of observations in most regions and seasons at both CSN and IMPROVE sites. The higher fall/winter elemental carbon concentrations in both the model and observations are also depicted in Figure 18 which shows monthly distributions. The highest observed and modeled concentrations both occur in November. As shown in Tables 7, correlation between the model and the observation at CSN sites were generally between 0.45-0.74 in most seasons/regions with the exception of some lower correlations in the the Northern Rockies (all seasons), the Southwest (summer), and the Northwest (winter/summer/fall). Correlations were somewhat higher at IMPROVE sites (Table 8), generally between 0.5 and 0.86 in most regions and seasons except for some lower values in the Ohio River Valley (winter/spring), the Southeast (spring), The Northern Rockies and Plains (all seasons), the Northwest (summer), and the West (summer). 26 ------- >2 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Figure 15. Observed elemental carbon concentrations (ng/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). Figure 16. Modeled elemental carbon concentrations (ng/m3) at CSN (triangles) and IMPROVE (circles) 27 ------- monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). Figure 17. CAMx elemental carbon mean bias (|ig/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). 28 ------- CSN, CAMx_2O16fj_v71O_CB6r5_NH3Rscale0_12US2, EC, 20150101 to 20161231, State=None IMPROVE, CAMx_2016fLv710_CB6r5_NH3Rscale0_12US2, EC, 20150101 to 20161231, State=None • IMPROVE ' CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months Figure 18: Boxplots of observed and modeled elemental carbon concentrations (ng/m3) by month at CSN (right) and IMPROVE (left) monitoring sites. Lines indicate median concentrations across monitors in each month. Boxes delineate the 25th and 75th percentile ranges. Table 7: EC model performance at CSN sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 751 0.67 0.74 0.15 23.00 0.37 55.48 0.58 spring 815 0.58 0.61 0.06 9.92 0.28 48.75 0.53 Northeast summer 819 0.58 0.59 0.04 6.91 0.25 42.99 0.56 fall 805 0.63 0.75 0.17 26.77 0.35 55.76 0.54 Annual 3190 0.61 0.67 0.10 16.92 0.31 50.91 0.55 winter 334 0.33 0.51 0.20 60.33 0.25 76.58 0.54 Upper Midwest spring 347 0.43 0.42 0.01 3.16 0.20 47.12 0.54 summer 332 0.40 0.40 0.02 4.39 0.18 44.27 0.48 fall 338 0.45 0.50 0.08 17.01 0.23 50.76 0.65 Annual 1351 0.40 0.46 0.08 18.89 0.22 53.40 0.52 Ohio River Valley winter 535 0.48 0.57 0.11 22.04 0.24 50.65 0.59 spring 571 0.53 0.50 -0.03 -4.94 0.21 40.56 0.55 summer 532 0.58 0.53 -0.04 -6.50 0.22 38.49 0.45 29 ------- fall 535 0.66 0.63 0.00 0.60 0.25 37.85 0.60 Annual 2173 0.56 0.56 0.01 1.94 0.23 41.37 0.56 Southeast winter 436 0.57 0.53 0.00 0.81 0.26 44.76 0.56 spring 478 0.54 0.43 -0.08 -14.74 0.23 42.29 0.56 summer 445 0.44 0.42 0.02 3.43 0.22 49.37 0.49 fall 430 0.66 0.53 -0.11 -16.42 0.29 43.36 0.66 Annual 1789 0.55 0.47 -0.04 -7.71 0.25 44.62 0.58 South winter 272 0.57 0.56 0.00 -0.85 0.25 43.92 0.61 spring 297 0.43 0.43 -0.02 -3.63 0.18 41.06 0.56 summer 251 0.37 0.45 0.05 13.78 0.22 58.42 0.48 fall 238 0.54 0.57 0.02 3.65 0.25 46.72 0.56 Annual 1058 0.48 0.50 0.01 2.31 0.22 46.60 0.57 Northern Rockies and Plains winter 141 0.25 0.20 -0.03 -11.10 0.22 88.76 0.09 spring 145 0.20 0.16 -0.02 -11.35 0.11 54.45 0.44 summer 161 0.22 0.20 -0.02 -7.81 0.10 45.54 0.39 fall 146 0.24 0.21 -0.04 -15.94 0.16 67.03 0.15 Annual 593 0.23 0.19 -0.03 -11.55 0.15 64.21 0.20 Southwest winter 228 0.88 0.73 -0.07 -7.69 0.34 38.56 0.53 spring 254 0.31 0.38 0.14 46.02 0.19 60.12 0.69 summer 237 0.30 0.35 0.11 35.61 0.17 55.46 0.41 fall 240 0.56 0.52 0.06 10.79 0.25 45.14 0.58 Annual 959 0.51 0.49 0.06 12.63 0.23 46.42 0.66 Northwest winter 140 0.75 0.95 0.26 35.43 0.61 81.29 0.34 spring 150 0.46 0.70 0.43 94.57 0.51 111.90 0.57 summer 158 0.40 0.68 0.51 125.70 0.53 130.71 0.43 fall 155 0.58 0.89 0.59 101.23 0.71 122.23 0.39 Annual 603 0.54 0.81 0.45 83.53 0.59 108.61 0.42 West winter 286 1.06 0.84 -0.24 -22.46 0.43 40.74 0.53 spring 294 0.41 0.51 0.09 22.40 0.19 47.11 0.74 summer 290 0.44 0.57 0.10 22.96 0.17 38.52 0.74 fall 277 0.68 0.71 0.04 5.26 0.25 37.41 0.63 Annual 1147 0.64 0.66 0.00 -0.31 0.26 40.55 0.67 National winter 3123 0.61 0.62 0.06 9.87 0.32 51.64 0.54 spring 3351 0.48 0.48 0.04 7.38 0.23 48.64 0.54 summer 3225 0.46 0.49 0.05 11.31 0.22 47.97 0.52 fall 3164 0.59 0.61 0.07 12.37 0.30 50.28 0.53 Annual 12863 0.54 0.55 0.06 10.28 0.27 49.77 0.55 30 ------- Table 8: EC model performance at IMPROVE sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 429 0.19 0.26 0.09 46.35 0.12 61.28 0.81 spring 478 0.15 0.19 0.04 28.44 0.07 46.50 0.86 Northeast summer 482 0.16 0.20 0.04 24.68 0.07 42.84 0.81 fall 459 0.20 0.24 0.05 25.24 0.10 49.89 0.78 Annual 1848 0.17 0.22 0.05 31.19 0.09 50.30 0.81 winter 228 0.14 0.19 0.06 40.29 0.08 52.98 0.82 Upper Midwest spring 239 0.19 0.18 -0.02 -9.41 0.08 43.15 0.54 summer 237 0.18 0.16 -0.02 -10.06 0.07 37.68 0.82 fall 245 0.20 0.20 -0.01 -3.52 0.08 39.14 0.83 Annual 949 0.18 0.18 0.00 1.62 0.08 42.50 0.70 winter 217 0.21 0.23 0.03 12.91 0.09 43.54 0.45 Ohio River Valley spring 242 0.21 0.20 -0.01 -6.43 0.09 42.45 0.27 summer 242 0.18 0.17 -0.02 -11.52 0.05 27.29 0.67 fall 232 0.30 0.26 -0.05 -17.91 0.09 31.04 0.67 Annual 933 0.23 0.21 -0.02 -7.18 0.08 35.71 0.47 winter 398 0.27 0.25 -0.01 -2.19 0.14 52.60 0.50 spring 447 0.36 0.22 -0.13 -35.35 0.20 56.02 0.18 Southeast summer 455 0.22 0.18 -0.03 -12.00 0.10 45.12 0.61 fall 423 0.35 0.24 -0.09 -24.55 0.14 39.08 0.83 Annual 1723 0.30 0.22 -0.06 -20.76 0.14 48.33 0.40 winter 240 0.17 0.16 0.00 -2.82 0.07 39.30 0.70 spring 273 0.17 0.18 0.00 2.26 0.09 51.62 0.60 South summer 250 0.12 0.10 -0.01 -9.28 0.05 41.67 0.64 fall 264 0.19 0.14 -0.04 -22.59 0.07 35.60 0.71 Annual 1027 0.16 0.15 -0.01 -8.39 0.07 42.11 0.64 winter 565 0.06 0.06 0.01 12.71 0.04 73.96 0.39 Northern spring 603 0.08 0.08 -0.01 -8.33 0.06 72.72 0.48 Rockies summer 631 0.10 0.15 0.05 46.72 0.09 86.82 0.28 and Plains fall 602 0.09 0.08 -0.01 -14.89 0.05 59.26 0.30 Annual 2401 0.08 0.09 0.01 11.24 0.06 73.79 0.33 winter 910 0.18 0.11 -0.06 -35.06 0.11 58.68 0.62 spring 994 0.09 0.09 0.01 8.20 0.06 68.82 0.56 Southwest summer 979 0.11 0.10 -0.01 -9.21 0.06 59.29 0.50 fall 964 0.14 0.10 -0.03 -20.83 0.08 57.62 0.56 Annual 3847 0.13 0.10 -0.02 -18.17 0.08 60.28 0.56 winter 447 0.08 0.12 0.04 53.70 0.08 102.20 0.83 Northwest spring 513 0.08 0.14 0.07 87.19 0.09 119.64 0.77 summer 519 0.14 0.22 0.08 57.63 0.16 114.54 0.48 31 ------- fall 500 0.12 0.23 0.11 97.37 0.16 135.94 0.71 Annual 1979 0.11 0.18 0.08 73.54 0.13 119.28 0.57 West winter 562 0.13 0.10 -0.02 -18.48 0.08 63.47 0.78 spring 605 0.08 0.09 0.02 24.98 0.05 64.13 0.73 summer 611 0.19 0.18 -0.01 -2.81 0.12 63.03 0.47 fall 576 0.15 0.16 0.00 1.75 0.08 54.84 0.71 Annual 2354 0.14 0.13 0.00 -1.01 0.08 61.04 0.56 National winter 3996 0.15 0.15 0.00 0.86 0.09 59.18 0.62 spring 4394 0.13 0.14 0.00 1.14 0.08 60.28 0.33 summer 4406 0.15 0.16 0.01 8.55 0.09 60.48 0.39 fall 4265 0.17 0.17 0.00 -2.32 0.09 53.64 0.63 Annual 17061 0.15 0.15 0.00 1.94 0.09 58.18 0.47 Soil Concentrations of crustal material (or soil) are calculated based on concentrations of 5 key crustal elements with mass adjustment factors that account for oxygen and other elements commonly bonded to those metals: Soil = 2.20 x Al + 2.49 x Si + 1.63 x Cct + 2.42 x Fe + 1.94 x Tj Maps of observed soil concentrations are shown in Figure 19. During winter and spring concentrations are largest in the Southwestern US (1-3 ng/m3) due to windblown dust at that time of year. Winter and spring concentrations in other parts of the US generally remain below 0.5 ng/m3. During summer and fall, concentrations between 1-3 ng/m3 are also observed across the Southern US, the plains states and in California in addition to in the Southwest. CAMx model predictions are shown in Figure 20 and generally overpredict soil concentrations over much of the US in all seasons (Figure 21) except in the Southwest and West because windblown dust emissions are not included in the simulation. Soil mean biases are in the range of ± 0.2-0.8 at CSN sites and ± 0.1-0.4 ng/m3 at IMPROVE sites for most regions and seasons (Table 9 and Table 10) with the exception of Ohio River Valley (summer/fall), the Upper Midwest (fall), the South (summer/fall), the Northwest (all seasons), and the West (summer). Underpredictions of soil in the summer across the South, Southwest, and West range from -0.5 to -0.9 Hg/m3. The monthly boxplots for soil show that similar to other primary PM components (i.e. elemental carbon) the urban (CSN) concentrations peaked in November in both the observations and the model with a consistent bias of around 0.5 ng/m3 across all months. At the rural IMPROVE sites, the observed concentrations peak in summer while the modeled concentrations peak during spring months leading to overestimates for most of the year averaging around 0.1-0.2 ng/m3 except for summer months for which the model average underestimates are in the range of 0.1-0.2 ng/m3. Correlation between model and monitored values shown in Tables 9 and 10 for soil are somewhat lower than for other PM species and range from 0.2-0.6 for most regions and seasons at CSN sites and 0.3-0.7 at IMPROVE sites. Correlations below 0.2 at CSN sites are found in the Ohio River Valley in summer, in the South during winter/spring/fall, in the Rockies Mountains and Plains in the winter, in the Southwest during all seasons and in the West during summer/fall. Correlations below 0.3 at IMPROVE sites are 32 ------- found in the Ohio River Valley in winter/summer, in the Southeast in summer, in the South in winter/fall, in the Southwest in fall, in the Northwest in summer/fall and in the West in summer/fall. >3 2.8 2.6 2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Figure 19. Observed soil concentrations (|ig/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). 33 ------- • IMPROVE * CSN • IMPROVE * CSN Figure 21. CAMx soil mean bias (pg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). Figure 20. Modeled soil concentrations (pg/m3) at CSN (triangles) and IMPROVE (circles) monitoring sties during winter (upper left), spring (upper right), summer (lower left) and fall (lower right). 34 ------- CSN, CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2, soil, 20150101 to 20161231, State=None IMPROVE, CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2, soil, 20150101 to 20161231, State=None —• IMPROVE - - a CAMx_2016fj_v710_CB6r5_NH3Rscale0_12US2 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months 1 1 1 1 1 1 1 1 1 1 1 2016_01 2016_03 2016_05 2016_07 2016_09 2016_11 Months Figure 22: Boxplots of observed and modeled soil concentrations (ng/m3) by month at CSN (right) and IMPROVE (left) monitoring sites. Lines indicate median concentrations across monitors in each month. Boxes delineate the 25th and 75th percentile ranges. Table 9: soil model performance at CSN sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 749 0.39 0.84 0.49 123.91 0.64 163.85 0.24 spring 813 0.50 0.98 0.49 99.44 0.61 122.82 0.38 Northeast summer 802 0.50 1.03 0.50 100.51 0.63 126.35 0.30 fall 761 0.51 1.19 0.70 137.12 0.86 166.99 0.28 Annual 3125 0.48 1.01 0.55 114.47 0.68 143.48 0.29 winter 306 0.31 0.74 0.41 134.76 0.49 161.35 0.34 Upper Midwest spring 323 0.50 1.12 0.68 136.29 0.77 152.44 0.53 summer 305 0.65 1.27 0.59 90.92 0.77 117.44 0.33 fall 310 0.58 1.48 0.86 146.94 1.00 171.77 0.30 Annual 1244 0.51 1.15 0.64 124.88 0.76 148.28 0.40 Ohio River Valley winter 546 0.47 1.06 0.65 138.79 0.82 176.47 0.25 spring 559 0.58 1.33 0.80 136.84 0.91 155.87 0.44 summer 560 0.74 1.63 1.04 141.04 1.20 162.82 0.18 35 ------- Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor fall 549 0.68 1.80 1.31 191.95 1.45 211.76 0.20 Annual 2214 0.62 1.46 0.95 153.55 1.10 177.08 0.28 winter 417 0.29 0.86 0.65 224.48 0.69 237.73 0.33 spring 456 0.52 1.06 0.57 111.14 0.73 140.22 0.20 Southeast summer 435 1.04 0.98 -0.02 -1.50 0.80 76.43 0.31 fall 424 0.57 1.15 0.58 100.12 0.68 118.45 0.44 Annual 1732 0.61 1.01 0.44 73.04 0.72 118.85 0.26 winter 327 0.58 1.32 0.79 136.38 1.10 190.38 0.03 spring 354 0.77 1.23 0.53 68.81 0.97 126.62 0.05 South summer 344 1.99 1.39 -0.60 -30.14 1.65 82.60 0.36 fall 330 0.84 1.70 0.86 102.24 1.33 157.28 0.09 Annual 1355 1.05 1.41 0.39 36.69 1.26 119.89 0.17 winter 147 0.27 0.54 0.29 104.81 0.45 166.17 0.16 Northern spring 150 0.43 1.01 0.54 125.46 0.56 130.55 0.60 Rockies summer 149 0.69 0.95 0.26 38.11 0.48 70.33 0.40 and Plains fall 140 0.53 1.18 0.62 115.65 0.76 142.10 0.40 Annual 586 0.48 0.92 0.42 88.12 0.56 116.74 0.41 winter 249 1.00 1.14 0.27 26.94 0.79 78.92 0.03 spring 253 1.40 1.33 0.09 6.43 0.87 61.89 0.12 Southwest summer 247 1.57 0.90 -0.57 -36.28 0.96 61.14 -0.15 fall 258 1.86 1.25 -0.42 -22.62 1.35 72.81 -0.05 Annual 1007 1.46 1.16 -0.16 -10.83 1.00 68.13 -0.01 winter 162 0.31 0.98 0.89 291.43 0.91 296.93 0.38 spring 162 0.47 1.58 1.54 325.48 1.55 327.66 0.60 Northwest summer 167 0.49 1.50 1.50 302.88 1.50 304.17 0.58 fall 160 0.44 1.44 1.54 352.20 1.57 359.14 0.36 Annual 651 0.43 1.37 1.37 319.45 1.38 323.16 0.49 winter 345 0.73 1.04 0.32 43.89 0.49 67.08 0.57 spring 352 0.76 1.18 0.45 59.42 0.58 75.61 0.57 West summer 349 1.23 0.83 -0.38 -30.89 0.74 60.14 -0.01 fall 329 1.35 0.99 -0.31 -23.33 0.75 55.76 0.18 Annual 1375 1.01 1.01 0.03 2.47 0.64 62.98 0.20 winter 3248 0.48 0.95 0.53 112.50 0.71 149.56 0.23 spring 3422 0.63 1.16 0.59 93.83 0.78 124.03 0.30 National summer 3358 0.94 1.18 0.29 30.70 0.94 99.68 0.18 fall 3261 0.78 1.38 0.67 85.57 1.05 135.52 0.12 Annual 13289 0.71 1.17 0.52 73.47 0.87 123.15 0.20 36 ------- Table 10: soil model performance at IMPROVE sites Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor winter 463 0.10 0.34 0.19 189.28 0.21 200.45 0.40 spring 481 0.23 0.49 0.22 95.21 0.23 101.37 0.68 Northeast summer 481 0.19 0.44 0.20 107.41 0.24 130.05 0.37 fall 459 0.13 0.50 0.31 247.55 0.32 251.01 0.57 Annual 1884 0.16 0.44 0.23 142.43 0.25 153.71 0.50 winter 216 0.12 0.35 0.17 141.34 0.19 157.87 0.53 Upper Midwest spring 208 0.28 0.65 0.29 102.01 0.31 109.05 0.74 summer 210 0.39 0.61 0.15 38.27 0.28 71.72 0.62 fall 215 0.26 0.72 0.36 141.99 0.38 149.54 0.69 Annual 849 0.26 0.58 0.24 93.02 0.29 111.06 0.62 winter 203 0.14 0.58 0.42 287.55 0.43 296.54 0.26 Ohio River Valley spring 209 0.36 0.78 0.43 118.24 0.45 124.66 0.63 summer 211 0.65 0.91 0.31 47.08 0.74 113.78 0.20 fall 198 0.39 1.10 0.76 191.96 0.84 213.11 0.34 Annual 821 0.39 0.84 0.47 121.10 0.61 157.27 0.30 winter 403 0.14 0.44 0.30 214.21 0.31 219.31 0.73 spring 413 0.35 0.68 0.33 94.37 0.37 105.52 0.60 Southeast summer 419 0.85 0.58 -0.27 -31.95 0.64 75.24 0.29 fall 391 0.32 0.61 0.31 98.36 0.37 116.08 0.59 Annual 1626 0.42 0.58 0.17 39.50 0.43 101.26 0.33 winter 250 0.32 0.58 0.27 83.35 0.45 140.87 0.02 spring 268 0.74 0.69 -0.05 -7.01 0.50 67.37 0.30 South summer 248 1.47 0.60 -0.87 -59.40 1.09 74.14 0.49 fall 265 0.54 0.74 0.20 37.48 0.54 100.27 0.11 Annual 1031 0.76 0.65 -0.11 -14.08 0.64 83.95 0.25 winter 558 0.13 0.25 0.12 94.82 0.16 126.47 0.50 Northern spring 571 0.41 0.61 0.20 49.20 0.29 70.37 0.63 Rockies summer 599 0.60 0.46 -0.14 -23.24 0.28 46.07 0.39 and Plains fall 574 0.36 0.54 0.18 50.85 0.34 96.52 0.48 Annual 2302 0.38 0.47 0.09 23.44 0.27 71.12 0.49 winter 981 0.52 0.41 -0.11 -21.16 0.40 77.01 0.33 spring 1016 1.18 0.79 -0.38 -32.47 0.59 50.16 0.50 Southwest summer 997 1.05 0.30 -0.74 -70.73 0.75 71.39 0.45 fall 984 0.85 0.36 -0.48 -56.99 0.57 66.83 0.28 Annual 3978 0.90 0.47 -0.43 -47.69 0.58 64.05 0.37 winter 475 0.07 0.21 0.15 231.30 0.16 246.53 0.72 Northwest spring 513 0.32 0.68 0.35 109.85 0.43 132.33 0.49 summer 512 0.46 0.43 -0.03 -6.72 0.42 91.06 0.09 37 ------- Region season n Average observed Concentration (Hg/m3) Average Modeled Concentration (Hg/m3) Mean Bias (Hg/m3) Normalized Mean Bias (%) Mean Error (Hg/m3) Normalized Mean Error (%) cor fall 499 0.19 0.39 0.20 100.93 0.32 166.42 0.19 Annual 1999 0.26 0.43 0.17 63.61 0.34 127.02 0.26 winter 623 0.20 0.33 0.13 65.13 0.21 102.88 0.54 spring 626 0.52 0.73 0.21 39.50 0.32 62.14 0.57 West summer 633 0.95 0.38 -0.57 -60.04 0.63 66.00 0.25 fall 605 0.72 0.33 -0.39 -54.47 0.53 73.93 0.14 Annual 2487 0.60 0.44 -0.16 -26.07 0.42 70.56 0.19 winter 4172 0.24 0.37 0.12 52.32 0.28 117.52 0.30 spring 4305 0.58 0.68 0.10 16.42 0.41 69.40 0.44 National summer 4310 0.76 0.46 -0.31 -40.94 0.56 73.16 0.23 fall 4190 0.48 0.51 0.01 1.80 0.46 94.60 0.14 Annual 16977 0.52 0.51 -0.02 -4.29 0.42 81.98 0.27 IV. Use of 2016fj PM Modeling as Base Year for Estimating Future EGU Benefits In this section we examine model performance in terms of the specific ways in which the modeling is applied for the proposed rule RIA. There are two key aspects to consider: 1) the use of modeling as an input into the 2016 and 2026 eVNA surfaces and 2) the use of modeling to determine the contribution of EGU emissions to PM2.5 concentrations nationwide. For calculating benefits, speciated PM2.5 model predictions are combined with observed speciated PM2.5 data to create a 2016 eVNA surface which is the basis, along with 2026 model predictions, for creating the 2026 eVNA surface. That is, the speciated PM2.5 surfaces are adjusted to conform with the magnitude spatial characteristics and of observed concentrations (see US EPA, 2022b for details). For instance, Figure 5 shows that model sulfate concentrations are overpredicted in the range of 0.1-0.5 Hg/m3 throughout much of the US but are underestimated during summer and fall in the Southwest and Texas. Figure 23 compares the 2016 CAMx and 2016 eVNA sulfate surfaces. This figure shows that the eVNA methodology adjusted annual average modeled sulfate concentrations downward by 0.1-0.5 Hg/m3 in the Eastern US and along the West coast but adjusted annual average sulfate concentrations upward by 0.1-0.4 ng/m3 in Texas. Similarly, Figure 17 shows (1) mostly unbiased EC CAMx predictions across the US with some isolated locations of EC overpredictions which are most pronounced in winter in urban areas and (2) EC underpredictions along the Appalachian Mountains and in the Northwestern US. Figure 24 compares 2016 CAMx and 2016 eVNA EC surfaces. This figure shows that the eVNA methodology did not significantly change modeled EC concentrations through most of the country but adjusted annual average EC downwards by 0.1-1 ng/m3 in urban areas such as Minneapolis, Chicago, New York and Houston and annual average EC upwards by 0.1-1 ng/m3 ug/m3 along the Appalachian Mountains and in the Northwest. Therefore, the fused eVNA surfaces minimize differences between the modeled and observed PM2.5 concentrations at monitoring locations. 38 ------- PM2.S sulfate: 2016 CAMx 2016-01-01 00:00:00 PM2.S sulfate: 2016 aVNA 2016-01-01 00:00:00 Figure 23: Comparison of annual average PM2.5 sulfate (ng/m3) for 2016 CAMx (left top) and 2016 eVNA (top right) and absolute PM2.5 sulfate difference (|ig/m3) between 2016 CAMx and eVNA (bottom). Blue colors on bottom plot represent higher sulfate concentrations in eVNA than in CAMx and green though red colors represent higher sulfate concentrations in CAMx. 39 ------- Figure 24: Comparison of annual average PM2.5 EC (pg/m3) for 2016 CAMx (left top) and 2016 eVNA (top right) and absolute PM2.5 EC difference (pg/m3) between 2016 CAMx and eVNA (bottom). Blue colors on bottom plot represent higher EC concentrations in eVNA than in CAMx and green though red colors represent higher EC concentrations in CAMx. The speciated PM2.s eVNA surfaces for the 2026 baseline are combined with the speciated state-EGU source apportionment contributions to modulate the baseline surfaces to reflect the impact of EGU emissions reductions from the various EGU policies in multiple rulemakings. Figures 25, 26 and 27 show the modeled contributions of EGU emissions to the 2026 eVNA surface for sulfate, nitrate, and primary PM2.5, respectively. Since modulating the PM2.5 surfaces to replicate baseline and policy emissions only occur in locations impacted by EGU emissions (i.e. red and purple colors in Figures 25, 26, and 27), model performance in other locations (i.e light yellow in Figures 25, 26, and 27) has little impact on the air quality impacts relevant for EGU policies. For instance, as shown in Figure 25, EGU sulfate contributions are most pronounced in the Eastern half of the US and in urban areas of California. In this respect, model performance for sulfate in other areas of the Western US would not be consequential for estimated the changes in sulfate expected to result from EGU policies. Similarly, EGU nitrate contributions (Figure 26) are highest in the Midwestern US, Salt Lake City, and California, so nitrate model performance in other parts of the country would have little impact on the predicted AQ. changes associated with EGU policies. Model biases in the Southeast US caused by not fully capturing large wildfires in the southern Appalachian Mountains (US Department of Agriculture, 2020) are unlikely to affect OC and EGU contributions from EGUs in that region. Primary PM2.5 contributions which include organic carbon, elemental carbon, and soil (Figure 27) are more heterogenous with sharper gradients from source locations. Again, model performance for EC, OC, and soil primary PM2.5 are not expected to impact AQ changed associated with EGU policies in locations that are distant from EGU sources where EGU contributions are lower (i.e. light yellow areas in Figure 27). 40 ------- Taken together, the model performance for PM2.5 species, as described in the previous section, is acceptable for use in determining EGU impacts when using eVNA surfaces and EGU modeled contributions in a relative manner to estimate the spatial fields of PM2.5 concentrations that properly reflect the impact of changes in EGU emissions for the purposes of estimating benefits associated with EGU policies. EGU S02 contribution to PM2.5 2026fj 80 159 239 318 Min = 0.00E+0 at (1,1), Max = 0.663 at (268,124) Figure 25: Sulfate concentrations (ng/m3) from EGU S02 emissions in 2026 0.700 0.675 0.600 0.525 0.450 0.375 0.300 0.225 0.150 0.075 0.000 41 ------- EGU NOx contributions to PM2.5 2026fj 80 159 239 318 Min = 0.00E+0 at (1,1), Max = 0.169 at (36,113) Figure 26: Nitrate concentrations (|ig/m3) from EGU NOx emissions in 2026 42 ------- 0.700 0.675 EGU Primary PM contributions to PM2.5 2026fj 80 159 239 318 Min = 0.00E+0 at (1,1), Max = 6.339 at (24,124) Figure 27: Primary PM25 concentrations (ng/m3) from EGU PM2.s emissions in 2026 V. References Karl, TR; Koss, WJ. (1984). Regional and National Monthly, Seasonal and Annual Temperature Weighted by Area, 1895-1983. National Oceanic and Atmospheric Administrations, National Environmental Satellite, Data, and Information Service, National Climatic Data Center, Asheville, NC, June 1984 Simon, H., Baker, K.R., and Phillips, S. (2012) Compilation and interpretation of photochemical model performance statistics published between 2006 and 2012. Atmospheric Environment 61, 124-139. US Department of Agriculture (2020) Fire in the Southern Appalachians: Understanding Impacts, Interventions, and Future Fire Events by Natasha A. James, Karen L. Abt, Gregory E. Frey, Xue Han, Jeffrey P. Prestemon, Gen. Tech. Rep. SRS-249. Asheville, NC: U.S. Department of Agriculture Forest Service, Southern Research Station. Available at: https://doi.org/10.2737/SRS-GTR-249 US Environmental Protection Agency, 2018. Modeling Guidance for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze, Research Triangle Park, NC. https://www3.epa.gOv/ttn/scram/guidance/g:uide/03-PM-RH-Modeling: Guidance-2018.pdf US EPA (2022a) Regulatory Impact Analysis for Proposed Federal Implementation Plan Addressing Regional Ozone Transport for the 2015 Ozone National Ambient Air Quality Standard. EPA-452/D-22-001. 43 ------- February 2022. U.S. Environmental Protection Agency Office of Air Quality Planning and Standards Health and Environmental Impacts Division, Research Triangle Park, NC US EPA (2022b) Regulatory Impact Analysis for Proposed Reconsideration of the Steam Electric Power Generating Effluent Limit Guidelines. U.S. Environmental Protection Agency Office of Water, Washington DC. US EPA (2022c) Technical Support Document (TSD): Preparation of Emissions Inventories for the 2016v2 North American Emissions Modeling Platform. EPA-454/B-22-001. February 2022. U.S. Environmental Protection Agency Office of Air Quality Planning and Standards Air Quality Assessment Division, Research Triangle Park, NC 44 ------- |