Skip to main content
NCBI home page
EPA Author Manuscripts logoLink to EPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Sep 19.
Published in final edited form as: Environ Sci Technol. 2017 Mar 15;51(7):3833–3842. doi: 10.1021/acs.est.6b05069

Assessing Model Characterization of Single Source Secondary Pollutant Impacts Using 2013 SENEX Field Study Measurements

Kirk R Baker †,*, Matthew C Woody ‡
PMCID: PMC6145072  NIHMSID: NIHMS982359  PMID: 28248097

Abstract

Aircraft measurements made downwind from specific coal fired power plants during the 2013 Southeast Nexus field campaign provide a unique opportunity to evaluate single source photochemical model predictions of both O3 and secondary PM2.5 species. The model did well at predicting downwind plume placement. The model shows similar patterns of an increasing fraction of PM2.5 sulfate ion to the sum of SO2 and PM2.5 sulfate ion by distance from the source compared with ambient based estimates. The model was less consistent in capturing downwind ambient based trends in conversion of NOX to NOY from these sources. Source sensitivity approaches capture near-source O3 titration by fresh NO emissions, in particular subgrid plume treatment. However, capturing this near-source chemical feature did not translate into better downwind peak estimates of single source O3 impacts. The model estimated O3 production from these sources but often was lower than ambient based source production. The downwind transect ambient measurements, in particular secondary PM2.5 and O3, have some level of contribution from other sources which makes direct comparison with model source contribution challenging. Model source attribution results suggest contribution to secondary pollutants from multiple sources even where primary pollutants indicate the presence of a single source.

graphic file with name nihms-982359-f0001.jpg

Introduction

Impacts from specific sources have traditionally been estimated for pollutants with National Ambient Air Quality Standards through permit programs including New Source Review and Prevention of Significant Deterioration. However, little specific guidance exists in the 2005 version of the Guideline for Air Quality Models for single source air quality impact demonstrations for ozone (O3) and secondary particulate matter less than 2.5 μm in diameter (PM2.5) impacts to support permit applications under these programs. Traditional dispersion models (e.g., AERMOD) used to assess single source impacts for permit related programs do not contain either the plume chemistry or atmospheric chemistry and physics necessary to appropriately characterize O3 and secondary PM2.5 impacts on the environment. The U.S. Environmental Protection Agency recently finalized changes to the Guideline for Air Quality Models and established new guidance that describes the type of models and methods used to assess single source secondary pollutant impacts for permit programs.(1) Public comment on this recent revision included requests that models for this purpose be further evaluated for suitability of estimating single source O3 and secondary PM2.5 impacts.(1)

Photochemical grid modeling systems have been used extensively to estimate source specific impacts on O3 and secondary PM2.5 with various techniques to differentiate those impacts from other sources.(2–6) Previous studies have shown that Eulerian photochemical grid models are able to replicate near-source secondary pollutant scavenging (e.g., titration) and downwind production of O3.(2, 6) Some photochemical grid models have been instrumented with a Lagrangian subgrid plume model to provide source impacts at subgrid scales with the intent of more finely resolved spatial representation of plumes by not immediately diluting emissions into the grid volume.(5, 7–15) Emissions mass is typically tracked in the Lagrangian submodel until plume sizes are comparable to the size of the host model grid cell at which time the subgrid pollutant mass is transferred to the host grid cell.(7, 11) Past evaluations of subgrid plume model predictions of secondary pollutants in 3-dimensional photochemical models include comparisons against routine surface monitors which suggest differences in overall model performance with and without subgrid treatment for specific sources are usually insignificant.(8, 16) It is not clear whether the small differences in model performance are related to the different chemistry realized at the subgrid level or differences in the transport of primary and secondary pollutants by the subgrid plume treatment implementation.

Source oriented field measurements of ambient air dominated by a single facility provide an opportunity to evaluate air quality modeling systems that estimate single source impacts. In-plume measurements of O3, O3 precursors such as nitrogen oxides (NOX) and volatile organic compounds (VOC), and other secondary pollutants have been sparse leading to limited comparisons with photochemical model predictions. Recent in-plume measurements of these same pollutants and secondary PM2.5 made during the 2013 Southeast Nexus (SENEX) field campaign in the southern United States provide the first opportunity to extend these earlier evaluations to include secondary PM2.5.(17) Plume enhancements of particulate matter have been identified downwind of multiple single sources from this field campaign based on relationships in the ambient data.(18) Here, a photochemical transport model was applied coincident with the in-plume field study measurements downwind of four separate electrical generating units using source sensitivity and source apportionment methods to differentiate these sources from other local and regional sources. In addition, subgrid plume treatment was applied with the source sensitivity approach to better understand whether treating source emissions with a Lagrangian submodel results in better agreement with in-plume source measurements. Model estimates are compared with ambient estimates of single source impacts both in space and time to illustrate model skill in plume placement and also just time to emphasize chemical plume evolution recognizing that for some types of applications skill in near-source plume placement is less important than chemical production.

Materials and Methods

The Community Multiscale Air Quality (CMAQ) model (www.cmascenter.org/cmaq) version 5.0.2 was used to model the periods of 2013 coincident with in-plume measurements made as part of the SENEX field campaign: June 16 and 22 for Scherer and Harllee electrical generating unit (EGU) facilities in northern Georgia, June 26 for Independence power plant in northeast Arkansas, and July 3 and 8 for New Madrid power plant in southeast Missouri (Table 1). The CMAQ model includes gas-phase chemistry using the Carbon Bond approach,(19) inorganic aerosol partitioning based on ISORROPIA II,(20) 2-product semivolatile organic aerosol partitioning,(21) and aqueous phase chemistry that includes sulfur oxidation.(22) CMAQ was applied with 3 different model domains using 4 km sized grid cells covering an area centered over field measurements in Arkansas, Missouri, and Georgia (Supporting Information (SI) Figure S1). The atmosphere is resolved with 35 layers from the surface (layer 1 height approximately 22 m) to the 50 mb top of the model with more layers in the boundary layer to better represent diurnal changes in mixing layer heights.

Table 1.

Stack Release Characteristics for the Facilities Modeled for Contribution Assessment

Facility Name Stack Height (m) Stack Diameter (m) Exit Temperature (C) Exit Flow Rate (m3/s) Exit Velocity (m/s) Longitude Latitude
Scherer 305 8 128 1283 24 −83.8064 33.0613
305 8 137 1234 23 −83.8074 33.0595
258 10 56 1463 17 −83.8064 33.0634
New Madrid 244 6 157 690 24 −89.5683 36.5136
244 6 177 623 21 −89.5617 36.5147
244 6 162 606 21 −89.5683 36.5136
Independence 305 8 161 1322 27 −91.4078 35.6782
305 8 161 1322 27 −91.4078 35.6782
Harllee 305 9 127 1435 23 −83.2997 33.1945
305 7 121 920 25 −83.2997 33.1945
Open in a new tab

Anthropogenic emissions from area and mobile sources are based on the 2011 National Emission Inventory.(23) Wild and prescribed fires are day specific for 2013 and point sources are based on 2013 Continuous Emissions Monitoring information where available and 2011 otherwise. The sources modeled here for impacts are based on 2013 emissions. Hourly emissions from each of the 4 facilities tracked for contribution are shown in SI Figure S2. The Harllee Branch facility has the highest emissions due to the lack of emissions control equipment during 2013. Location and stack release characteristics for these facilities are shown in Table 1. Biogenic emissions are estimated using version 3.61 of the Biogenic Emission Inventory System model with Biogenic Emissions Landuse Database version 4 vegetation information.(24) Previous work shows good agreement between biogenic impacts and field study measurements in the southeast United States.(25) Meteorological inputs to CMAQ were generated using version 3.7 of the Weather Research & Forecasting model.(26) Previous applications of this model show generally good agreement with surface variables and mixing layer height.(27)

Multiple approaches were used to differentiate single source impacts on modeled primary and secondary pollutants. A source sensitivity approach estimates source impacts by difference between 2 CMAQ model simulations, one of which contains all sources and another which contains all sources except the source of interest (e.g., brute-force zero-out).(28) The Integrated Source Apportionment Method (ISAM) was also used to track the contribution of each source of interest through all chemical and physical processes in the model and estimate the contribution to total bulk model predictions for O3(29) and PM2.5.(30) Source sensitivity(2, 4–6) and source apportionment(2, 31) approaches have been used in the past to estimate single source impacts using photochemical grid models.

Advanced Plume Treatment (APT)(8) is a subgrid plume treatment approach integrated in CMAQ and was applied with the brute-force zero-out sensitivity method to differentiate source impacts. CMAQ-APT uses the Second-Order Closure Integrated puff model with CHEMistry (SCICHEM) to treat subgrid puffs within CMAQ.(32) SCICHEM uses overlapping 3-D puffs to represent plumes for emissions from a specified source. Puffs evolve on a subgrid scale, undergoing transport and various other plume dynamics (plume rise, dispersion, puff merging/splitting) which use information from the host grid cell but are not consistent with the host model advection and diffusion schemes. The subgrid chemical evolution uses an identical chemical mechanism and aerosol module as the parent grid model. Once puffs reach a similar size to the model grid, the puffs are “merged” with the grid, meaning the mass contained in the puff is added to the grid and the puff is no longer tracked. Because CMAQ-APT does not support inline photolysis, which was used in CMAQ and CMAQ-ISAM simulations, photolysis look-up tables were used in CMAQ-APT. Other relevant CMAQ-APT parameters are summarized in SI Table S1. The average number of puffs in a domain per facility at a given time during these simulations was 60 with the maximum being 243. The average time between puff releases was 30 s with a maximum value of 100 s which is comparable to tolerances set in other subgrid plume treatment approaches.(2)

As a postprocessing step, the “merge concs” CMAQ-APT utility was used to merge CMAQ-APT puffs with gridded CMAQ-APT results. This technique adds puff mass to the model grid to allow consistent comparisons against traditional gridded CMAQ results (SI Figure S3). In practice, all puff mass would merge into the grid if enough time has elapsed for the puff size to reach the size of the grid. However, when comparing hourly gridded impacts without using “merge concs”, the grid-only impacts would exclude pollutants contained within puffs at that specific hour. Another CMAQ-APT postprocessing utility, “receptor concs” was used to assess subgrid scale impacts at aircraft measurement locations (SI Figure S3). These results represent subgrid scale impacts (i.e., point-based impacts) estimated as host grid cell concentrations plus subgrid puff concentrations. Here, CMAQ-APT results postprocessed using “merge concs” are referred to as merged or grid based while those postprocessed using “receptor concs” are referred to as point based.

Details regarding measurements used here to evaluate the model are provided elsewhere.(17)Briefly, aircraft based measurements were collected at time intervals ranging from every 1 s (for NOX, O3, and SO2) to every 10 s (for PM1 sulfate ion). Measurements used in this study include: submicron speciated PM measured using Aerosol Mass Spectrometry (AMS);(33) SO2 measured by pulsed UV fluorescence;(34) and NO, NO2, NOY, and O3 measured by gas phase chemiluminescence.(34–36) Modeled PM2.5, calculated as the sum of the Aitken and accumulation modes, was compared without adjustment to measured submicron PM1. This approach introduces uncertainty into the comparison, but converting CMAQ PM2.5 to match AMS PM1 would introduce different uncertainties since CMAQ sometimes overpredicts fine particle mean diameters(37, 38)and AMS transmission efficiency changes as a function of particle size.(39) A review of AMS PM1sulfate ion and particle volume (SI Figure S4) for all flights examined here showed relationships consistent with those for the Georgia flights published elsewhere.(18) PM1 sulfate ion tends to follow temporal variation seen in the particle volume measurements and often will increase when volume increases.

To compare modeled single-source impacts to observations, model predictions and observations were matched in horizontal and vertical space and time. Ambient based source specific contribution were estimated by subtracting observed “background” concentrations, here defined as the median observed pollutant concentration on a flight-by-flight basis, from each transect measurement.(2) The median values for each flight day and pollutant are shown in SI Table S2. When comparisons of modeled and observed plume centerlines were made, modeled plume centerlines were adjusted horizontally such that the maximum modeled and observed NOY were matched in space. Note that no adjustments were made vertically for centerlines. Flight days and transects were prioritized for use in this model evaluation where multiple well oriented downwind transects were made from a large power plant and extra consideration was given for flight days and plants examined elsewhere.(18)

Results and Discussion

Conceptual Understanding of Source Contributions to Flight Days

Southerly winds transported the New Madrid plume toward the north during the day on July 8th (see Figure 1) and to the southeast during the July 3rd nighttime flight transects (SI Figure S5). Figure 1 shows the ambient based (measurements minus median value from flight) and CMAQ source apportionment estimates of July 8 daytime downwind impacts from the New Madrid facility, all other EGUs in the model domain, and all non-EGU point sources in the model domain. The additional source apportionment results are presented to help identify other local sources that may be contributing to the ambient based impacts. Figure 1 shows while the New Madrid contribution to SO2 and PM2.5 sulfate ion is typically largest, a nearby industrial source (Noranda Aluminum) also contributed SO2 and PM2.5 sulfate to the downwind transects along with nearby power plants (Sikeston and Joppa). Contributions from these facilities, especially Sikeston and Joppa, are also evident in the ambient enhancements downwind of those facilities (Figure 1).

Figure 1.

Figure 1.

Open in a new tab

SO2, PM2.5 sulfate ion, NOX, and O3 contribution estimates downwind of the New Madrid power plant on July 8, 2013 based on measurements (left column) and a source apportionment (ISAM) photochemical model simulation. The ISAM approach was also used to estimate the contribution from all other EGUs in the model domain and non-EGU point sources in the model domain. Modeled contribution from the facility shown here is the maximum contribution between the top of the stack and top of the surface mixing layer.

Figure 2 shows downwind transect ambient and source apportionment modeled SO2 and PM2.5sulfate ion impacts from the Independence power plant and New Madrid power plant. Other local sources, in particular Noranda Aluminum and Sikeston Power, have notable modeled impacts on the nighttime transects made downwind of New Madrid on July 3. However, the nighttime flight has a large contribution from sources outside the model domain (or recirculation of sources inside the domain) which may be PM2.5 sulfate ion formed the previous day that is above the surface mixed layer. Source apportionment modeling showed little notable contribution from other local sources in transects made downwind to the northeast (southwesterly winds) of the Independence facility on June 22 (Figure 2 and SI Figure S6).

Figure 2.

Figure 2.

Open in a new tab

SO2 and PM2.5 sulfate ion contribution estimates downwind of the Independence and New Madrid (night-time flights) power plants based on measurements and source apportionment photochemical model simulations. Modeled contribution from the facility shown here is the maximum contribution between the top of the stack and top of the surface mixing layer.

The model estimated plume tends to be slightly lower in altitude than the aircraft during the nighttime flight downwind of New Madrid (SI Figure S9). Plots of daytime modeled facility impacts show the plume to be well mixed through the boundary layer during the period of aircraft measurements (SI Figure S10). During the nighttime flight, the model does predict large SO2impacts in the vertical layers between the stack release height and mixing layer height that are not well mixed upward or downward suggesting that the stack release point is above the modeled surface mixing layer during the hours of the flight. This feature suggests the CMAQ vertical advection local closure scheme is dominant over the nonlocal closure mixing scheme, which allows mass to be transported in the vertical direction to many layers as opposed to just the layers immediately adjacent to the emissions.(40)

Downwind daytime transects were made from Harllee Branch and Scherer power plants in Georgia (June 16 and 22). The model was able to differentiate the impacts of Scherer and Harllee Branch from other sources on both flight days. Ambient impacts from these facilities are most distinguishable on June 16 (SI Figure S7) due to favorable winds and downwind transect arrangement but the impacts are more problematic to differentiate in the June 22 (SI Figure S8) downwind transect measurements due to easterly winds. Other EGUs in the model domain have a notable contribution to O3 and PM2.5 sulfate ion at the transect measurements made downwind of Harllee Branch and Scherer.

Single Source Plume Comparison

Ambient based and modeled impacts are shown as a function of plume centerline, which is based on NOY, in Figure 3 for the Independence plant (and SI Figures S11–S13 for other plants). This analysis more clearly illustrates plume structure by distance from the source since spatial differences exist between the ambient and modeled plume due to the modeled representation of local transport. The modeled impacts are shown using source apportionment and source sensitivity with and without subgrid plume treatment.

Figure 3.

Figure 3.

Open in a new tab

Model and observed concentrations depicting plume centerlines (matched in space by adjusting max NOy model results horizontally to match max NOy observations) at various distances downwind of the New Madrid Plant on July 8, 2013. Model results include the advanced plume treatment grid-based results (APT Grid), point-based (i.e. receptor) APT results (APT Point), brute-force (BF), and integrated source apportionment method (ISAM).

Modeled single source SO2 impacts are highest nearest the source and PM2.5 sulfate ion impacts show less of a downwind gradient. Ambient based impacts of facility specific SO2 are often narrower in horizontal spatial extent and larger in magnitude than what is estimated by the modeling system at transects in close proximity to the source. The spatial scale and magnitudes become closer at the transects farther from these sources. Plume structure is most notable in the downwind transects of the Independence plant (Figure 3). The point-based subgrid plume treatment did well to capture the observed SO2, but this did not translate to improved model performance farther downwind. Estimated PM2.5 sulfate contribution is both higher and lower compared with ambient based estimates. Ambient based PM2.5 sulfate ion plume estimates do not tend to show large centerline peak concentrations compared with SO2, but are sampled over a slightly longer time interval. Figure 3 shows a sharp centerline PM2.5 sulfate ion peak predicted by the APT approach but not grid-based approaches. This feature is also not evident in the ambient data.

Similar to SO2, the modeling system tends to estimate lower NOX contribution from these plants compared to ambient based estimates, especially at transects in closest proximity to the source. Typically, each of the methods resulted in similar NOX contribution estimates downwind. One exception is the APT approach contribution estimates downwind of the Independence plant (Figure 3), which are notably larger than the grid-based approaches in the nearer transects. The source sensitivity approaches (brute-force and brute-force with APT) capture near-source NO titration of O3 where indicated by ambient data although the model tends to not capture the magnitude of the O3 decrease. APT is the only approach to capture notable near-source O3decreases due to titration from fresh NO emissions but this does not translate into better O3production estimates at transects farther downwind based on ambient data. The model often predicts smaller contribution compared to the ambient approach but did better at matching ambient based downwind O3 when comparing near-peak 98th percentile estimates (see Table 2). It is important to note that it is particularly difficult to differentiate O3 from specific sources in ambient data due to the regional nature of the pollutant and many local sources in these areas. The spatial and plume cross-section comparisons demonstrate model ability to capture single source primary and secondary pollutant impacts over time and space which is similar to that shown in other studies. (2)

Table 2.

Source Contribution to Ambient Measurements (Measured–Median of Flight Segment) and Modeled Contribution Using Multiple Methods

98th percentile estimate bias (OBS-MODEL)
pollutant flight day related facility units OBS ISAM BF APT ISAM BF APT
SO2 130616 Scherer ppb 5.1 3 1.8 1.6 −2.2 −3.3 −3.6
PM2.5 sulfate 130616 Scherer ug/m3 1.4 1.5 0.5 0.5 0.1 −0.9 −0.9
NOX 130616 Scherer ppb 1.7 1.7 0.1 0.1 0 −1.6 −1.6
O3 130616 Scherer ppb 10.5 2.5 1.2 1.5 −7.9 −9.2 −9.0
SO2 130616 Harllee Branch ppb 0.7 0.3 19.4 12.1 −0.4 18.7 11.4
PM2.5 sulfate 130616 Harllee Branch ug/m3 0.3 0.4 3.3 2.1 0.1 3 1.8
NOX 130616 Harllee Branch ppb 0.2 0.2 5.5 3.1 0 5.2 2.9
O3 130616 Harllee Branch ppb 2.3 0.7 6.9 5.4 −1.5 4.6 3.1
SO2 130626 Independence ppb 7 3.8 3.8 3 −3.2 −3.2 −4.0
PM2.5 sulfate 130626 Independence ug/m3 0.6 1.5 0.9 0.3 0.9 0.3 −0.3
NOX 130626 Independence ppb 1.9 2.2 2.1 1.6 0.3 0.2 −0.3
O3 130626 Independence ppb 7.8 2.6 2.2 1.7 −5.2 −5.6 −6.1
SO2 130703 New Madrid ppb 20.6 6.3 6.4 0.9 −14.3 −14.1 −19.7
PM2.5 sulfate 130703 New Madrid ug/m3 4.1 0.4 0.4 0.1 −3.7 −3.7 −4.0
NOX 130703 New Madrid ppb 13.9 8.6 8.5 1.2 −5.3 −5.4 −12.7
O3 130703 New Madrid ppb 14.8 0 0.1 0.1 −14.8 −14.6 −14.7
SO2 130708 New Madrid ppb 13.9 3.6 3.6 2.7 −10.3 −10.3 −11.2
PM2.5 sulfate 130708 New Madrid ug/m3 5.6 2.7 1.6 1.7 −2.9 −4.0 −4.0
NOX 130708 New Madrid ppb 4.3 5.7 4.9 3.2 1.4 0.6 −1.1
O3 130708 New Madrid ppb 8.6 7.5 8.2 9.2 −1.1 −0.4 0.6
Open in a new tab

Daytime Plume Chemical Ratios

Ratios, as opposed to absolute concentrations, provide a method of normalization across model and observed values to focus the interpretation of results on chemical processes and minimize differences due to pollutant transport or emissions. The PM2.5 sulfate ion to SO2 + PM2.5 sulfate ion ratio is intended to provide information about how quickly SO2 is converted to particulate sulfate in these plumes and shown averaged by transect for each facility in Figure 4. Table 3 shows the flight averaged ratio and bias between the ambient based ratio and those estimated with modeling approaches. Downwind particulate sulfate production is well characterized by all methods for the New Madrid transects (bias less than 2%). The ISAM approach tends to overestimate downwind sulfate production at Independence (bias of 7%) and Scherer (bias of 20%) but is the only approach to fully capture the peak downwind production at Harllee Branch (bias of 1%) and was the closest to peak downwind production at Scherer. The production of particulate sulfate in these downwind plumes was generally well represented by these approaches for isolating single source impacts with average ratio bias less than 8% except for overestimation by the ISAM approach for the closest transects downwind of Scherer.

Figure 4.

Figure 4.

Open in a new tab

Average observed and modeled species ratio of particulate sulfate to the sum of SO2 and particulate sulfate at multiple downwind distances from each source. Only the daytime transect information is included for New Madrid.

Table 3.

Average Observed and Modeled Species Ratio over All Transects of (1) Particulate Sulfate to the Sum of SO2 and Particulate Sulfate and (2) NOX to NOYfor Each Source

average ratio bias (MODEL - OBS)
ratio plant OBS ISAM BF APT APT Point ISAM BF APT APT Point
Sulfur Ratio Harllee Branch 0.14 0.15 0.10 0.09 0.09 1.3 −4.1 −4.4 −4.6
Independence 0.02 0.09 0.06 0.03 0.02 7.2 4.1 0.9 0.1
New Madrid 0.16 0.17 0.15 0.18 0.16 1.1 −1.6 1.8 −0.2
Scherer 0.18 0.38 0.13 0.13 0.12 19.5 −5.1 −5.6 −6.1
NOX:NOY Ratio Harllee Branch 0.61 0.54 0.49 0.48 0.50 −7.3 −12.2 −13.4 −10.8
Independence 0.79 0.86 0.82 1.00 0.91 7.2 3.4 21.0 11.6
New Madrid 0.70 .63 .51 1.00 0.48 −6.3 −18.9 29.9 −22.2
Scherer 0.50 0.21 0.17 0.16 0.17 −29.5 −33.3 −33.7 −33.0
Open in a new tab

The NOX to NOY ratio provides a proxy for photochemical age,(41–43) or how quickly or slowly primary emissions are converted to secondary pollutants. All of the methods converted NOX to NOY too quickly downwind of Scherer and at transects furthest from Harllee and New Madrid (SI Figure S14), which are noted in Table 3 as negative bias in the ratio. The conversion downwind of Independence is generally well characterized by the BF and ISAM approaches. However, the APT approaches, in particular the APT grid approach, estimated less chemical conversion compared to the ambient data. The APT-grid approach shows almost no chemical production downwind of either Independence or New Madrid which is inconsistent with ambient measurements and the other modeling approaches (SI Figure S14). The difference could be attributed to the use of photolysis lookup tables, subgrid plume treatment in APT, or other factors. However, sensitivity brute-force simulations using photolysis lookup tables in place of in-line photolysis actually decreased the ratio of NOX to NOY by 6% at the 48 km transect downwind of the Independence facility suggesting the lower conversions with APT are likely attributable to subgrid plume treatment or the postprocessing approach of generating merged grid and puff estimates.

Finally, the NOY to SO2 ratio is used to assess modeled emissions compared to observations (Figure S15). Good agreement with the modeled and observed ratio suggests the modeled emissions are well characterized. The one exception is the APT results at the New Madrid facility, which predicted less NOY producing a lower ratio that is inconsistent with ambient data. The contrast of grid and point-based APT results suggests that the postprocessing to merge puffs into grids may be placing puff mass into adjacent grid cells (either vertically or horizontally) leading to differences in adjacent grid cells and possibly previous and subsequent hours.

Modeled Contribution Estimate Comparisons

Figure 5 shows modeled contribution from the New Madrid power plant on July 8, 2013 to illustrate differences in methods. A morning period is chosen to illustrate differences in subgrid puff placement and the afternoon to illustrate both chemical and physical differences in these approaches. ISAM estimated surface SO2 contribution (Figure 5a) at 10 am shows the highest levels nearest the plant and decreasing contribution in the direction of prevailing winds to the northeast. Downwind contribution is similar using the BF approach (Figure 5b). The BF approach shows higher contribution than APT nearer the source and lower contribution starting at approximately 12–15 km downwind. This feature is related to APT keeping mass in subgrid puffs closer to the source then dumping that mass into the host grid cells farther downwind (Figure 5c). ISAM, BF, and APT estimated contribution of the power plant to surface layer NOX and O3 are shown for 2 pm. Negative contribution for the source sensitivity methods represents how O3 levels would change in the absence of NOX emissions which results in less O3 due to titration by fresh NO emissions. Here, titration is more muted when APT is employed (Figure 5h, 5i) which is related to less grid-resolved NOX compared to the BF method (Figure 5f). ISAM is designed to only track contribution to O3 production meaning no contribution will be estimated when destruction processes dominate (Figure 5g).

Figure 5.

Figure 5.

Open in a new tab

Model estimated contribution from New Madrid power plant on July 8, 2013 to SO2 (top row), NOX (middle row), and O3 (bottom row).

For the transects downwind of the power plants shown here, the ISAM and BF approaches estimate similar contribution for each of the pollutants except situations where O3 destruction outpaces production. The APT approach results in notable differences from the BF method but are not directionally systematic (SI Figure S16). ISAM tends to predict higher PM2.5 sulfate ion contribution compared to the sensitivity based approaches. However, it is not clear that this resulted in a systematic difference in bias compared to ambient based contribution estimates. The APT approach changes model estimated contributions through both chemical and physical treatment. A comparison is provided using a conserved tracer with little chemical reactivity (e.g., CO) to illustrate nonchemical impacts of subgrid plume treatment. SI Figure S17 shows the episode average CO estimates using the brute-force zero out approach with and without APT. For this source and time period the APT approach results in higher contribution aloft and lower at the surface. This tendency toward less mixing to the surface has been shown in another subgrid plume scheme implemented in a different host model.(7) While the subgrid impacts on pollutant chemistry may be desired, the impacts on model estimates due to puff transport are notable and should not be discounted when interpreting results generated with APT or when comparing estimates with other approaches. It is possible that the incommensurate advection and diffusion between the subgrid and host model contribute to a physically less realistic outcome in terms of source impacts compared to using either modeling system singularly.

Each of these approaches for differentiating single source impacts tends to provide a reasonable representation of downwind secondary chemical production. However, particular approaches may be better suited for different types of assessments. Where understanding how much air quality might change due to changes in emissions then a source sensitivity approach may be best, especially in situations where highly nonlinear chemistry is anticipated. If the contribution to modeled pollutants is desired, then source apportionment would be appropriate. Finally, in situations where an emissions source and key receptors are within the same grid cell then subgrid plume treatment approaches would provide additional information.

Supplementary Material

Sup

Acknowledgment

We recognize the contributions of James Kelly, Norm Possiel, Allan Beidler, James Beidler, Chris Allen, Lara Reynolds, and Eladio Knipping. We also acknowledge all the participants of the 2013 SENEX field campaign for providing the aircraft measurements and in particular Thomas Ryerson, John Holloway, Ann Middlebrook, and Joost de Gouw. Although this work was reviewed by EPA before publication, it may not necessarily reflect official Agency policy.

References

  • 1.U.S. Environmental Protection Agency, Air Quality ModelGuidelines: Enhancements to the AERMOD Dispersion Modeling Systemand Incorporation of Approaches to Address Ozone and Fine ParticulateMatter; Revisions. Docket ID No. EPA-HQ-OAR-2015–0310 2015. https://www.regulations.gov/#!documentDetail;D=EPA-HQ-OAR-2015-0310-0001
  • 2.Baker KR; Kelly JT Single source impacts estimated with photochemical model source sensitivity and apportionment approaches Atmos. Environ 2014. 96, 266–274 DOI: 10.1016/j.atmosenv.2014.07.042 [DOI] [Google Scholar]
  • 3.Baker KR; Kotchenruther RA; Hudman RC Estimating ozone and secondary PM 2.5 impacts from hypothetical single source emissions in the central and eastern United States Atmos. Pollut. Res 2016, 7(1) 122–133 DOI: 10.1016/j.apr.2015.08.003 [DOI] [Google Scholar]
  • 4.Bergin MS; Russell AG; Odman MT; Cohan DS; Chameldes WL Single-Source Impact Analysis Using Three-Dimensional Air Quality Models J. Air Waste Manage. Assoc 2008, 58 (10) 1351–1359 [PubMed] [Google Scholar]
  • 5.Kelly JT; Baker KR; Napelenok SL; Roselle SJ Examining single-source secondary impacts estimated from brute-force, decoupled direct method, and advanced plume treatment approaches Atmos. Environ 2015, 111, 10–19 DOI: 10.1016/j.atmosenv.2015.04.004 [DOI] [Google Scholar]
  • 6.Zhou W; Cohan DS; Pinder RW; Neuman JA; Holloway JS; Peischl J; Ryerson TB;Nowak JB; Flocke F; Zheng WG Observation and modeling of the evolution of Texas power plant plumes Atmos. Chem. Phys 2012, 12 (1) 455–468 DOI: 10.5194/acp-12-455-2012 [DOI] [Google Scholar]
  • 7.Baker KR; Hawkins A; Kelly JT Photochemical grid model performance with varying horizontal grid resolution and sub-grid plume treatment for the Martins Creek near-field SO2 study Atmos. Environ 2014,99, 148–158 DOI: 10.1016/j.atmosenv.2014.09.064 [DOI] [Google Scholar]
  • 8.Karamchandani P; Johnson J; Yarwood G; Knipping E Implementation and application of sub-grid-scale plume treatment in the latest version of EPA’s third-generation air quality model, CMAQ 5.01 J. Air Waste Manage. Assoc 2014, 64 (4) 453–467 DOI: 10.1080/10962247.2013.855152 [DOI] [PubMed] [Google Scholar]
  • 9.Karamchandani P; Lohman K; Seigneur C Using a sub-grid scale modeling approach to simulate the transport and fate of toxic air pollutants Environ. Fluid Mech 2009, 9 (1) 59–71 DOI: 10.1007/s10652-008-9097-0 [DOI] [Google Scholar]
  • 10.Karamchandani P; Vijayaraghavan K; Chen S-Y; Seigneur C; Edgerton ES Plume-in-grid modeling for particulate matter Atmos. Environ 2006, 40 (38) 7280–7297 DOI: 10.1016/j.atmosenv.2006.06.033 [DOI] [Google Scholar]
  • 11.Karamchandani P; Vijayaraghavan K; Yarwood G Sub-grid scale plume modeling Atmosphere 2011, 2(3) 389–406 DOI: 10.3390/atmos2030389 [DOI] [Google Scholar]
  • 12.Kim Y; Seigneur C; Duclaux O Development of a plume-in-grid model for industrial point and volume sources: application to power plant and refinery sources in the Paris region Geosci. Model Dev 2014, 7 (2)569–585 DOI: 10.5194/gmd-7-569-2014 [DOI] [Google Scholar]
  • 13.Korsakissok I; Mallet V, Subgrid-scale treatment for major point sources in an Eulerian model: A sensitivity study on the European Tracer Experiment (ETEX) and Chernobyl cases. J. Geophys. Res 2010,115 DOI: 10.1029/2009JD012734 [DOI] [Google Scholar]
  • 14.Korsakissok I; Mallet V Development and application of a reactive plume-in-grid model: evaluation over Greater Paris. Atmos Chem. Phys 2010, 10 (18) 8917–8931 DOI: 10.5194/acp-10-8917-2010 [DOI] [Google Scholar]
  • 15.Woody M; Wong H-W; West J; Arunachalam S Multiscale predictions of aviation-attributable PM 2.5 for US airports modeled using CMAQ with plume-in-grid and an aircraft-specific 1-D emission model Atmos. Environ 2016, 147, 384–394 DOI: 10.1016/j.atmosenv.2016.10.016 [DOI] [Google Scholar]
  • 16.Karamchandani P; Zhang Y; Chen S-Y Development and initial application of a sub-grid scale plume treatment in a state-of-the-art online Multi-scale Air Quality and Weather Prediction Model Atmos. Environ 2012, 63, 125–134 DOI: 10.1016/j.atmosenv.2012.09.014 [DOI] [Google Scholar]
  • 17.Warneke C; Trainer M; de Gouw JA; Parrish DD; Fahey DW; Ravishankara AR;Middlebrook AM; Brock CA; Roberts JM; Brown SS; Neuman JA; Lerner BM; Lack D;Law D; Huebler G; Pollack I; Sjostedt S; Ryerson TB; Gilman JB; Liao J; Holloway J; Peischl J;Nowak JB; Aikin K; Min KE; Washenfelder RA; Graus MG; Richardson M; Markovic MZ;Wagner NL; Welti A; Veres PR; Edwards P; Schwarz JP; Gordon T; Dube WP; McKeen S;Brioude J; Ahmadov R; Bougiatioti A; Lin J; Nenes A; Wolfe GM; Hanisco TF; Lee BH;Lopez-Hilfiker FD; Thornton JA; Keutsch FN; Kaiser J; Mao J; Hatch C Instrumentation and Measurement Strategy for the NOAA SENEX Aircraft Campaign as Part of the Southeast Atmosphere Study 2013 Atmos. Meas. Technol. Discuss 2016, 2016, 1–39 DOI: 10.5194/amt-2015-388 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Xu L; Middlebrook AM; Liao J; Gouw JA; Guo H; Weber RJ; Nenes A; Lopez-Hilfiker FD; Lee BH; Thornton JA Enhanced formation of isoprene-derived organic aerosol in sulfur-rich power plant plumes during Southeast Nexus J. Geophys. Res 2016, 121 (18) 11137–11153 DOI: 10.1002/2016JD025156 [DOI] [Google Scholar]
  • 19.Sarwar G; Appel KW; Carlton AG; Mathur R; Schere K; Zhang R; Majeed MA Impact of a new condensed toluene mechanism on air quality model predictions in the US Geosci. Model Dev 2011, 4 (1)183–193 DOI: 10.5194/gmd-4-183-2011 [DOI] [Google Scholar]
  • 20.Fountoukis C; Nenes A ISORROPIA II: a computationally efficient thermodynamic equilibrium model for K+-Ca2+-Mg2+-Nh(4)(+)-Na+-SO42--NO3--Cl--H2O aerosols Atmos. Chem. Phys 2007, 7 (17) 4639–4659DOI: 10.5194/acp-7-4639-2007 [DOI] [Google Scholar]
  • 21.Carlton AG; Bhave PV; Napelenok SL; Edney EO; Sarwar G; Pinder RW; Pouliot GA;Houyoux M Treatment of secondary organic aerosol in CMAQv4.7 Environ. Sci. Technol 2010, 44 (22)8553–8560 DOI: 10.1021/es100636q [DOI] [PubMed] [Google Scholar]
  • 22.Sarwar G; Fahey K; Kwok R; Gilliam RC; Roselle SJ; Mathur R; Xue J; Yu J; Carter WPL Potential impacts of two SO2 oxidation pathways on regional sulfate concentrations: Aqueous-phase oxidation by NO2 and gas-phase oxidation by Stabilized Criegee Intermediates Atmos. Environ 2013, 68,186–197 DOI: 10.1016/j.atmosenv.2012.11.036 [DOI] [Google Scholar]
  • 23.U.S. Environmental Protection Agency.2011 National EmissionsInventory, version 1 Technical Support Document. (2014. http://www.epa.gov/ttn/chief/net/2011nei/2011_nei_tsdv1_draft2_june2014.pdf
  • 24.Bash JO; Baker KR; Beaver MR Evaluation of improved land use and canopy representation in BEIS v3. 61 with biogenic VOC measurements in California Geosci. Model Dev 2016) 9 (6) 2191 DOI: 10.5194/gmd-9-2191-2016 [DOI] [Google Scholar]
  • 25.Carlton AG; Baker KR Photochemical Modeling of the Ozark Isoprene Volcano: MEGAN, BEIS, and Their Impacts on Air Quality Predictions Environ. Sci. Technol 2011, 45 (10) 4438–4445 DOI: 10.1021/es200050x [DOI] [PubMed] [Google Scholar]
  • 26.Skamarock WC; Klemp JB; Dudhia J; Gill DO; Barker DM; Duda MG; Huang X; Wang W;Powers JG A description of the Advanced Reserch WRF version 3NCAR Technical Note NCAR/TN-475+STR. 2008.
  • 27.Baker KR; Misenis C; Obland MD; Ferrare RA; Scarino AJ; Kelly JT Evaluation of surface and upper air fine scale WRF meteorological modeling of the May and June 2010 CalNex period in California Atmos. Environ 2013, 80 (0) 299–309 DOI: 10.1016/j.atmosenv.2013.08.006 [DOI] [Google Scholar]
  • 28.Cohan DS; Hakami A; Hu Y; Russell AG Nonlinear response of ozone to emissions: Source apportionment and sensitivity analysis Environ. Sci. Technol 2005, 39 (17) 6739–6748 DOI: 10.1021/es048664m [DOI] [PubMed] [Google Scholar]
  • 29.Kwok R; Baker K; Napelenok S; Tonnesen G Photochemical grid model implementation of VOC, NO x, and O 3 source apportionment Geosci. Model Dev 2015, 8, 99–114 DOI: 10.5194/gmd-8-99-2015 [DOI] [Google Scholar]
  • 30.Kwok R; Napelenok S; Baker K Implementation and evaluation of PM2.5 source contribution analysis in a photochemical model Atmos. Environ 2013, 80, 398–407 DOI: 10.1016/j.atmosenv.2013.08.017 [DOI] [Google Scholar]
  • 31.Baker KR; Foley KM A nonlinear regression model estimating single source concentrations of primary and secondarily formed PM2.5 Atmos. Environ 2011, 45 (22) 3758–3767 DOI: 10.1016/j.atmosenv.2011.03.074 [DOI] [Google Scholar]
  • 32.Chowdhury B; Karamchandani P; Sykes R; Henn D; Knipping E Reactive puff model SCICHEM: Model enhancements and performance studies Atmos. Environ 2015, 117, 242–258 DOI: 10.1016/j.atmosenv.2015.07.012 [DOI] [Google Scholar]
  • 33.Bahreini R; Ervens B; Middlebrook A; Warneke C; De Gouw J; DeCarlo P; Jimenez J; Brock C;Neuman J; Ryerson T Organic aerosol formation in urban and industrial plumes near Houston and Dallas, Texas J. Geophys. Res 2009, 114 (D7) 2156–2202 DOI: 10.1029/2008JD011493 [DOI] [Google Scholar]
  • 34.Ryerson T; Buhr M; Frost G; Goldan P; Holloway J; Hübler G; Jobson B; Kuster W; McKeen S;Parrish D Emissions lifetimes and ozone formation in power plant plumes J. Geophys. Res 1998, 103(D17) 22569–22583 DOI: 10.1029/98JD01620 [DOI] [Google Scholar]
  • 35.Pollack IB; Lerner BM; Ryerson TB Evaluation of ultraviolet light-emitting diodes for detection of atmospheric NO2 by photolysis-chemiluminescence J. Atmos. Chem 2010, 65 (2–3) 111–125 DOI: 10.1007/s10874-011-9184-3 [DOI] [Google Scholar]
  • 36.Ryerson T; Huey L; Knapp K; Neuman J; Parrish D; Sueper D; Fehsenfeld F Design and initial characterization of an inlet for gas-phase NOy measurements from aircraft J. Geophys. Res 1999, 104 (D5)548–5492 DOI: 10.1029/1998JD100087 [DOI] [Google Scholar]
  • 37.Kelly JT; Baker KR; Nolte CG; Napelenok SL; Keene WC; Pszenny AA Simulating the phase partitioning of NH 3, HNO 3, and HCl with size-resolved particles over northern Colorado in winter Atmos. Environ 2016, 131, 67–77 DOI: 10.1016/j.atmosenv.2016.01.049 [DOI] [Google Scholar]
  • 38.Nolte C; Appel K; Kelly J; Bhave P; Fahey K; Collett J Jr; Zhang L; Young J, Evaluation of the Community Multiscale Air Quality (CMAQ) model v5. 0 against size-resolved measurements of inorganic particle composition across sites in North America, Geosci. Model Dev, 2015, 8, 2877–2892, doi: DOI: 10.5194/gmd-8-2877-2015. [DOI] [Google Scholar]
  • 39.Liu PS; Deng R; Smith KA; Williams LR; Jayne JT; Canagaratna MR; Moore K; Onasch TB;Worsnop DR; Deshler T Transmission efficiency of an aerodynamic focusing lens system: Comparison of model calculations and laboratory measurements for the Aerodyne Aerosol Mass Spectrometer Aerosol Sci. Technol 2007, 41 (8) 721–733 DOI: 10.1080/02786820701422278 [DOI] [Google Scholar]
  • 40.Pleim JE A combined local and nonlocal closure model for the atmospheric boundary layer. Part I: Model description and testing J. Appl. Met. Clim 2007, 46 (9) 1383–1395 DOI: 10.1175/JAM2539.1 [DOI] [Google Scholar]
  • 41.Woody MC; Baker KR; Hayes PL; Jimenez JL; Koo B; Pye HO Understanding sources of organic aerosol during CalNex-2010 using the CMAQ-VBS Atmos. Chem. Phys 2016, 16 (6) 4081–4100DOI: 10.5194/acp-16-4081-2016 [DOI] [Google Scholar]
  • 42.Hayes PL; Ortega AM; Cubison MJ; Froyd KD; Zhao Y; Cliff SS; Hu WW; Toohey DW;Flynn JH; Lefer BL; Grossberg N; Alvarez S; Rappenglück B; Taylor JW; Allan JD;Holloway JS; Gilman JB; Kuster WC; de Gouw JA; Massoli P; Zhang X; Liu J; Weber RJ;Corrigan AL; Russell LM; Isaacman G; Worton DR; Kreisberg NM; Goldstein AH; Thalman R;Waxman EM; Volkamer R; Lin YH; Surratt JD; Kleindienst TE; Offenberg JH; Dusanter S;Griffith S; Stevens PS; Brioude J; Angevine WM; Jimenez JL Organic aerosol composition and sources in Pasadena, California, during the 2010 CalNex campaign J. Geophys. Res 2013, 118 (16) 9233–9257 DOI: 10.1002/jgrd.50530 [DOI] [Google Scholar]
  • 43.Kleinman LI; Springston SR; Daum PH; Lee Y-N; Nunnermacker L; Senum G; Wang J;Weinstein-Lloyd J; Alexander M; Hubbe J The time evolution of aerosol composition over the Mexico City plateau Atmos. Chem. Phys 2008, 8 (6) 1559–1575 DOI: 10.5194/acp-8-1559-2008 [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Sup
NIHMS982359-supplement-Sup.pdf (4.4MB, pdf)

RESOURCES