Evaluation of the SO₂ and NO_X offset ratio method to account for secondary PM_2.5 formation

Abstract

The U.S. Environmental Protection Agency (EPA), state and local agencies have focused their efforts in assessing secondary fine particulate matter (aerodynamic diameter ≤2.5 µm; PM_2.5) formation in prevention of significant deterioration (PSD) air dispersion modeling. The National Association of Clean Air Agencies (NACAA) developed a method to account for secondary PM_2.5 formation by using sulfur dioxide (SO₂) and nitrogen oxides (NO_x) offset ratios. These ratios are used to estimate the secondary formation of sulfate and nitrate PM_2.5. These ratios were first introduced by the EPA for nonattainment areas in the Implementation of the New Source Review (NSR) Program for Particulate Matter Less than 2.5 Micrometers (PM_2.5), 73 FR 28321, to offset emission increases of direct PM_2.5 emissions with reductions of PM_2.5 precursors and vice versa. Some regulatory agencies such as the Minnesota Pollution Control Agency (MPCA) have developed area-specific offset ratios for SO₂ and NO_x based on Comprehensive Air Quality Model with Extensions (CAMx) evaluations for air dispersion modeling analyses. The current study evaluates the effect on American Meteorological Society/Environmental Protection Agency Regulatory Model (AERMOD) predicted concentrations from the use of EPA and MPCA developed ratios. The study assesses the effect of these ratios on an electric generating utility (EGU), taconite mine, food processing plant, and a pulp and paper mill. The inputs used for these four scenarios are based on common stack parameters and emissions based on available data. The effect of background concentrations also evaluates these scenarios by presenting results based on uniform annual PM_2.5 background values. This evaluation study helps assess the viability of the offset ratio method developed by NACAA in estimating primary and secondary PM_2.5 concentrations. An alternative Tier 2 approach to combine modeled and monitored concentrations is also presented.

Implications:

On January 4, 2012, the EPA committed to engage in rulemaking to evaluate updates to the Guideline on Air Quality Models (Appendix W of 40 CFR 51) and, as appropriate, incorporate new analytical techniques or models for secondary PM_2.5. As a result, the National Association of Clean Air Agencies (NACAA) developed a screening method involving offset ratios to account for secondary PM_2.5 formation. The use of this method is promising to evaluate total (direct and indirect) PM_2.5 impacts for permitting purposes. Therefore, the evaluation of this method is important to determine its viability for widespread use.

Introduction

The U.S. Environmental Protection Agency's (EPA) preferred air dispersion model, the American Meteorological Society/Environmental Protection Agency Regulatory Model (AERMOD), does not consider atmospheric chemistry to account for the secondary formation of fine particulate matter (aerodynamic diameter ≤2.5 μm; PM_2.5) emissions. For this reason, the EPA along with state and local agencies have focused their efforts in assessing secondary fine particulate matter (PM_2.5) formation for air dispersion modeling. The National Association of Clean Air Agencies (NACAA) developed a method to account for secondary PM_2.5 formation by using sulfur dioxide (SO₂) and nitrogen oxides (NO_x) offset ratios to estimate the secondary formation of sulfate and nitrate PM_2.5 (NACAA, 2011). These ratios were first introduced by the EPA for nonattainment areas to offset emission increases of direct PM_2.5 emissions with reductions of PM_2.5 precursors and vice versa (73 FR 28321; EPA, 2008). However, on July 21, 2011, the EPA changed their position and established that these offset ratios were no longer considered presumptively approvable but must be subject to a technical demonstration (McCarthy, 2011).

On March 4, 2013, the EPA released the Draft Guidance for PM_2.5 Permit Modeling to the public for consideration, review, and comment (EPA, 2013). Section III.2.2 from this draft guidance deals with the hybrid approach involving a qualitative and quantitative assessment of secondary PM_2.5 formation. The pollutant offset ratio method developed by NACAA (2011) is mentioned as a viable option to provide the quantitative portion of a hybrid assessment of secondary PM_2.5 formation. This method uses area-specific offset ratios for PM_2.5, permitting converting emissions of precursors into equivalent amounts of direct PM_2.5 emissions. NACAA recommended a four-tiered approach for conducting air quality analyses for comparison with National Ambient Air Quality Standards (NAAQS), Significant Impact Levels (SILs), and prevention of significant deterioration (PSD) increments (NACAA, 2011). Tier I relates to SIL modeling demonstrations where AERMOD is used with region- or state-specific offset ratios to account for primary and secondary PM_2.5. Tiers II, III, and IV relate to NAAQS and PSD increment cumulative analyses. Tier II relates to the use of AERMOD with region- or state-specific offset ratios for primary and secondary PM_2.5. Tier III proposes the use of AERMOD for primary PM_2.5 and the use of a chemistry plume model (i.e., Second-order Closure Integrated puff model with CHEMistry [SCICHEM]) to estimate secondary PM_2.5 formation. Tier IV proposes the use of AERMOD for primary PM_2.5 and chemical models such as Comprehensive Air Quality Model with Extensions (CAMx) or the Community Multiscale Air Quality (CMAQ) modeling system for secondary PM_2.5.

The Minnesota Pollution Control Agency (MPCA) favors the use of NACAA's Tiers I and II based on Minnesota-specific offset ratios for SO₂ and NO_x (McCourtney, 2012)^. These offset ratios originate from evaluations performed by MPCA modelers using CAMx. Based on these evaluations, the secondary PM_2.5 emission rate is defined as the sum of the SO₂ emission rate divided by 10 and the NO_x emission rate divided by 100. Thus, the secondary PM_2.5 formation from SO₂ and NO_x is 1:10 and 1:100, respectively. MPCA's draft Modeling Guidance defines the total equivalent PM_2.5 emission rate to be the sum of the primary and secondary PM_2.5 emission rates (MPCA, 2013). The total equivalent emission rate is to be used in AERMOD modeling demonstrations to show compliance with the PM_2.5 NAAQS.

The offset ratio method was developed to account for total (primary and secondary) PM_2.5 impacts in a simplified manner. It is convenient because its review does not require knowledge or expertise on the use of complex photochemical grid models. However, the selection of these ratios seems to be arbitrary. The EPA ratios are based on the 75th percentile distribution for NO_x and on the 90th percentile distribution for SO₂ (Fox, 2007). On the other hand, the Minnesota-specific offset ratios were set by a visual analysis of the 98th percentile gas to particle-phase ratios obtained from CAMx. These data were plotted with the y-axis depicting the ratio of precursor gas to secondary PM_2.5 particles. These ratios are indicative of the conversion of SO₂ and NO_x to secondary PM_2.5. A high ratio means that most of the PM_2.5 precursors are in the gaseous phase and the opposite is true for a low ratio. The x-axis is the distance from the source where each distance value has a box plot showing the distribution of the 98th percentile ratios at each distance. In these plots, it is evident that there is a negligible conversion of SO₂ and NO_x to secondary PM_2.5 in the close proximity of the source since the gas-to-particle ratio is very high. However, this conversion becomes more noticeable as the distance from the source increases since the gas-to-particle ratios become smaller. The ratios ascertained from each case were based on the smallest value (lower whisker) of the box plot that happened at a distance of 20 km from the source. There was some variability on the cases presented by MPCA. For example, the ratios ascertained from the cases presented ranged from 10:1 to 20:1 for SO₂ and from 77:1 to 10,000:1 for NO_x. However, the Minnesota-specific offset ratios were set at 10:1 for SO₂ and 100:1 for NO₂ (McCourtney, 2012).

The PM_2.5 offset ratio method is useful in accounting for total equivalent PM_2.5 emissions due to its relative simplicity that does not require the use of a photochemical grid model. However, this method is not without flaws, since offset ratios have been shown to vary by distance, season, grid resolution, stack height, and emission rate (Boylan and Kim, 2012). With an already stringent 24-hr PM_2.5 NAAQS standard and often high background concentrations, it is most challenging for many industries to meet this standard when considering direct PM_2.5 alone due in part to the current conservatism in modeling practices. In addition, primary particulates are at a maximum at the point of release from the source. Therefore, it is assumed that the highest impact will be at the receptor that is in the location where the plume from the source intersects with the ground, which is likely closer to the source. However, secondary particulates can take hours or days to form; therefore, the maximum impact will likely be farther from the source. Furthermore, the modeled uncertainties associated with CAMx are much larger than the modeled uncertainties associated with AERMOD. EPA guidance states that absolute model output from photochemical grid-based models should not be used directly in evaluating impacts (EPA, 2007). Therefore, comparisons between concentrations obtained with CAMx and AERMOD should be viewed with caution. Also, the use of background concentrations of PM_2.5 in the model accounts for secondary fine particles and contributions from nearby sources. Thus, when significant amounts of PM_2.5 precursors are emitted, the offset ratio method has the potential to double count these emissions. Likewise, primary fine particles from background sources may be double counted since a modeling analysis commonly includes both nearby sources and background concentrations. This situation increases the uncertainty of the results obtained in AERMOD with this method.

The current study evaluates the effects of using PM_2.5 offset ratios on predicted AERMOD concentrations for various industries. The industries evaluated include an electric generating utility (EGU), a taconite mine, a food processing plant, and a pulp and paper mill. The inputs used for these four scenarios are based on typical stack parameters and emission rates for that specific industry. This evaluation study seeks to help in assessing the viability of the offset ratio method developed by NACAA in estimating total PM_2.5 concentrations.

Experimental Methods

The current study evaluates the use of constant offset ratios to account for secondary PM_2.5 in the following types of industries: EGU, taconite mine, food processing facility, and pulp and paper mill. The EGU case is based on an older coal-fueled boiler with a capacity of 270 MMBtu/hr controlled by an electrostatic precipitator (ESP) and a high-efficiency cyclonic collector. The taconite mine example is based on a new natural-gas-fueled indurating furnace of 540 MMBtu/hr that exhaust also flues from other iron ore processes. The controls in this furnace include low-NO_x burners (LNB), baghouse (PM/PM₁₀/PM_2.5 control), and a dry scrubber (SO₂ control). The food processing facility example is based on a new 200 MMBtu/hr boiler permitted burning natural gas with fuel oil as backup. However, the emissions from this boiler are based on a boiler burning gas 1 fuels. Based on the gas 1 fuel definition under 40 CFR Part 63 Subpart DDDDD, liquid fuels (i.e., fuel oil) can only be burned during periods of gas curtailment or gas supply interruptions. The control for this boiler includes good combustion practices (GCP). The last example is an older 250 MMBtu/hr boiler at a pulp and paper mill able to combust natural gas, fuel oil, and wood. The controls on this boiler include an ESP and a cyclone. Also, this boiler is permitted to combust noncondensable gases (NGCs) that produce very high SO₂ emissions.

One representative stack was considered in each analysis with stack parameters based on actual permitted facilities (Table 1). AERMOD version 12345 was used in this evaluation of the different cases and scenarios.

Table 1. Input parameters for four source types

Input parameter		Case 1	Case 2	Case 3	Case 4
Facility		Electric generating utility	Taconite mine	Food processing facility	Pulp and paper mill
Emission source		Boiler	Indurating furnace	Boiler	Boiler
Capacity (MMBtu/hr)		270	540	200	250
Fuel(s)		Coal	Natural gas	Natural gas/fuel oil (backup)	Natural gas/fuel oil/wood
Controls		ESP/cyclone	Baghouse/dry scrubber/LNB	GCP	ESP/cyclone
Stack height (m)		60	100	50	75
Emission rate (g/sec)	PM2.5	0.1	7	0.5	2.5
	NOx	40	4.5	3.5	10
	SO2	80	7	1.0	150
Exit temperature (K)		340	320	427	450
Diameter (m)		2.5	5	1.2	1.8
Exit velocity (m/sec)		22	15	15	13

The same baseline meteorological data were used for all cases. This included AERMET (version 11059) processed 5-yr surface data for the Minneapolis/St. Paul Airport (WBAN no. 14922) and upper air data from Minneapolis (UA ID 94983) for 2006–2010. Flat terrain was assumed for all runs, thus, no terrain data was used in this evaluation. The same polar receptor grid was used for all runs with a fenceline located 100 m from the source.

Initially, each case was run with a unitized (1 g/sec) PM_2.5 emission rate that was scaled to the primary and total equivalent (primary and secondary) PM_2.5 emission rates. The total equivalent emission rate was calculated based on MPCA's recommendation by adding the emission rate of primary PM_2.5 (noncondensable and condensable) to the NO_x emission rate divided by 100, and to the SO₂ emission rate divided by 10. Then, the same runs were repeated including building effects from a building of 40 m in the x, y, and z dimensions. Building effects have been ignored in preliminary assessment analyses of the offset ratio method. Thus, the current study seeks to highlight the relevance of these effects on predicted concentrations. Each case includes the following results:

1.	Primary PM2.5, no building effects
2.	Total equivalent PM2.5, no building effects
3.	Primary PM2.5 with building effects
4.	Total equivalent PM2.5 with building effects

All modeling analyses were completed assuming a background of 25 μg/m³, which is the 98th percentile background concentration of 24-hr PM_2.5 averages for years 2010-2012 at the Shakopee Monitor 505 (Gavin and McMahon, 2013). This background value is meant to account for nearby sources, long-range transport of emissions, and natural background levels of fine particulates (Gavin and McMahon, 2013). This is a relatively common background concentration, since some urban and industrial areas can experience concentrations very close to the 24-hr NAAQS. The results obtained from each of the cases were compared with the 24-hr PM_2.5 NAAQS of 35 μg/m³.

Results and Discussion

The modeling results for the four cases evaluated are summarized below (Table 2). The maximum primary PM_2.5 concentrations are based on the emission rate of primary PM_2.5 (noncondensable and condensable). The maximum total PM_2.5 concentrations are based on an emission rate that includes primary PM_2.5 plus NO_x emission rate divided by 100 and SO₂ emission rate divided by 10. The cases are simplistic in nature, since a facility would have more than one source of PM_2.5 and PM_2.5 precursors. However, the main intent is to evaluate the effect of modeling total PM_2.5 concentrations by using the offset ratio method as described in NACAA's Tiers I and II. This is the preferred method for regulatory agencies because it does not require the use of complex photochemical grid models.

Table 2. Results of primary and total (primary and secondary) 24-hr PM_2.5 concentrations

Case	Building Effects Included?	Predicted Impact from Primary Emissions(μg/m^3)	Predicted Impact from Secondary Emissions(μg/m^3)	Total Equivalent PM2.5 (μg/m^3)	Background Concentration(μg/m^3)	Total Predicted Impact PM2.5 (μg/m^3)	Primary PM2.5 (% Total Predicted)	Secondary PM2.5 (% Total Predicted)	Background Concentration(% Total Predicted)
Case 1	Yes	0.20	16.72	16.92	25	41.92	0.48%	39.39%	59.64%
(EGU)	No	0.06	4.86	4.92	25	29.92	0.20%	16.24%	83.56%
Case 2	Yes	1.75	0.19	1.94	25	26.94	6.50%	0.71%	92.80%
(Taconite mine)	No	1.75	0.19	1.94	25	26.94	6.50%	0.71%	92.80%
Case 3	Yes	6.11	1.65	7.76	25	32.76	18.65%	5.04%	76.31%
(Food processing plant)	No	0.60	0.16	0.76	25	25.76	2.33%	0.62%	97.05%
Case 4	Yes	2.71	16.38	19.09	25	44.09	6.15%	37.15%	56.70%
(Pulp and paper mill)	No	1.30	7.87	9.18	25	34.18	3.80%	23.03%	73.14%

The inclusion of building effects has the greatest effect on the food processing facility (Case 3) by almost an order of magnitude due to its shorter stack. A negligible effect was observed for the taconite mine (Case 2), which has the tallest stack. The EGU (Case 1) and pulp and paper mill (Case 4) had also noticeable effects from the inclusion of a building structure. This is relevant because some of the preliminary assessments of the offset ratio method were performed without considering the effect from building structures.

The effects from using the offset ratio method to account for secondary formation are significant in Cases 1 and 4 when compared with the impacts from using only direct PM_2.5 emission rates. These simplified examples show that the use of this method for a facility with one PM_2.5 source can increase the predicted emission concentrations significantly. The cases evaluated also suggest that additional emission controls, if available, may be necessary to achieve modeled concentrations below the PM_2.5 NAAQS. This may be an issue for facilities with large amount of condensables for which controls are either nonexistent or prohibitively expensive. Additionally, a real facility will most likely have to account for additional emission sources, potentially higher background concentrations, road traffic emissions, and nearby sources. Thus, facilities that have modeled compliance with the PM_2.5 NAAQS in the past may not show compliance if required to account for secondary PM_2.5 formation by means of the offset ratio method described herein. Additionally, some facilities that currently have trouble meeting the PM_2.5 NAAQS will have an even harder time if required to use the offset ratio method.

In contrast, under certain scenarios (i.e., Cases 2 and 3), the offset ratio method may yield a trivial contribution from PM_2.5 precursors. In these cases, the offset ratio method can be used to satisfy the qualitative portion of the hybrid assessment proposed in EPA's Draft Guidance for PM_2.5 Permit Modeling (2013). Therefore, in spite of its conservatism, the offset ratio method is a screening tool that may help certain applicants in ascertaining that total PM_2.5 impacts are below the NAAQS.

Combining Modeled Results and Background Concentrations

An important element in a NAAQS modeling demonstration is the selection of background concentrations that are added to the modeled design value from AERMOD. The probabilistic nature of the new standards allow for alternate methods to combine monitored and predicted concentrations. For example, the 1-hr NO₂ NAAQS was promulgated as the 98th percentile of maximum daily concentrations and the 24-hr PM_2.5 NAAQS was promulgated as the 98th percentile of daily concentrations. Thus, the probability of these standards is 1.00 − 0.98 = 0.02. This is equivalent to 1 exceedance every 50 days (1/50 = 0.02). When we extrapolate this ratio to the number of days in a year (365), we get 7.3 exceedances in a year, which is rounded up to the eighth highest value in a year (the form of the standard). However, the EPA (2010) had recommended that to demonstrate compliance with the 24-hr PM_2.5 NAAQS the high-first-high modeled value be used as the modeled design value instead of the high-eighth-high. This policy was such, in part, to compensate for secondary PM_2.5 concentrations. However, by assuming that the 98th percentile modeled concentration is combined with the 98th percentile background concentration, the probability equals 0.0004 or (0.02) × (0.02). This is equivalent to the 99.96th percentile or one exceedance every 2500 days (1/2500 = .0004). The probabilistic inappropriateness of such an approach has been described previously (Murray and Newman, 2013; Newman and Murray, 2013). Furthermore, this degree of conservatism is well beyond the level necessary to protect the NAAQS.

Another factor of conservatism in dispersion modeling relates to determining a representative background concentration to be used in a modeling analysis. Nicholson (2013) described a screening technique to obtain a representative background concentration by analyzing hourly PM_2.5 monitored data from the Santa Fe, New Mexico, airport monitoring site. This approach included the analysis of monitored data in time-series plots and histograms to identify major exceptional events in the area, including forest fires and dust storms. Secondly, the analysis found that high monitored concentrations were also observed at times when winds were downwind from large emission sources. Initially, the histogram for the raw data showed a positively skewed distribution with a long right tail due to few, yet very high monitored observations. The approximate 98th percentile value of the raw data was 18 μg/m³. To determine a truly representative background, Nicholson screened out monitoring observations from exceptional events and occurrences when the monitor was downwind of a major emission source. It is common that the effect from major emission sources is double counted, since major nearby sources are modeled explicitly in a NAAQS modeling analysis. Therefore, double counting can be avoided by screening out the observations impacted by these nearby sources. As these exceptional events were screened out of the distribution, the long tail of the distribution decreased significantly. This supported the notion that these were outlier observations that skewed the distribution and were not representative of background concentrations. Furthermore, the 98th percentile from the adjusted distribution was about 6 μg/m³ after screening out exceptional events. This is about one third of the 98th percentile value from the original unscreened distribution. Nicholson concluded his analysis by cautioning against the use of background concentrations based upon extreme values, since they are not representative of the background in a dispersion modeling domain.

An alternate more realistic approach in NAAQS dispersion modeling analyses is to combine the 98th percentile predicted concentration with the 50th percentile background concentration. This approach conserves the use of the modeled 98th percentile value from AERMOD and allows for a more representative background level by selecting the median instead of the tail of the distribution. A case study identifying the conservatism of using the tail of the distribution (99th or 98th percentile) has been discussed by Nicholson (2013). Additionally, this approach will still be protective of the NAAQS because it results in 0.01 or (0.02) × (0 .50). This is equivalent to the 99th percentile combined concentration, which is more conservative than the form of the standard (98th percentile). Similarly, for the 1-hr SO₂ NAAQS, this approach would result in 0.05 or (0.01) × (0 .50). This is equivalent to the 99.5th percentile concentration, which is also more conservative than the form of the standard (99th percentile). This method is statistically sound to be used as a Tier 2 approach, since it still provides a reasonable level of conservatism that would ensure the protection of the NAAQS.

This approach was carried out in the current study by evaluating the PM_2.5 monitored observations for the years of 2010 to 2012 from the Shakopee monitor (MPCA ID 505). This monitor collects 24-hr PM_2.5 samples every 3 days and the monitored design value is defined as the 98th percentile 3-yr average from these observations. A histogram and a box plot showing the percentiles of these data are presented in Figure 1. This figure depicts that the 98th percentile is 24.8 μg/m³, which is already 71% of the 24-hr PM_2.5 NAAQS. However, the 90th percentile diminishes to 15.4 μg/m³, which is already 44% of the 24-hr PM_2.5 NAAQS. This means that these high concentration values are, in all likelihood, caused by extreme events. Thus, using them as background values in a compliance modeling demonstration is overly conservative. As mentioned previously, the authors recommend using the 50th percentile background value, which in this case is 7.5 μg/m³. The combination of this value with the 98th percentile modeled concentration results in the 99th percentile combined concentration, which is more conservative than the form of the standard.

Figure 1. Histogram and box plot for observation data from the Shakopee, Minnesota, monitoring station for the years of 2008–2012.

Table 3 shows a comparison of the results from combining the modeled concentrations for each case with the 98th and 50th percentile background values. The reduction in total concentrations (predicted and background) from using the 50th percentile background distribution for the four cases presented varies from 40% to 68%. Understandably, facilities with low emissions would see the greatest reduction in total concentrations (Case 2 and Case 3 with downwash). However, other facilities with larger emissions would also experience significant reductions in their total impact by using the 50th percentile monitored concentration. This approach is statistically sound and protective of the NAAQS.

Table 3. Comparison of total concentrations (modeled and background) based on the 98th and 50th percentiles

Case	Building Effects Included?	Total Equivalent PM2.5 (μg/m^3)	Background Concentration(μg/m^3)	Total Predicted Impact PM2.5 (μg/m^3)	Background Concentration(μg/m^3)	Total Predicted Impact PM2.5 (μg/m^3)	Percent Reduction in Total PredictedImpact PM2.5
Case 1	Yes	16.92	25	41.92	7.5	24.42	41.8%
(EGU)	No	4.92	25	29.92	7.5	12.42	58.5%
Case 2	Yes	1.94	25	26.94	7.5	9.44	65.0%
(Taconite mine)	No	1.94	25	26.94	7.5	9.44	65.0%
Case 3	Yes	7.76	25	32.76	7.5	15.26	53.4%
(Food processing plant)	No	0.76	25	25.76	7.5	8.26	67.9%
Case 4	Yes	19.09	25	44.09	7.5	26.59	39.7%
(Pulp and paper mill)	No	9.18	25	34.18	7.5	16.68	51.2%

Summary and Recommendations

The benefit from using the offset ratio method is that it conservatively accounts for secondary PM_2.5 formation without the need of using complex chemical models. Once region- or state-specific ratios have been developed or justified by a regulatory agency, these ratios can be used to adjust the total PM_2.5 emission rate used in AERMOD. Thus, this approach is simple for applicants to perform and for regulatory agencies to review.

However, some of the assumptions of this method are overly conservative. First, the offset ratio method is conservative because it assumes that both primary and secondary PM_2.5 are emitted from the stack when in reality secondary formation happens gradually over time and space. Furthermore, another level of uncertainty relates to the CAMx generated ratios that vary by season, stack height, emission rate, and grid resolution. These elements of conservatism get compounded by other common assumptions related to the use of maximum emission rates, uniform background concentrations, inclusion of intermittent equipment, and AERMOD's treatment of low-speed winds. Together, these conservative assumptions make NAAQS and increment consumption compliance demonstrations more challenging while at the same time increasing the uncertainty of the predicted results.

In spite of being conservative, this technique may be a useful screening method to account for secondary PM_2.5 formation to satisfy the quantitative portion of a hybrid assessment. Therefore, in spite of its conservatism, in some cases this method may complement the qualitative assessment to satisfy the requirements spelled out in EPA's Draft Guidance for PM_2.5 Permit Modeling.

Also, recent shifts in EPA's guidance suggest that by using techniques such as the offset ratio method to account for secondary PM_2.5 formation, an applicant could be allowed to use the true form of the 24-hr PM_2.5 standard (i.e., the 98th percentile predicted concentration) instead of the high-first-high concentration as previously directed by EPA (2010). This change in policy is considered in EPA's Draft Guidance for PM_2.5 Permit Modeling (2013) and some regulatory agencies such as the MPCA have started allowing it already (MPCA, 2013).

The facilities that may benefit the most from using the offset ratio method are those with inherently low emissions of PM_2.5 precursors of SO₂ and NO_x. Therefore, it is likely that newer facilities may benefit from this method, since they would be subject to the new 1-hr NO₂ and SO₂ standards that would require them to limit these PM_2.5 precursor emissions (Case 3). Also, facilities with very tall stacks (i.e., 100 m) may benefit from this approach in situations where emissions of PM_2.5 precursors are moderate (Case 2), since dispersion of the additional portion of PM_2.5 precursors can be accommodated by the enhanced dispersion. Furthermore, facilities with boilers subject to 40 CFR Part 63 Subpart DDDDD may elect to use natural gas fuel as their primary fuel in order to avoid some of the additional emission limits from using a non–gas 1 fuel. This would result in lower PM_2.5 precursors from SO₂, since the emission rate of natural gas is significantly lower than the one for low-sulfur fuel oil.

In general, facilities with high emissions of SO₂ and NO₂ will not benefit from this method, since the equivalent PM_2.5 emissions from precursors may be significant (Cases 1 and 4). Also, older facilities that have not been required to model compliance with the new short-term standards (1-hr and 24-hr) will find it more difficult to benefit from this method. These facilities may need to limit their PM_2.5, SO₂, and NO_x emissions so they can meet the short-term standards, including the 24-hr PM_2.5 NAAQS. This is a cumbersome task that may be achieved by a combination of mitigation strategies, including the installation of emission controls, extension of stack heights, and fuel and operational limitations. However, it is possible that the expense necessary to make these changes may be unfeasible and a facility could be forced to shut down.

Some other facilities may elect to account for secondary PM_2.5 formation by means of photochemical modeling (i.e., CAMx, CMAQ). However, EPA has not established a regulatory default photochemical model so a screening-level analysis analogous to the screening nature of Section 5.2.4 of Appendix W for NO₂ would be necessary. Furthermore, there is no established methodology or protocol suitable for the assessment of photochemical models and it is not clear if regulatory agencies have the expertise to review such analyses.

A critical recommendation in addressing the unreasonable level of conservatism in NAAQS compliance modeling of the new probabilistic standards relates to how background concentrations are combined with dispersion modeling data. Simply adding the 98th percentile predicted concentration with the 98th monitored concentration does not result in the 98th percentile of the combined total concentration. Instead of combining the 98th percentile modeled concentration with the 98th percentile background concentration, an alternate Tier 2 approach is to combine the 98th percentile predicted concentration with the 50th percentile background concentration. This recommendation provides a reasonable level of conservatism and will still be protective of the NAAQS because it results in a 99th percentile combined concentration.

Funding

The authors thank the China Section of the Air & Waste Management Association for the generous scholarship funds they received to cover publication costs of page charges and make the publication of this paper possible.