We now find ourselves in a unique regulatory period in which the Environmental Protection Agency (EPA) is responsible for promoting the use of modeling procedures to fit the form of the 1-hour probabilistic National Ambient Air Quality Standards (NAAQS) for NO2 and for SO2. Associated with these pollutants and standards are important modeling issues that are either not currently addressed in the Guideline on Air Quality Models1, have been addressed through the Support Center for Regulatory Air Models (SCRAM) Clearinghouse memoranda, or are on the table within the EPA-industrial-academia realm of review and consideration. Of paramount concern since the 2010 promulgation of the new NAAQS has been the fact that past modeling techniques and practices were often conservative. They considered the combined effects of worst case emissions, worst case meteorology, and worst case receptors all within the context of atmospheric physics in models that err toward conservative estimates of the modeled design concentration. Many of these measures were used to simplify the analyses absent more robust, scientifically sound information about the emissions sources and pertinent atmospheric conditions. With the advent of the more stringent and short-term 1-hour air quality standards, the use of redundant, overly conservative assumptions is no longer appropriate.
Dispersion modeling plays a central role in the U.S. air quality management regulatory program. It is used in assessing air quality impacts for federal, regional, state and local permitting and relies on EPA models and guidance to do so. Currently, permit applications require the use of model methodologies that are yet evolving as the science improves, while EPA simultaneously uses those models to complete impact studies in support of regulatory programs. This makes for a difficult situation if one is modeling a facility to obtain a new permit or being evaluated as part of a State Implementation Plan (SIP).
Associated with these pollutants and standards are important modeling issues not currently addressed in EPA’s Guideline on Air Quality Models or for which modeling guidance has changed over the past few years because models like AERMOD and CALPUFF do not handle probabilistic standards effectively. Such issues include policy decisions relating to attainment designations, monitoring network challenges, and emissions quantification issues associated with a hybrid monitoring/modeling approach to assessing attainment of the NAAQS. With the state of today’s models it is important to recognize strategies for anticipated modeling study practices, timelines, practical examples of mitigation measures to achieve compliance, regulatory issues in compliance, and possible roadblocks to meeting the 1-hour NO2 and SO2 NAAQS. From a technical perspective, specific model issues that must be addressed include low wind speeds, NOX to NO2 in stack ratios and conversion rates, consideration of building downwash, Good Engineering Practice stack heights, and maximum emissions versus more representative emission profiles. These combined permitting and modeling stringencies as well as evolving model physics to better represent atmospheric transport and dispersion can affect facilities both directly (i.e., when permitting a facility) and indirectly (when included as a nearby source in a multi-source SIP modeling analysis). Conundrums and successful strategies toward meeting these new NAAQS are offered herein.
The National Ambient Air Quality Standards (NAAQS) for NO2 and SO2 promulgated in 2010 are probabilistic in form as follows:
- 1-hr NO2 NAAQS of 100 ppb, based on 98th-percentile of the annual distribution of daily maximum 1-hr values
- 1-hr SO2 NAAQS of 75 ppb, based on 99th-percentile of the annual distribution of daily maximum 1-hr values
These 1-hour standards are quite stringent compared to the previous NAAQS for these pollutants, which had an annual averaging period for NO2 and 3-hour, 24-hour and annual averaging periods for SO2. Dispersion modeling historically has been used to assess the potential for a source or group of sources to comply with the standards. Continuing that practice, the regulated community along with state and local regulators are using models and methodologies that may be questionable for demonstrating compliance with 1-hour NAAQS and from which permit terms and conditions are based. Examples may be found in many industries from power production to steel manufacturing to refineries.
To assist regulated entities and regulatory agencies in demonstrating compliance with these standards through dispersion modeling, EPA recently published the following guidance documents:
- Applicability of Appendix W Modeling Guidance for the 1-hour NO2 National Ambient Air Quality Standard, June 28, 20102
- Applicability of Appendix W Modeling Guidance for the 1-hour SO2 National Ambient Air Quality Standard, August 23, 20103
- Additional Clarification Regarding Application of Appendix W Modeling Guidance for the 1-hour NO2 National Ambient Air Quality Standard, March 1, 20114
Thus, while EPA considers final modeling guidelines and while models are in a state of technical flux, sources must utilize an evolving compilation of methodologies. Regulated entities can apply “creative methods” that are deemed acceptable and not are considered “gaming” the modeling results within the bounds of reasonable and representative source characterizations, following EPA and state guidance, and demonstrating environmental compliance.
The focus herein is to determine avenues to model compliance with the NAAQS by following the procedures and guidance of the regulatory agencies, by being aware of the potential inaccuracies and conservative features of the available models, and being innovative in terms of conventional and unconventional measures to demonstrate modeled compliance. When model performance improves – some of these strategies may yet be of interest and use.
Discussions during recent conferences such as EPA’s 10th Modeling Conference, Regional/State/Local modeling meetings, and modeling stakeholder meetings as well as specialty sessions at national conferences of the Air & Waste Management Association (A&WMA) and the Council of Industrial Boiler Owners (CIBO), and other supported research by the Electric Power Research Institute (EPRI) and the American Petroleum Institute (API) imply that current regulatory models warrant some critical scrutiny. A summary of modeling issues and limitations of AERMOD were presented recently.5 These include:
- Over-estimates in low wind speeds – AERMOD predicts the highest concentrations for low level sources (point and fugitive) during low wind speed conditions (less than 1 m/s). The accuracy of AERMOD was not evaluated for a significant number of low wind speed cases and recent studies have shown an “over prediction tendency for AERMOD in light wind/stable conditions.”
- Poor AERMOD performance using permit allowable emissions – AERMOD was evaluated using actual emissions for sources in past research studies. Matching maximum permit allowable emissions to worst case meteorology is unrealistic. The use of representative emissions would be more in concert with a realistic modeling scenario. There is also concern that modeling emissions from intermittent sources such as emergency generators and fire pumps will over predict concentrations, although EPA recognizes that these sources often operate infrequently and allows some flexibility in addressing their operations in AERMOD.
- Good Engineering Practice (GEP) stack heights – for the past 25+ years, sources have used the concept of GEP stack height as a guiding precept for designing and building stacks that would eliminate building and structure downwash effects which can cause higher concentrations near the stacks. In 2011, the WAKFLG option in AERMOD was disabled thereby allowing the application of downwash to GEP and higher stacks which previously may have been modeled as meeting the NAAQS. As a result, new modeling could predict higher ambient impacts for sources that have been unchanged for years or decades.
- NOX conversion to NO2 - while EPA previously allowed a tiered approach for atmospheric conversion from NOX to NO2, it was generally not used for the annual average NAAQS. Under the more stringent NO2 1-hour NAAQS, the use of the tiered approach (Tiers 2 and 3) is more common but the applicability and use are varied from source to source and region to region. In-stack ratios of NO2/ NOX, conversion rates in the atmosphere, and background ambient ozone levels which control such conversion are all required to use Tier 2 and 3 methods.
- Poor AERMOD performance for ground level fugitive emissions – while not generally an SO2 or NO2 issue, AERMOD’s poor performance for fugitive emissions can nonetheless contribute to the mix of confounding AERMOD evaluation problems.
- Combined low wind speed and building downwash issues – not only is AERMOD predicting higher concentrations with low winds and when the GEP cutoff for downwash effects (WAKFLG) has been turned off, the combination of low winds and downwash may be exacerbating each of these individual problems. Downwash is generally weaker in low wind speed cases whereas plume rise is expected to be higher thereby resulting in less downwash. Contrary to expectations, however, AERMOD sometimes predicts peak concentrations under these conditions.
Couple all of these apparent conservative performance issues (i.e., the model over predicts due to each one of these items) with the form and level of the new 1-hour SO2 and NO2 NAAQS, and modeling compliant scenarios for existing sources as well as new or modified sources becomes extremely difficult.
Measures to Meet the 1-Hour NAAQS
Meeting the ambient 1-hour SO2 and NO2 NAAQS has become a challenge for sources that are high emitters or high potential emitters. This includes planned sources as well as existing sources that could be pulled into modeling requirements because of other new sources nearby or because they are in non-attainment counties or areas or within 50km of such areas. So what are some conventional and even non-conventional ways to achieve compliant modeling scenarios and allow sources to meet the short term SO2 and NO2 NAAQS? The following section offers several measures that may help sources in demonstrating compliance with the NAAQS.
Over the past two years a number of factors have come to light that require additional consideration when conducting air dispersion modeling: 1) challenges in demonstrating compliance with the 1-hour SO2 and NO2 1-hour NAAQS has become apparent; 2) facility managers and environmental staff have had to contend with the realities of meeting such NAAQS even when in a status quo mode of operation, and 3) the issues concerning model performance have surfaced as AERMOD is now being used to provide estimates at the level of the time step (1-hour) where the model does not perform well. The following are conventional and unconventional recommendations that may reduce modeled impacts from stationary sources. Conventional recommendations are those that have been applied and accepted in the past whereas unconventional recommendations are those that are new, relatively unproven, and/or rely more on modeling experience combined with a knowledge of cause and effect than agency precedent.
Measure 1 – Source Parameter and Emission Changes
In the past, many sources have resorted to raising stack heights, changing air flow using higher speed fans or reducing the cross sectional areas of stack openings to increase flow velocity, thereby increasing plume rise and perhaps reducing ground level air concentration estimates. Another potential parameter change is combining flues (being mindful of EPA’s regulation on not allowing “dispersion techniques” solely for the purposed of demonstrating compliance with an ambient level). Adjusting the permit limit for emissions will also reduce the direct impact of a source proportionally to the emission reduction. Limiting the use and testing of emergency generators and limiting the hours of operation or reducing capacity of certain operations (such as the annual number of start-ups and shutdown events) will also reduce ambient impacts directly depending on how the agency considers infrequent and/or unplanned short-terms emissions in 1-hour modeling.
Measure 2 – Converging Regulatory Benefit
Many sources will be subject to new and upcoming regulations like Boiler Maximum Available Control Technology (MACT), the Clean Air Interstate Rule (CAIR), or the Mercury and Air Toxics Standards (MATS, aka Utility MACT). Thus, taking into consideration any one of these programs or the combination may lead to fuel switching, additional controls, or permit emissions reductions for existing sources in a given modeling domain.
Measure 3 – Changing the Boundaries of “Ambient Air”
If unacceptable modeled impacts are occurring on nearby adjacent properties, roadways, rail lines or hillsides, a company may be able to purchase and restrict public access to such areas, thus eliminating these land areas from “ambient air” for which the source must demonstrate compliance. Similarly, properties that are currently under a company’s ownership, but not restricted from public access, are considered ambient air under EPA’s definition and could be fenced to prevent access and thus eliminate these areas from modeling. Overwater receptors can also be problem if an agency determines that they represent ambient air. Large turnaround basins and shipping lanes near ports can often be characterized as under the control of the property owner with access to the public restricted, and thus, not ambient air. Gates can be erected across rail feed lines into plants thus limiting the need to consider property generally within the confines of the plants as ambient air.
Measure 4 – For New and Modified Sources, Model Below the Significant Impact Levels
Under the rules and guidance for modeling as part of an application for a Prevention of Significant Deterioration (PSD) permit, a new or modified source can avoid NAAQS modeling if its predicted impacts are below certain concentrations called Significant Impact Levels (SILs). Predicted impacts below the SIL demonstrate that a source or group of sources are not likely to cause or contribute to a possible exceedance of the NAAQS. Even if you are above the SIL for a small area and for a few selected periods, it may not be at the location or during the time period when other sources have combined to be above the NAAQS. Another part to this SIL demonstration that is often overlooked is the modeling of “contemporaneous” increases in emissions and decreases (from coincidental shutdowns, operational changes, or other reasons).
Measure 5 – Evaluate Meteorological Data Representativeness
Most dispersion modeling studies over the past 30 years have relied on National Weather Service (NWS) data that are good quality data intended for application in aviation not dispersion modeling. Therefore, many of the parameters needed for modeling are not contained in NWS datasets (e.g., net radiation, temperature profiles, etc.) and must be derived independently through atmospheric physics and from sites that are not necessarily representative of the atmosphere where the facility under consideration is located. Two suggestions to improve modeling results are: 1) make sure the meteorological data selection process includes an evaluation of representativeness to the site and, 2) consider collecting onsite data. Note that onsite data collection is a resource and time consuming endeavor that will influence permitting costs and timelines.
Measure 6 – Evaluate Ambient Air Monitoring Representativeness
Selecting the appropriate ambient air monitoring to represent background concentrations is very important and is highlighted in EPA’s new guidance on combining sources and estimating background concentrations. The monitored background may be inclusive of other nearby localized sources, especially in urban areas, that are already represented in the inventory of sources included in the NAAQS analysis. This may result in the double counting of emissions sources and an over prediction of concentrations. With the new 1-hour probabilistic NAAQS, choosing an appropriate background monitor that minimizes double counting while still being representative of background concentrations for a specific area is paramount.
Measure 7 – Pollutant Specific Strategies
Emissions at facilities are generally reported in terms of oxides of nitrogen (NOX) rather than NO2 or any of the other various forms of NOX. Much of these NOX emissions are in the form of NO. In the presence of ozone in the atmosphere, NO can be oxidized into NO2, although an equilibrium state is usually established among NO, NO2, and ozone concentrations. Of fundamental importance is the issue that an ozone-limited atmosphere will restrict the amount of conversion from NO to NO2. Current EPA guidance, contained in a June 28, 2010 memorandum6, describes guideline recommendations for modeling annual averaged NO2 concentrations and puts such recommendations in the context of the new 1-hour NAAQS. A three tiered analysis scheme is provided from most conservative to least conservative. Tier 1 assumes that all NOx emissions are converted to NO2. Tier 2 specifies that the result of the NOx modeling be multiplied by an empirically-derived NO2/NOx ratio, using avalue of 0.75 for the annual standard and 0.8 for the 1-hour standard. Tier 3 allows consideration of other models or methods within models on a “case-by-case” basis and includes the Ozone Limiting Method (OLM) and the Plume Volume Molar Ratio Method (PVMRM). These methods should be considered for use as they may allow the model estimation of more realistic NO2 concentrations based on background ozone concentrations and in stack NO2/NOx ratios.
For SO2, one pollutant specific strategy may be to modify the SO2 decay coefficient (currently hardcoded into AERMOD as 14,400 seconds for urban sources) in rural areas. This strategy will need to be defended on a case by case basis but may be warranted given the right mix of nearby sources and land use.
Measure 8 – Use a Variable Emission Processor
Generally, regulatory agencies allow time of day and capacity restrictions to be accounted for in dispersion modeling. Furthermore, many facilities do not operate on a continuous basis at peak (maximum hourly) emissions. A proposed processor, EMVAP (Emissions Variability Processor) would simulate variable emissions7. EPA has not approved its use but has it under consideration. Some agencies allow for the “smoothing” of maximum hourly emissions when such emission scenarios are not representative of normal operation. Modeling of emissions that are less than maximum allowable hourly emission rates requires prior approval by the reviewing regulatory agency.
Measure 9 – Building Configuration Changes
For greenfield sites, architects can configure the layout of buildings and sources to minimize downwash. For existing sites, removing abandoned or deactivated buildings can reduce downwash effects. Sources are limited to modeling stacks up to GEP heights (no credit is given for heights above GEP), however, a source may consider adding tiered structures—or even a parapet—to increase GEP and allow for greater modeled stack heights. Another consideration for Greenfield facilities is the selection of site locations that are higher than the surrounding topography, thus increasing the effective stack height of sources and reducing modeled ambient concentrations. Finally, with regard to reducing the effects of building downwash, another option would be to model all site structures in a wind tunnel. EPA regulations allow the regulatory agency the discretion to require wind tunnel modeling in order to document excessive ground-level concentrations due to the effects of surrounding buildings and also to determine the GEP height.
Measure 10 – Use an Alternative Model to AERMOD
The Guideline states that AERMOD is the preferred model for industrial source modeling in simple and complex terrain within 50km of a source. However, for “complex wind flow” conditions, the CALPUFF model may be more appropriate and may help avoid the steady state, plume impacting conditions that are prevalent in AERMOD applications in complex terrain under low wind speeds and stable atmospheric conditions. The application of CALPUFF within 50km requires prior approval of the reviewing regulatory agency based on a case-by-case demonstration that AERMOD is not acceptable or clearly less appropriate than CALPUFF.
Measure 11 – Use the Urban Option in AERMOD
The “urban option” in AERMOD generally results in more turbulence during nighttime hours and, therefore, lower concentrations in the stable boundary layer (SBL). The physics of nighttime turbulence enhancement is rooted in the additional heating caused by an urban area. Hence, applicability of the urban option in AERMOD is related to the population density in the area around the facility. However, a facility with many heaters and other sources of excess heat (boilers, space heaters, pumps, processes, pavement, etc.) may emulate an urban area and increased turbulence during nighttime hours. Recent qualitative studies have shown that a refinery can experience up to a four degree (°F) increase in temperature over similar rural temperature readings. A four degree temperature change is equivalent to the temperate differential between a medium-sized city and a rural area. One method that could be employed to account for such heating and turbulence at large industrial facilities in AERMOD, and thus take advantage of increased turbulence during nighttime hours, is to use a measured or estimated temperature differential associated with sources of excess heat and assign a population density that simulates an equivalent a temperature differential.
Measure 12 – Stratify Plume Impact Points to Lower Concentrations
If several similar emission units are located at a facility and the combined results of their individual modeled impacts occur at the same locations, some emission unit stacks may be raised, lowered, moved, or any combination thereof to spatially stratify modeled impact locations of the individual emission units and lower maximum modeled concentrations. Another strategy that can induce the same effect is to modify one or more emission units’ exhaust parameters (e.g., add a heat recovery device) which changes its buoyancy and momentum plume rise , resulting in a different plume transport and dispersion pattern than other emission units.
Challenges surrounding the application of dispersion models for the 1-hour NO2 and SO2 NAAQS are considerable but not insurmountable. Some technical issues have recently been or are currently being addressed by the scientific community, EPA, and state and local permitting authorities. Meanwhile, permitting for new and modified sources goes on, air dispersion modeling is still required, and sources must still show compliance with ambient air quality standards. Guidance is available to modelers which provides the tools necessary to demonstrate modeled compliance with the NAAQS, but creativity on the part of the air quality analyst and dispersion modeler is required to select the most advantageous methodologies and to rigorously defend the validity of their application in each specific case. In depth knowledge of the regulations, recent state and federal guidance, and the physics of the model are needed to apply all reasonable means in the context of demonstrating compliance with the new 1-hr NAAQS. To streamline the permitting process, the applicant should propose all modeling methodologies prior to submitting final modeling results to gain agency approval, and work closely with state and federal air permitting authorities to ensure that new internal modeling guidelines are addressed in the modeling analysis.
1Guideline on Air Quality Models, 40 CFR Part 51, Appendix W, December 9, 2005.
6Applicability of Appendix W Modeling Guidance for the 1-hour NO2 National Ambient Air Quality Standards, memorandum from Tyler Fox to Regional Air Division Directors, Office of Air Quality Planning and Standards, June 28, 2010.
7 Robert Paine, Status Report on Development and Testing of the EMVAP System, presented at the 10th EPA modeling Conference, AECOM through sponsorship of the Electric Power Research Institute, Research Triangle Park, North Carolina, March 14, 2012.