An applicant for major source construction permits must demonstrate that the new emissions from its proposed new source or regulated modification, plus any emissions resulting from associated industrial growth at the facility, plus emissions from existing sources, will not have an unacceptable impact on ambient air quality. To do this, the applicant must analyze each regulated pollutant that will be emitted in significant amounts; and the analysis must be conducted in accordance with 40 CFR § 52.21 or with any EPA-approved state PSD program. Air dispersion modeling must be used to evaluate ambient air quality standards, existing ambient air quality levels, meteorological conditions, sensitive receptors, and area topography, as well as to predict the impact of new emissions on air quality.
EPA describes how this analysis should proceed in its 1990 New Source Review Workshop Manual (published as a draft and never finalized). According to the manual, an applicant should determine a new source’s impact by: (1) modeling for the predicted impact from the source for each covered criteria pollutant; (2) adding this concentration to the background concentration for the pollutant; and (3) adding this total to any secondary emissions for the pollutant that are expected as a result of the modification. This article will focus on the first item—completing the major source construction permit application modeling process.
Complicated processes occurring in nature can be described, albeit incompletely, using mathematical models. Atmospheric processes disperse the emissions downwind into a less concentrated form. Air dispersion models are computer tools that use mathematical equations to try to describe the dispersion process and to approximate the way substances are distributed downwind from an emission source. By knowing the initial emission characteristics, one can use a dispersion model to statistically predict the behavior and movement of emissions plumes and the chemical air concentrations at selected downwind receptor locations. Regulatory agencies have long employed Gaussian models as their basic analytical tool to describe the distribution of continuous, buoyant plumes of air pollution coming from either ground level or elevated sources. (Gaussian models assume that the air pollutant dispersion follows a Gaussian distribution, meaning that the distribution reflects a normal bell-curve probability distribution.)
Air modeling must mathematically simulate the highly complex nature of plume behavior. For instance, models must capture atmospheric turbulence, one of the most significant factors affecting the dispersion of air pollution plumes. Turbulence increases the mixing of unpolluted air into the plume, spreading the plume both horizontally and vertically as it moves downwind, thereby reducing the concentration of pollutants in the plume. Improved air dispersion models also take into account other forces that impact the dispersion of air pollutant plumes. These include building downwash, gravitational settling, and washout due to rain. The effects of building downwash occur as an air plume goes over a building and creates enhanced air eddies on the downwind side of the building. This causes the plume and its pollutants to mix and deposit on the ground much sooner than would occur in the absence of the building. Moreover, through the process of “wet deposition,” rain causes washout of pollutants from a plume, which also impacts the rate at which pollutants are deposited on the ground.
In addition, the height at which an emissions plume mixes within the atmosphere represents a significant limiting behavior. Atmospheric conditions affecting the temperature of the air above the earth affect the height at which a plume and its contaminants mix within the atmosphere. For example, so-called inversion layers in the atmosphere—in which temperatures increase with altitude, rather than decrease—cause limited mixing and, thus, generally result in higher concentrations of emissions. Despite these complexities and the limitations of models, the air dispersion model continues to represent a basic and essential tool in the field of air quality regulation. See AERMOD Implementation Guide.
The Emergence of Air Dispersion Models under the Clean Air Act
The foundation of the CAA is the designation of the NAAQS and the classification of each county in the United States as nonattainment or attainment/unclassified. Once these classifications are made, EPA and their delegated state agencies are required to protect the air quality of attainment areas when permitting new major sources. Because necessity follows need, the first official dispersion models were established.
Section 165(e)(3)(D) of the CAA describes air permitting preconstruction requirements and requires EPA to “specify with reasonable particularity each air quality model or models to be used under specified sets of conditions for purposes of this part.” EPA first published the Guideline on Air Quality Models (EPA 450/2-78-027R) in April 1978, and it was incorporated by reference in the regulations for the PSD of Air Quality in June 1978. After many subsequent revisions, the Guideline now appears in 40 CFR § 51 as Appendix W. Early in the rulemaking process, EPA solicited comments regarding models developed outside of EPA. These included stack tip downwash algorithms (a forerunner of today’s Building Profile Input Program, or BPIP) and models used only by specific state agencies, such as the Texas Models. The Tenth Conference on Air Quality Models took place in March 2012. Held approximately every three years, this conference (required by Section 320 of the CAA) serves to standardize modeling procedures and practices for programs like PSD.
The Art and Science of Dispersion Modeling
To quote EPA, “Dispersion models are source-oriented models that characterize atmospheric processes by dispersing a directly emitted pollutant to predict concentrations at selected downwind receptor locations.” The most commonly used air dispersion model, American Meteorological Society/Environmental Protection Agency Regulatory Model (AERMOD), addresses short-range (<50 km) dispersion from stationary industrial sources. AERMOD is used by facilities and consultants to demonstrate whether new emission sources will or will not violate the NAAQS or PSD increment standards. Although other models are occasionally used, AERMOD is EPA’s workhorse model, used in probably 99 percent of situations.
Dispersion modeling is more of an art form than an actual science. Knowledge not only of how the model works, but also of the various sensitivities to small adjustments and an ability to balance these parameters, is required by the modeler in order to appropriately represent the emission source and to maximize operational flexibility. No current model, not even AERMOD, precisely replicates real world dispersion, so modelers face a complicated problem when attempting to fine-tune the model outputs in order to establish appropriate permit conditions. For example, calm winds (a wind speed less than three knots) dominate at night, so a plant road used by truck traffic to haul materials on and off the facility modeled to operate at night will result in higher concentrations of particulate matter than if night hours are excluded. However, if the model excludes night operation, the facility being modeled will have to monitor and record on a daily basis the time that the first and last truck enters the facility. Excluding nighttime operation from the model leads to a restriction on nighttime operation in the permit.
This level of recordkeeping can become extremely onerous over the life of a permit. Plus, any unforeseen circumstances that require a nighttime delivery would be prohibited as outside of what is authorized by the permit. A more flexible permit would not limit trucks to daytime operation. In other words, if a limit on nighttime operation is necessary to model compliance, that limit outweighs operational flexibility. Multiply this balancing act by several dozen parameters and you have an understanding of the artwork required by the modeler. At a minimum, the following parameters are required in order to run the model. See AERMOD Implementation Guide.
Control Options. These include pollutant and averaging period(s), default or non-default regulatory options (such as dry deposition or use of the ozone limiting method/plume volume molar ration method for 1-hour NO2 concentrations), and rural or urban coefficients.
Source Options. This is an emission source described by its location (usually in Universal Transverse Mercator (UTM) coordinates, which is a metric comparable to latitude and longitude), base elevation, emission rate, and source type (point source or fugitive source, in the form of area or volume sources, are the most common). Release parameters, which vary depending on the source type, but for a point source include stack height, temperature, exit diameter, and exit velocity. Area and volume sources replace the stack parameters with initial horizontal and/or vertical dispersion parameters. Source grouping facilitates use a single model to cover a variety of operating scenarios. Emission factors limit operation to hour of day, selected months, certain wind speeds, or a multitude of other combinations.
Receptor Options. The model will calculate a concentration at each X, Y coordinate that is identified as a receptor. Most models will have thousands of receptors. Real world terrain data from U.S. government maps are used to establish the ground height of each receptor. Receptors are located at residences and sample ambient air beyond the fence line of a facility out to a certain distance, such as ten kilometers.
Meteorological Options. The model will calculate a concentration at each hour of meteorological data that is identified. Usually a five-year data set is used (5 x 8,760 hours for a single model run).
Boundary Options. Fence lines are established because NAAQS are only relevant in ambient air, or air where the general public can gain access. This is not the same as a facility’s property line. Only the land protected by a contiguous physical fence can be excluded from the model. In the absence of a fence, all on-property land must be included in the receptor grid. It is imperative that all substantial facility buildings be accurately represented in the model in order to allow a proper calculation of building downwash, which is one of the model preprocessors. Many of the modern, advanced dispersion modeling programs include a preprocessor module for the input of meteorological and other data; and many also include a postprocessor module for graphing the output data and/or plotting the area impacted by the air pollutants on maps. The plots of areas impacted may also include isopleths showing areas of minimal to high concentrations that define areas of the highest health risk. The isopleths plots are useful in determining protective actions for the public and responders.
Output Options. The form of the output is selected including which high values are desired. Some standards allow the highest modeled value to be excluded or establish a statistical form of the result, such as the ninety-eighth percentile, averaged over three years.
In addition to the model inputs listed above, modelers must also run various preprocessor programs, entirely separate algorithms used to transform meteorological and observational data, such as wind speed and direction, temperature, and building height, into a format that can be used in AERMOD. One program creates a file that contains elevation and hill-height scaling factors for each receptor in the air dispersion study. For example, the wind will have one effect on a receptor located in a flat plain but a different effect on a receptor located on the side of a steep hill. The scaling factors adjust for these disparities in impacts. The Building Profile Input Program for PRIME (BPIPPRM) incorporates the concepts and procedures expressed in the good engineering practice (GEP) technical support document, building downwash guidance, and other related references that correctly calculate building heights and projected building widths for simple, multi-tiered, and groups of structures. An analysis of GEP is required by EPA in 40 CFR § 52.21(h). GEP is the mechanism by which EPA prevents a facility from constructing excessively tall stacks for the purpose of diluting their emissions into the air. GEP sets a maximum stack height that can be represented in the model by allowing only those stacks that are comparable in height to the surrounding buildings.
EPA Guidance on Air Quality Models
Appendix W to 40 CFR § 51 provides the EPA guideline on air quality models. It has been updated many times over the years to reflect the latest developments in algorithms and standard practices and to provide consistent regulatory application of air models. The 1977 CAA mandated “consistency” and encouraged the standardization of model applications. EPA’s Guideline on Air Quality Models (hereafter, Guideline) was first published to “satisfy these requirements by specifying models and providing guidance for their use. . . .” The Guideline provides a common basis for estimating the air quality concentrations of criteria pollutants used in assessing control strategies and developing emission limits. Preface to Appendix W to 40 CFR § 51.
EPA uses three primary ongoing activities to keep the Guideline up to date: annual EPA modeling workshops conducted for the purpose of ensuring consistency and providing clarification in the application of models; the solicitation and review of new models from the technical and user community; and ongoing research efforts by EPA and others in air quality and meteorological modeling. The Guideline is updated only on an as-needed basis.
According to the Guideline, the development of a project’s dispersion model begins with an accurate site layout map, indicating all fence lines, property lines, and source and building locations. If the location is an existing facility, an accurate inventory of existing sources, including when they were first operated, along with existing buildings, is required. Information on new sources is taken from engineering specifications and is often a first round estimate and not how the facility will eventually be constructed, resulting in the need for as-built dispersion modeling to confirm that ambient air is protected. See Appendix W to 40 CFR § 51.
The model is initially run as a screening model for just the new emission sources for comparison to the modeling significant impact levels (SILs) (See 40 CFR § 51.165(b)(2)) and significant monitoring concentrations (SMCs) (See 40 CFR § 52.21(i)(5)(i)). EPA had not, prior to the adoption of SILs for PM2.5 , issued a regulation that would allow a source to forego a cumulative source impact analysis if the source demonstrated that its emissions would not result in an ambient air quality impact above a SIL, but EPA’s longstanding guidance authorized this approach. See 72 Fed. Reg. 54112, 54139 (Sept. 21, 2007) (proposed rule); New Source Review Workshop Manual, at C.24-C.25 (draft 1990). The SILs (concentrations) are measured in µg/m (micrograms per cubic meter) for the criteria pollutants SO2, PM10, PM2.5, NO2, and CO. The SILs and SMCs are “screening tools” the EPA uses to determine whether a new source may be exempted from certain New Source Review requirements. The SIL is a de minimis threshold used to determine whether a source may cause or contribute to a violation of PSD increment or the NAAQS (i.e., a significant deterioration of air quality). The SMC is an increase in ambient concentrations of pollution, below which EPA exempts a source from the PSD program requirement to gather and submit one-year pre-application ambient monitoring data.
If the screening model results for a pollutant/averaging period combination are below the significance levels, no further modeling is required for that combination. For example, the SIL and SMC for a 24-hour average PM10 are 5 µg/m and 10 µg/m, respectively. If the screening model results in a 24-hour average PM10 concentration of 8 µg/m, the project would not be required to do preconstruction ambient air monitoring but would need to run a refined (cumulative—includes nearby sources) dispersion model. When the screening model results in concentrations below both the SIL and SMC, no further modeling is required. Oftentimes, if the SMC is exceeded, the applicant will be allowed to use preconstruction monitoring data from a representative existing state-agency-operated ambient air monitor instead of constructing and operating their own onsite monitor to gather a year’s worth of data.
When an SIL is exceeded, the full dispersion model must be refined to include all existing sources and all neighboring sources within the radius of impact plus fifty kilometers. The radius of impact is the farthest distance from the facility in any direction where a receptor exceeding the SIL is located. Receptors are locations at which the model calculates the pollutant concentrations. The model results are compared with the NAAQS and to the PSD Increments. The PSD Increment is applicable in areas where the baseline has been triggered and represents a fixed level of air quality degradation allowed in attainment areas. In other words, if the air quality is good, new facilities are not allowed to increase pollution up to the point that the air quality verges on nonattainment. The baseline date is roughly the date when the first PSD application was submitted for an area. The model must demonstrate that there are either no modeled exceedances of the NAAQS and PSD Increments, or that the proposed project contributes no more than the SIL to any modeled exceedance.
Before comparing the model results to the NAAQS, a background concentration is added to the model results. This background value is usually determined from state-operated ambient air monitors and represents the air quality of the area if all the sources represented in the dispersion model were not in operation. The background value represents air quality levels due to mobile sources, pollution carried in the atmosphere from sources that are far away, and other sources that are not typically modeled. No background value is used for comparison with the PSD Increments because this standard is a relative, and not absolute, air quality value. PSD Increment is the degradation from whatever the air quality was when the baseline date was set. NAAQS is an absolute value of air quality that is considered acceptable. See 40 CFR Part 50.
The exception to this rule is for PM2.5. In Sierra Club v. EPA, 705 F.3d 458 (D.C. Cir. 2013), the Court of Appeals for the District of Columbia vacated EPA’s rules governing the SIL and SMC for new sources of PM2.5 on the grounds that there are no SILs or SMCs for PM2.5 and all PSD projects must conduct a refined modeling analysis in addition to providing one year of preconstruction monitoring (or use a representative existing monitor). In other words, for PM2.5 only, there are no federal screening levels and every model must be a fully refined dispersion model unless the screening levels are in the state regulations. The result for now is a state-by-state determination of whether or not this ruling has immediate impact or whether it must wait until state implementation plans are modified.
It is clear that PSD modeling is complicated; therefore, a dispersion modeling protocol is used to document and communicate with the state agency about what parameters and model options will be used. It is usually provided to the state permitting agency after initial runs have been made and the modeler has a feel for the complexity of the project. The modeling protocol is included in the permit application, along with the model results in tabular and often graphical form.
A model rarely, if ever, shows compliance with all the standards in the first run. Inevitably, adjustments must be made to the inputs to reduce the concentrations. This is where the conservative assumptions that were initially made are readjusted, making them closer to reality and resulting in less operational flexibility. It is possible to isolate each individual exceedance to determine which source or sources are the largest contributors. The modeler has an assortment of tools that can be used to reduce the modeled concentrations. A sample, but by no means all-encompassing, list of these methods includes the following:
- If the exceedances are near the fence line, the facility could opt to fence in its entire property or purchase adjoining land. An increase in stack height can drastically increase dispersion; however, limits on stack height exist both physically (a five hundred-foot tall, six-inch diameter stack could not withstand the effects of wind and gravity) and regulatory (a model cannot exceed good engineering practice (GEP) stack height, which is determined using BPIP-PRIME).
- Nighttime hours tend to have calm winds, which decrease dispersion. Limiting sources to daytime operation can reduce concentrations but also increase a facility’s compliance burden.
- Redundant sources can be limited to never operate simultaneously.
- Emergency sources, such as backup generators and fire pumps, can be limited in when they can be tested for operational readiness.
- A change in the layout of a new facility can move sources outside of the impact of building downwash.
- Emission rates can be lowered by reducing the maximum allowed fuel sulfur content, by adding a control device, or by reducing operating margin (operating closer to manufacturer guarantees).
- A stack diameter can be reduced, resulting in a proportional increase in stack velocity, thus increasing dispersion.
- A stack can be reoriented to vent in a vertical unobstructed manner by removing raincaps or by turning a horizontal stack upwards. “A tractor flap” can be employed to allow for vertical unobstructed release while simultaneously preventing rain from entering the stack.
- Because a NAAQS model includes neighboring sources within at least fifty kilometers, the exceedance can be examined for culpability to determine exactly which emission source is the largest contributor to the high concentration. An exceedance can be retained in a permit application if it can be demonstrated that the applicant’s contribution to the exceedance is below the SIL. The state agency would then approach the neighboring source under the authority of the State Implementation Plan (SIP) and require it to reduce its emission impact.
After receiving the complete application, the state agency will review all of the applicant’s determinations and come to its own conclusions regarding those determinations. The critical reviews will typically take place with the selection of the best available control technology (BACT) and the adequacy of the demonstrations regarding the potential impact on air quality.
As the above summary shows, the process of conducting an air quality analysis is extremely complex. In recent years, this process has been further complicated by EPA’s adoption of stricter NAAQS for various pollutants, including PM2.5, SO2, and NO2, both in terms of lower numerical standards and by shortening the averaging periods. For the most part, new and modified sources seeking PSD permits must demonstrate compliance with new NAAQS as soon as they are adopted. However, it typically takes EPA several years to promulgate rules on implementing PSD for newly adopted NAAQS. To provide help in the interim, EPA has issued numerous guidance documents regarding SILs for specific pollutants, which provide directions on addressing unique issues associated with implementing PSD for that pollutant.
In light of the many variables associated with conducting an air quality analysis, it is crucial that PSD permit applicants engage the permitting authority at each stage in the modeling process. To begin, the applicant must submit a modeling protocol to the permitting agency describing the modeling techniques and databases that will be used to conduct the preliminary analysis. If this analysis identifies one or more pollutants that exceed the specified significance level(s), the applicant must investigate the availability of existing ambient air quality and meteorological data and obtain confirmation from the permitting agency that the data can be used to perform the ambient air quality analyses. If the agency concludes that the data are inadequate, the applicant must consult with the agency to develop a plan for obtaining the necessary site-specific data. The applicant also must consult with the agency when identifying the appropriate computer model and developing the required NAAQS and increment inventories.
Dispersion modeling is protective of the NAAQS, but it is not necessarily predictive of air quality. With the development of faster computers and the refinement of modeling algorithms that better mimic reality, dispersion modeling will continue to be a dominant aspect of the PSD application process.