Comparison of model performance for source apportionment of visibility impacts and regional haze rule assessment

The U.S. EPA's Regional Haze Rule establishes implementation guidelines for the determination of Best Available Retrofit Technology (BART) that may be required for industrial and utility stationary sources of air emissions within one or more of twenty-six stationary source categories that have otherwise been grandfathered from emissions control requirements. BART is an emission limitation representing available and cost-effective emissions control technology for major sources of air pollutants including oxides of nitrogen (NOX), sulfur dioxide (SO2), and particulate matter (PM) that affect visibility in federally protected Class I areas such as National Parks, Wilderness Areas, and National Forests. Air quality modeling plays a critical role in the BART determination; EPA's implementation guidance specifies the use of modeling at the applicability stage to determine whether a single source “causes” or “contributes” to visibility impairment and is therefore subject to the rule, and at the BART determination stage to quantify the relative effectiveness of visibility improvement among multiple control technology alternatives. This paper presents a comparison of model performance between two air quality models, CALPUFF and CMAQ, which may be used to assess visibility impacts and control strategies for individual stationary sources under the Clean Air Visibility Rule.

EPA's guidelines recommend that the CALPUFF modeling system be used for BART applicability assessments and control technology analyses. The system comprises three components: a meteorological model that utilizes mesoscale diagnostic fields (e.g., MM5), non-steady state puff dispersion and chemical transformation algorithms, and data postprocessing algorithms. CALPUFF computes ambient concentrations from tens of meters to several hundred kilometers downwind of an emissions source, and is distinguished among regulatory air quality models because it has been proven more rigorous in simulating ambient concentrations, in particular those that result from chemical transformations (e.g., SO2 to sulfates and NOX to nitrates) of emissions from a single source.

CALPUFF demands modest resources for input data, computational processing, and data storage, but is not as demanding (nor as robust) as other more refined air quality models such as the Community Multiscale Air Quality (CMAQ) model that is used for regional-scale, multiple pollutant analyses. Another key distinction of the CALPUFF system is that the model is designed to compute air quality impacts attributable to an individual facility with fine scale resolution (on the order of 1 km) as opposed to simulating emissions from a broad source category or geographic region with lesser resolution (12 km). When used for source apportionment analyses, CMAQ is typically applied for the latter analyses that do not require differentiation of contributions from individual sources.

EPA acknowledges in its BART implementation guidelines several criticisms of CALPUFF, foremost that the chemical transformation algorithms consider only the secondary formation of sulfates and nitrates and neglect the formation and visibility impacts of secondary organic aerosols and primary emissions of volatile organic compounds. CALPUFF has also been shown to compute specious results at large downwind distances at the extreme applicability range of the model's puff-splitting algorithms. More refined models such as CMAQ simulate atmospheric constituents using advanced thermodynamic chemical transformation algorithms on a regional scale, and do not exhibit these problems. Nonetheless, EPA recommends the application of CALPUFF for BART assessment, stating that “CALPUFF is sufficiently reliable to inform the decision making process,” but that “other models, used in combination with CALPUFF may be helpful in providing a relative sense of the source's visibility impact and can aid in informing the BART decision.” EPA weighs the preceding factors in its final BART guidelines and ultimately concludes, “CALPUFF is currently the best application available to predict the impacts of a single source on visibility in a Class I area.” Individual state, local, and tribal regulatory agencies are ultimately responsible for implementing the Regional Haze Rule, and EPA leaves discretion to these agencies to use an appropriate model other than CALPUFF if circumstances warrant. However, given many regulatory agencies' and stationary sources' inexperience with, and limited resources to use, more sophisticated models to analyze numerous eligible stationary sources individually, CALPUFF is expected to be the preferred model for BART assessment.

Given the relatively low triggering thresholds for BART and well-known uncertainties in CALPUFF modeling results, it is not surprising that many BART-eligible sources are concerned that indeterminate model results could be used to require a major capital investment in emissions controls. These sources, especially those whose impacts are within a small margin of the triggering threshold due to modest emission levels and/or considerable distance from a Class I area, would be motivated to invest in more sophisticated modeling to ensure that any requirement for investment in control technology is warranted. Facilities need to be prepared to quickly undertake such analyses, since state regulatory agencies have a limited timeframe (approximately one year) to make determinations about which sources are subject to BART requirements. This paper seeks to facilitate the process by demonstrating viable modeling techniques that should be acceptable under the BART guidelines to evaluate alternative modeling protocols for assessment of individual stationary sources using CALPUFF and CMAQ.

The baseline for comparison is a CALPUFF analysis prepared using presumptively acceptable model inputs and processing options based on the general CALPUFF modeling protocol prepared by VISTAS, the regional planning organization in the Southeastern U.S., which is responsible for coordinating BART assessments for state and local agencies within its domain. The results of the baseline analysis will be compared with results from CMAQ simulations over an equivalent domain and using comparable inputs for meteorological conditions (e.g., MM5 diagnostic fields), source emissions profiles, and ambient chemistry data, but using innovative sensitivity and source apportionment techniques such as the direct decoupled method (DDM) to discern the visibility impacts from the single source. The viability of CMAQ as an alterative to CALPUFF modeling for BART are assessed not only in terms of the robustness of the visibility impairment results, but the practical aspects of the modeling exercise including simulation runtime, necessary computing resources, and availability of model inputs, all of which will influence the discretion of the reviewing agency to consider an alternative approach for BART analyses.