A case study in the role of air quality models in regulatory development: Creation of an response surface model for ozone over the eastern United States

Over the past two decades, photochemical grid models have frequently been used to inform air quality policy decisions because of their unique ability to integrate meteorology and chemistry, allowing for calculation of the expected effects of a set of proposed emissions changes. Typically, air quality planners use these models to simulate specific, individual, “what if” scenarios that assess the impacts of a given control strategy or a finite set of control strategies. Among the limitations of the this modeling approach is that it does not provide the decision maker with any information on any alternate strategy that may represent a more efficient route to the planner's air quality goal (i.e., attainment of the national ambient air quality standards). As an example, one of the most pressing air quality issues of the next five years will be how to reach attainment of the eight-hour ozone standard in persistent nonattainment areas like Houston, New York City, and Philadelphia by 2010 as mandated by the Clean Air Act. Currently, there is disagreement among the States and the affected stakeholders as to whether the most efficient path to attainment involves mostly local volatile organic compound (VOC) and nitrogen oxides (NOx) control, or continued reductions in regional NOx emissions. One modeling tool that can be used to address efficiency issues like the one outlined above is via a “model of the model” approach. If enough air quality model runs are completed, it is possible to build a response surface that will enable predictions of air quality changes resulting from various types of emissions reductions without having to continually rerun the air quality model itself. This type of tool can become even more powerful to the air quality planner when the model predictions are paired with control costs and/or economic benefits. This paper describes the development and application of a modeled ozone response surface over the eastern United States.

The modeled response surface was developed from 140 CAMx simulations for a 30-day set of meteorology. This eastern U.S. response surface considered seven types of emissions reductions: nonroad VOC, nonroad NOx, electrical generating unit (EGU) NOx, non-EGU NOx, on-road VOC, on-road NOx, and other area VOC emissions. These source sectors were varied by factors ranging from 0.0 to 1.2 of their 2015 base emissions over two geographic domains (projected attainment and nonattainment areas in 2015), for a total of 14 policy space dimensions. The paper will first describe the evaluation procedures that quantify the error associated with the response surface output, relative to any individual run or any of the multiple verification runs that were completed. The rest of the paper will be devoted to a presentation of the condensed model output that can be utilized by air quality planners to guide their strategy selections. Case studies investigating optimal control strategies will be performed for three specific ozone nonattainment areas: Atlanta, Chicago, and Philadelphia.