3. OZONE PRODUCTIVITY DATABASE 3.1 Introduction This chapter discusses the development of an ozone productivity database for all elevated point NOx sources in eastern and central Texas as a function of time and space using the Comprehensive Air Quality with Extensions (CAMx) photochemical grid model. The CAMx model, the emission inventory input of the elevated NOx sources, and three air quality episodes used for modeling are described. This chapter also details the methodology for the point source sensitivity analyses performed on each grid cell containing a large, elevated point source (or sources) as well as the conversion of the CAMx model output to the final ozone productivity database format. 3.2 CAMx model As discussed in the literature review, ground level ozone formation is an extremely complex phenomenon that depends on nonlinear chemical mechanisms, detailed meteorological information, and precursor emissions input. Based on these inputs and chemical and physical models, photochemical grid models are used to predict ozone concentrations over time and space. A number of common versions of photochemical grid models are available. Basically, only three such models are accepted by the EPA for regulatory purposes: UAMV, CAMx, and MODELS3. This research applies ENIVIRON’s Comprehensive Air Quality Model with Extensions (CAMx) as the primary tool to quantify ozone production due to elevated point sources at different times and locations. This section gives an overview of the CAMx model and describes the elevated point emissions data model input that is manipulated during the research. Additionally, the three 4243episodes that are modeled and used to represent a range of metrological conditions are discussed. 3.2.1 BACKGROUND CAMx is the model used for this research and by the TNRCC in air quality policy development because it is the most modern (post 1995 coding), flexible, and modular. The following summary of the CAMx model is adapted from the CAMx users guide (ENVIRON, 1998). The CAMx model is an Eulerian photochemical grid computer model that allows for the integrated assessment of photochemical and particulate air pollution over scales ranging from individual point sources to multi-state regional effects. According to the users manual, CAMx “is designed to unify all of the technical features required of ‘state of the science’ air quality models into a single system that is computationally efficient, easy to use, and publicly available” (ENVIRON, 1998). The CAMx model simulates the emissions, dispersion, and removal of pollutants in the lower troposphere by solving the pollutant continuity equation for each chemical species of interest in a three-dimensional grid: ()()movallEmissionlChemistrylltttHttctctccKtzhczccVtcRe2∂∂+∂∂+∂∂+∇⋅∇+⎥⎦⎤⎢⎣⎡∂∂∂−∂∂+⎟⎠⎞⎜⎝⎛⋅∇−=∂∂→η Where VBHB is the horizontal wind vector, η is the vertical entrainment rate, h is the layer interface height, and K is the turbulent exchange coefficient. The first term on the right-hand side represents horizontal advection, the second term represents net resolved vertical transport, and the third term represents sub-grid scale turbulent diffusion. The continuity equation describes the time dependency of the average species concentration within each grid cell volume as a sum of all the physical and chemical processes operating on that volume, and averaged over the volume.The key physical processes that CAMx models include: horizontal advection and diffusion, vertical transport and diffusion, chemistry, dry deposition, wet deposition, and plume in grid. The continuity equation is numerically marched forward over a series of time steps. At each step the separate contributions of each major process (advection, dispersion, chemistry, etc.) to concentration change is calculated within each grid cell for each species of interest. CAMx supports five specific chemical mechanisms. The chemical mechanism used in this research is based on the Carbon Bond mechanism version 4 (CBM-IV; Gery et al., 1989) that is compatible with EPA regulations. The chemical mechanism also contains updated isoprene chemistry based on Carter (1996). The regional modeling in this research has a grid cell resolution of 16 kilometers by 16 kilometers. However, finer resolution may sometimes be needed to adequately model the chemistry of point source NOx plumes. To deal with sub grid scale processes, CAMx incorporates a Plume-in-Grid (PiG) feature which allows CAMx to track individual plume segments, accounting for dispersion and chemical evolution until their mass can be adequately represented within the grid model framework. A simplified schematic illustration of the CAMx PiG submodel is given in Figure 10. A stream of plume segment “puffs” is emitted from an elevated NOx source for each grid time step. The puffs are assumed to possess a Gaussian shaped two-dimensional concentration distribution in the cross-section and a constant cross section along the axis of the plume. Each puff moves horizontally downwind, grows via diffusion and undergoes chemical reactions. The puffs dilute emitted NOx and entrain ambient O3. The puff is “slaughtered” (all remaining emissions released into the grid cell) due to dilute NOx or age. 44Figure 10 A simplified schematic of the CAMx PiG submodel A summary of required input for the CAMx model is presented in Table 4. Typically, files representing domain-wide inputs (area, biogenics, meteorology) require gridded data, i.e., data values for every grid cell in the modeling domain. Performing a model run requires preparing the input files, which involves preprocessing of data and manipulation of emissions input files. For example, emissions can be masked with a preprocessor program that reduces selected chemical species of specified types of emissions in a defined area (e.g., 50% mobile source NOx reductions in Dallas). In order to run the model, a run script must be developed that defines the episode, directs the model toward all of the required input files, and defines the format of the model output. The amount of time required to perform a model run is indicative of the computational complexity of the model, and this type of modeling has only recently become feasible with improved computer capacity and speed. For example, a model run for the 1995 June 18-22 episode used in this research