The Influence of Model Resolution on Ozone in Industrial Volatile Organic Compound Plumes

Regions with concentrated petrochemical industrial activity (e.g., Houston or Baton Rouge) frequently experience large, localized releases of volatile organic compounds (VOCs). Aircraft measurements suggest these released VOCs create plumes with ozone (O(3)) production rates 2-5, times higher than typical urban conditions. Modeling studies found that simulating high O(3) productions requires superfine (1-km) horizontal grid cell size. Compared with fine modeling (4-km), the superfine resolution increases the peak O(3) concentration by as much as 46%. To understand this drastic O(3) change, this study quantifies model processes for O(3) and odd oxygen (O(x)) in both resolutions. For the entire plume, the superfine resolution increases the maximum O(3) concentration 3% but only decreases the maximum O(x) concentration 0.2%. The two grid sizes produce approximately equal O(x) mass but by different reaction pathways. Derived sensitivity to oxides of nitrogen (NO(x)) and VOC emissions suggests resolution-specific sensitivity to NO(x) and VOC emissions. Different sensitivity to emissions will result in different O(3) responses to subsequently encountered emissions (within the city or downwind). Sensitivity of O(3) to emission changes also results in different simulated O(3) responses to the same control strategies. Sensitivity of O(3) to NO(x) and VOC emission changes is attributed to finer resolved Eulerian grid and finer resolved NO(x) emissions. Urban NO(x) concentration gradients are often caused by roadway mobile sources that would not typically be addressed with Plume-in-Grid models. This study shows that grid cell size (an artifact of modeling) influences simulated control strategies and could bias regulatory decisions. Understanding the dynamics of VOC plume dependence on grid size is the first step toward providing more detailed guidance for resolution. These results underscore VOC and NO(x) resolution interdependencies best addressed by finer resolution. On the basis of these results, the authors suggest a need for quantitative metrics for horizontal grid resolution in future model guidance.