A new response surface approach for interpreting Eulerian air quality model sensitivities

The Regional Acid Deposition Model (RADM) was used in a sensitivity study at 80 km and 20 km grid resolutions to simulate urban and regional tropospheric conditions in the eastern U.S. for the time period 19 July to 12 August 1988. Using one-at-a-time (OAT) sensitivity perturbations and a mini-global analysis we varied three model inputs in the range of their known uncertainties: mixing height; emissions of nitrogen oxides (NO,) and volatile organic compounds (VOC); and the rate constant for the reaction OH + NO2 --> HNO3. Two measures, ozone concentration ([O-3]) and the ratio [O-3]/[NOx], were used to assess the photochemical system's sensitivity response on an O-3 response surface. We find that: (1) these input uncertainties change the system's state in different ways, altering either the [O-3] or the system's response to emissions reductions, or changing both responses together; (2) the [O-3]/[NOx] measure successfully predicts the direction and relative degree of change in the system state while [O-3] does not; and (3) for most of the days and locations we have examined, the system's responses to uncertainties simulated OAT are nearly additive when compared with the mini-global case, and that both of the sensitivity measures we assessed correctly tracked this additivity. We conclude that incorporating the [O-3]/[NOx] ratio is a significant first step in constructing a useful sensitivity "vector" for understanding systems-level changes in the model. However, additional measures that can successfully track the change of shape of the response surface are still needed. (C) 1999 Elsevier Science B.V.