Model-to-model sensitivity testing was performed with an ensemble of CMAQ models using two chemical mechanisms, Carter’s SAPRC99 and the CB4 of Gery et al., on three grid domains, a continental 32km domain, an 8km southeast U.S. domain, and two 2km domains centered on Nashville, TN and Atlanta, GA. Also, three versions of an additional mechanism, Stockwell et al.’s RADM2, were simulated on the 32km and 8km domains. The modeling series includes full emissions runs and three simple control strategies of 50% NOX, or 50% VOC, or a combined 75% NOX and 25% VOC control in all grid domains and with each of the chemistries for a 4-14 July 1999 period of analysis.
We applied some of the suggested indicator species ratios and other diagnostic tests to observations and model simulations for days in July 1999 when the SOS99 and SEARCH99 stations were operational in Nashville, TN and Atlanta, GA providing high precision measurements for NO2, a key component of several proposed indicators. The observations at Cornelia Fort (Nashville) and Jefferson Street and Yorkville (Atlanta) were used to anchor the full emissions base case model results. We note slight differences across the chemistries from these comparisons with the older chemistries RADM2 and CB4 responding more similarly to each other than to the newer SAPRC99. Model-to-observations comparisons were also performed at these sites with some of the suggested chemical indicators in the full emissions ensemble members.
Earlier work by Dennis et al. with one of our indicators, [O3]/[NOX], suggested that Atlanta would realize local NOX disbenefits, i.e., increased [O3] for decreased NOX emissions, on some days with lower maximum [O3] but that the smaller urban area Nashville would benefit from the regional NOX control nearly every day. Dennis et al. also noted that the model-predicted control response for 32km and 8km results were not completely similar and that differences were apparent between older and newer chemical mechanisms. The analysis reported here was designed to help understand these differences as part of a larger experiment with CMAQ in which the model’s sensitivity response to separate and combined perturbations in physical and chemical components was analyzed.
For one part of the analysis we calculate a model control ratio as control case O3 ratioed to base case O3 which allows us to discriminate usefully between ensemble members across space and time. We note, for example, that grid scale does influence the model O3 control response for large urban areas like Atlanta with finer grid scales enhancing differences between the older chemistries CB4 and RADM2 and the newer SAPRC99. We also note differences in the O3 control response across space from the large urban area of Atlanta to a suburban Atlanta location to the smaller urban area Nashville. Further, NOX processing, and changing the rate of that processing with a large change in chemistry or grid scale, appears to be a key element in determining the model’s control sensitivity response. Most significantly, the rank or level of a model’s O3 control response cannot be reliably inferred from the rank or level of the O3 difference between ensemble members’ full emissions response, and the degree of this disparity changes also with grid scale, location, and degree of radical limitation in the local photochemical regime.
Potential causes in the chemical mechanisms for the disparities observed in the chemical sensitivity runs are explored with diagnostic probes of the mechanisms’ NOX and radical cycles. With these we note systematic differences in processing rates between mechanisms and across the three control strategies. Diagnosing and understanding these differences is crucial for choosing the correct control strategy for a designated area and for advancing the general state of knowledge of nonlinear O3 production and control.
Supplementary URL: