The design, development, validation, and implementation of a variational data-assimilation module to ingest the observations resulted in quantifiable improvements in CMAQ surface fine particle results. For a suite of 40 test "case days," the overall eastern US low bias in modeled total PM2.5 concentration improves by 13.35%, considering a weighted average of bias improvements among all surface measurement networks. Similarly, the modeled total PM2.5 RMSE improves by just over 6%, the R-squared statistic improves by 11.4%, and the Index-of-Agreement improves by 4.88%, again considering a weighted average across all measurement networks. With respect to total column aerosol optical depth, there is unequivocal improvement in total AOD east of the Rockies, such that the revised standard deviation of the AQF-DSS AOD error as measured against surface sun photometer data remains nearly flat as concentrations increase. Further, the assimilated model intercept is lower as well. In the western U.S., the impact of MODIS AOD data-assimilation is less evident. However, a better error model for the new Deep Blue retrieval method, which holds promise over the brighter reflecting land-surfaces characteristic there, should help rectify this.
Based on these promising improvements, adoption of the new MODIS data-assimilation module into real-time operations is underway, and early results from real-time operational implementation parallel runs will be described.