Key points are not available for this paper at this time.
Abstract National Oceanic and Atmospheric Administration (NOAA) and United States Environmental Protection Agency (USEPA) have been deriving surface particulate matter with a median diameter of 2.5 µm or less (PM2.5) from satellite aerosol optical depth (AOD) over Continental United States (CONUS) using a climatological PM2.5‐AOD regression relation. However, because PM2.5‐AOD relation can change over time, this method can have large errors when the relation deviates from the climatological values. In this work, the geographically weighted regression (GWR) model is used to estimate surface PM2.5 from AOD. The parameters of the regression model are derived dynamically in a daily or hourly manner using surface PM2.5 measurements and AOD from satellite. The method is evaluated using Visible Infrared Imaging Radiometer Suite (VIIRS) AOD and Advanced Baseline Imager (ABI) AOD to estimate daily and hourly PM2.5 over CONUS, and Advanced Himawari Imager AOD to estimate hourly PM2.5 over Taiwan. The algorithm performs much better than using simple climatological relationships. The estimated daily PM2.5 from VIIRS AOD has a cross validation (CV) R 2 of 0.59 with surface measured PM2.5, bias of 0.09 µg/m 3 and Root mean square error (RMSE) of 5.66 µg/m 3 . The hourly PM2.5 estimates from ABI AOD has a CV R 2 of 0.44, bias of 0.04 µg/m 3 , and RMSE of 4.53 µg/m 3 . The algorithm will run in near‐real‐time at NOAA to provide air quality community PM2.5 estimates over CONUS.
Zhang et al. (Sat,) studied this question.