Influence of Spatial Resolution and Retrieval Frequency on Applicability of Satellite-Predicted PM2.5 in Northern China

Abstract

Satellite aerosol optical depth (AOD) products have been widely used in estimating fine particulate matter (PM_2.5) concentrations near the surface at a regional scale, and perform well compared with ground measurements. However, the influence of limitations such as retrieval frequency and the spatial resolution of satellite AODs on the applicability of predicted PM_2.5 values has been rarely considered. With three widely used MODIS AOD products, including Multi-Angle Implementation of Atmospheric Correction (MAIAC), Deep Blue (DB) and Dark Target (DT), here we evaluate the influence of their spatial resolution and sampling frequency by estimating daily PM_2.5 concentrations in the Beijing-Tianjin-Hebei (BTH) region of northern China during 2017 utilizing a mixed effects model. The daily concentrations of PM_2.5 derived from MAIAC, DB and DT AOD all have high correlations (R²: 0.78, 0.8, and 0.78) with the observed values, but the predicted annual PM_2.5 exhibits a distinct spatial distribution. DT estimation obviously underestimates annual PM_2.5 in polluted areas due to lower sampling of heavy pollution events. By contrast, the retrieval frequency (~40-60%) of MAIAC and DB AOD can represent well annual PM_2.5 in nearly all 83 sites tested. However, MAIAC and DB-derived PM_2.5 have a larger bias compared with observed values than DT, indicating that the large spatial variation of aerosol properties can exert an influence on the reliability of the statistical AOD-PM_2.5 relationship. Also, there is notable difference between MAIAC and DB PM_2.5 due to their different cloud screening methods. The MAIAC PM_2.5 with high spatial resolution at 1 km can capture much finer hotpots than DB and DT at 10 km. Our results suggest that it is crucial to consider the applicability of satellite-predicted PM_2.5 values derived from different aerosol products according to the specific requirements besides modeling the AOD-PM_2.5 relationship.