Estimating PM2.5 in Xi'an, China using aerosol optical depth: a comparison between the MODIS and MISR retrieval models.

Satellite measurements have been widely used to estimate particulate matter (PM) on the ground, which can affect human health adversely. However, such estimation from space is susceptible to meteorological conditions and may result in large errors. In this study, we compared the aerosol optical depth (AOD) retrieved by the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Multiangle Imaging SpectroRadiometer (MISR) to predict ground-level PM2.5 concentration in Xi'an, Shaanxi province of northwestern China, using an empirical nonlinear model. Meteorological parameters from ground-based measurements and NCEP/NCAR reanalysis data were used as covariates in the model. Both MODIS and MISR AOD values were highly significant predictors of ground-level PM2.5 concentration. The MODIS and MISR models had overall comparable predictability of ground-level PM2.5 concentration and explained 67% and 72% of the daily PM2.5 concentration variation, respectively. Seasonal analysis showed that the MODIS and MISR models had overall comparable predictability of ground-level PM2.5 concentration, with the MISR model having a higher correlation coefficient (R) and thus giving a better fit in all seasons. The MISR model had high prediction accuracy in all seasons, with average R(2) and absolute percentage error (APE) of 0.84 and 15.3% in all four seasons, respectively. The prediction of the MODIS model was best during winter (R(2)=0.83) with an APE of 19%, whereas it was relatively poor in spring (R(2)=0.56) with an APE of 21%. Further analysis showed that there was a significant improvement in correlation coefficient when using the nonlinear multiple regression model compared to using a simple linear regression model of AOD and PM2.5. These results are useful for assessing surface PM2.5 concentration and monitoring regional air quality.