Estimating PM2.5 surface concentrations from AOD: A combination of SLSTR and MODIS

Abstract

Compared to surface in-situ observations, satellite data on aerosol optical depth (AOD) enables area-wide monitoring of tropospheric aerosols. However, coverage and reliability of satellite data products depend on atmospheric conditions and surface concentrations have to be retrieved from AOD. This study investigates the potential to produce reliable maps of PM2.5 surface concentrations for Germany and parts of the surrounding countries using AOD based on observations by three different satellite sensors. For the first time, AOD retrievals from the Sea and Land Surface Temperature Radiometer (SLSTR) onboard Sentinel-3A are used together with those from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the two NASA platforms Terra and Aqua. We investigate the differences and similarities of the three different satellite products in terms of coverage, resolution and algorithmic performances. Based on this analysis we examine the suitability and advantage of a combination of these data sets. We can substantiate an increase in mean daily coverage from a maximum of 10.2% for the individual products to 16.7% for the ensemble product. Using a semi-empirical linear regression model, we derive surface-level PM2.5 concentrations and attain an overall correlation of 0.76 between satellite-derived and in-situ measured PM2.5 concentrations. By considering surface measurements, the systematic error (bias) and the root mean square error (RMSE) can be significantly reduced. The general model performance is evaluated by a 5-fold cross validation and the relative prediction error (RPE).