Development of an Empirical Model for Predicting Dust Aerosol Optical Depth Using Satellite and Reanalysis Data for the Use in Earth System Models

Abstract:

Although the availability of satellite and ground based aerosol optical depth (AOD) data is increasing, application of AOD data in dust modeling is hindered because the AOD contains a non-dust signal as well. In this work, we extract the dust signal from AOD data using a novel approach by associating AOD with the surface wind. We then develop an empirical multiple regression model for predicting dust optical depth (DOD). For this purpose, we utilize long-term (2003-2012) MODIS satellite and GLDAS reanalysis gridded data. MODIS deep blue AOD was used as a dependent variable in the regression model whereas ERA-Interim surface wind and divergence, GLDAS soil moisture and soil temperature, and MODIS normalized difference vegetation index (NDVI) were used as independent variables. Results show that wind and divergence are positively correlated with the DOD, whereas soil moisture, soil temperature, and NDVI are negatively correlated. The developed multiple regression model has an overall R-squared of 0.73 and root mean squared error of 0.39. Our model is useful in testing the performance of more physically based dust models and can be implemented in Earth System Models.