Evaluation and modification of SARA high-resolution AOD retrieval algorithm during high dust loading conditions over bright desert surfaces

Many studies have been carried out for the validation of Aerosol Optical Depth (AOD) retrieved using MODerate resolution Imaging Spectroradiometer (MODIS) satellite data at the spatial resolution of several to several tens of km. Most of the validation efforts have focused on urban areas with fewer studies focused on deserts with high amounts of dust. On the other hand, calculating AOD at a high spatial resolution over desert surfaces is of high interest to identify local and accurate dust source regions. In this study, two different high-resolution MODIS AOD retrieval algorithms were evaluated to investigate the nature of high dust loading events over the Middle East region. These algorithms include the Simplified Aerosol Retrieval Algorithm (SARA) at 500 m resolution based on the MOD09 Level 2 daily (original SARA_MOD09) and MCD43A4 16 day (modified SARA_MCD43) surface reflectance products. These algorithms were validated using AERONET AOD observations at Kuwait University, and accuracy was evaluated using descriptive statistics and Expected Error (EE) over land. Also, the results of these algorithms were evaluated with the Collection 6 (C6) Deep Blue algorithm at 10 km spatial resolution (DB_10km). Results show that the SARA_MCD43 and SARA_MOD09 AOD retrievals are well correlated (r) with AERONET data (SARA_MCD43 = 0.89 and SARA_MOD09 = 0.84), with a larger number of retrievals falling within the Expected Error (SARA_MCD43 = 86% and SARA_MOD09 = 77%) than the DB (DB_10km = 60%) AOD product. The SARA_MCD43 AOD exhibited good agreement with DB_10km AOD over desert surfaces areas during high aerosol loading events, with an average correlation coefficient of 0.92, RMSE of 0.09, and mean absolute error (MAE) of 0.08. While SARA_MOD09 AOD over background desert areas is similar to SARA_MCD43, significant differences were observed over high aerosol loading areas and SARA_MOD09 dramatically underestimates AOD. The results demonstrate that: (i) the SARA algorithm (regardless of surface reflectance data used) is more robust than the MODIS AOD products (both DT and DB), and is better able to represent spatial dust aerosol conditions with an accurate AOD over desert areas over a range of low to high dust pollutant levels; (ii) SARA_MCD43 is well correlated with the DB AOD observations and provides AOD with much greater accuracy and reliability than SARA_MOD09 over desert areas during severe dust loading events.