Merging of MISR and MODIS Level-2 Aerosol Optical Depth Products Using Bayesian Principles

Atmospheric aerosol loading together with its absorption and scattering properties, plays an important role in the radiative forcing of the Earth's changing climate. Aerosol loading is measured in the units of Aerosol Optical Depth (AOD) routinely retrieved by many spaceborne sensors including the MISR and MODIS. AOD is widely used in aerosol and air quality characterization as well as aerosol-climate effect assessment. The differences in satellite-based retrieval algorithms, spatial/temporal resolution of the radiances/products, sampling, and cloud-screening schemes have led to significant differences among AOD products retrieved from various sensors. A fusion of AOD datasets retrievals from multiple satellites is therefore desirable, which utilizes the strengths of the individual products, and improves the accuracy when compared with the ground-based AOD data e.g., from the AERONET.In view of this, we have made an attempt to produce a "merged" Level-2 AOD product based on MISR and MODIS datasets, using Bayesian approach which utilizes the error distribution from the AERONET AOD. The "merged" AOD dataset is inter-compared with MISR and MODIS AOD products along with the AERONET data over the Indo-Gangetic Plains. It is found that the RMSE of the merged AOD (0.07-0.13) data is lower than both MISR (0.11 - 0.21) and MODIS (0.15 - 0.21) retrievals. The merged AOD data dataset has higher correlation with AERONET data (0.89 - 0.92), compared to MISR (0.81 - 0.86) and MODIS (0.69 - 0.86). We have also compared the AOD products using their Expected Error (EE) envelope. It is found that the percent of the "merged" AOD (77.3% - 84.9%) falling within EE envelope is greater than MISR (58.4% - 75.0%) and MODIS (49.1% - 69.8%). The comparison of "merged" AOD, and that of individual MISR and MODIS products with the AERONET data in terms of RMSE, correlation coefficient and EE envelope suggests the overall significantly improved accuracy of the "merged" AOD dataset. Our methodology and the resulting dataset are especially relevant in the scenario of fusing multi-sensor retrievals for producing long-term satellite-based climate data records.