New Integration Approaches for MODIS C6.1 DT and DB Products over Land and Ocean

In the fall of 2017, the newest Moderate Resolution Imaging Spectroradiometer (MODIS) Collection 6.1 (C6.1) aerosol products were available with various refinements and improvements made to both the radiation calibration and Dark Target (DT) and Deep Blue (DB) algorithms. A combined DT and DB dataset (DTB) was also added based on piecewise fixed thresholds using the Normalized Difference Vegetation Index (NDVI) for taking advantage of one's merits. This aerosol products have not been fully evaluated yet. Therefore, this study first comprehensively validates and compares the MODIS C6.1 aerosol optical depth (AOD) products with reference to the Aerosol Robotic Network (AERONET) Level 3 Version 2.0 data at 384 ground stations from 2013 to 2017 over land and ocean. The results suggest that the differences and improvements of three AOD products are rather non-uniform that varies with region. Overall, the DB products show the best performance in most regions at about half of the sites, especially in Europe and North America. Meanwhile, besides bright surfaces (i.e., deserts and arid/semi-arid areas), DB products match more closely with the AERONET AODs than that of DT over medium or densely vegetated areas. While the merged product using NDVI has some improvements over individual ones in general, worse performance is also shown in many cases. A more optimal method is thus wanting. Therefore, aimed at this problem, we develop improved merge approaches to increase the spatiotemporal data coverage and reduce the estimation uncertainty. For this, three tests, i.e., a land-use-type test, a surface-relief test, and an aerosol-type test are performed according to the strengths and weakness of the performances of the DT and DB algorithms with their high-quality assurance retrievals against the AERONET AOD measurements. Based on this, both Terra and Aqua new DTB products are generated and validated against ground measurements from site to global scales, and for varying underlying surfaces and elevated terrains. Results show that more than 90% of the sites now have more data points, and the performances of the new DTB products are improved with an increased percentage of the data falling within the expected error [ ± (0.05 + 15%)] envelope and reduced mean absolute errors and root-mean-square errors compared with official DTB products at most sites. Separate- and equal-number comparisons show that the new DTB products significantly improve both the data coverage and data quality. The new merged products are more accurate and less affected by varying surface structures than the operational products. These results suggest that the improved merge approach is more robust and can be used for generating more accurate global aerosol products.