Investigating multiple aerosol optical depth products from MODIS and VIIRS over Asia: Evaluation, comparison, and merging

The first purpose of this paper is to evaluate and compare four aerosol optical depth (AOD) products from the MODerate Resolution Imaging Spectroradiometer (MODIS) and the Visible Infrared Imaging Radiometer Suite (VIIRS) during 2013–2018 in Asia. In our study, a total of 81 AERONET sites are considered and land cover maps are utilized as well. The results show that the AOD product of deep blue from VIIRS (DB_V) achieves the best performance in the study areas, with the R of 0.91 and the RMSE of 0.14. Meanwhile, the deviations for the AOD products of deep blue from MODIS (DB_M), dark target (DT), DB_V, and environmental data record (EDR) periodically fluctuate with different levels as time moves forward. In general, DB_V overcomes others with the smallest overall deviation, while the largest positive and negative deviations are observed in DT and EDR, respectively. The performance of each AOD product is different in the regions with diverse land cover types. Especially, all AOD products will generally underestimate the AOD values in forest; DB_V performs better than DB_M in croplands and urban, while the overestimation of DB_V is larger than that of DB_M in arid lands. The distribution of high AOD values for DT and EDR shows difference in four seasons, which is dominated by multiple factors. With regard to DB_M and DB_V, apart from the seasonal variations, the high AOD values also distribute in arid lands from March to August. For the coverage of valid AOD values, the annual AOD completeness of DB_M and DB_V tends to be large in the Southwest (arid lands). As for DT and EDR, the large annual AOD completeness principally distributes in India, where the primary land cover type is croplands. Next, a novel grid-based merging framework (SL-SGW) is proposed to acquire the AOD product with the best performance and the largest AOD completeness of DB_M, DT, DB_V, and EDR as much as possible. The experiment results (2017–2018) show that the R and the RMSE for the merged AOD product are 0.904 and 0.13, respectively. It's believed that the merging framework could effectively absorb the strengths of DB_M, DT, DB_V, and EDR. In the meantime, the underestimations of the AOD values for all AOD products in forest and the overestimations for DB_M and DB_V in arid lands are both mitigated after merging. The AOD completeness of the merged exceeds those of other AOD products for all land cover types, particularly in croplands and urban.