Evaluation of MODIS Deep Blue and MAIAC AOD products over typical ecologically fragile areas in Central Asia and surrounding regions across different land cover types

Accurate aerosol optical depth (AOD) retrieval remains challenging in ecologically fragile areas because bright and heterogeneous land surfaces, complex aerosol types, and frequent dust activity can introduce substantial uncertainties into satellite products. In this study, the MODIS Collection 6.1 Deep Blue (DB) and MAIAC AOD products were evaluated against AERONET observations over a typical ecologically fragile area in Central Asia and surrounding regions (23–56°N, 46–97°E) during 2018–2022, with particular attention to seasonal, land-cover-dependent, and aerosol-loading-related uncertainties. The quantitative evaluation revealed that MAIAC AOD outperformed DB, as evidenced by a lower RMSE (0.157 vs. 0.196). Both products showed AOD underestimation, which was related to highly reflective surfaces and inaccurate aerosol model assumptions in the retrieval algorithms. Using MCD12C1 land cover data, we found that MAIAC AOD outperformed DB for all surface types except urban and built-up lands. Moreover, compared with sites characterized by bright surfaces, both products exhibited improved AOD accuracy over moderately bright and dark surfaces, with MAIAC showing greater improvement, achieving an average RMSE reduction of 0.09, whereas the corresponding reduction for DB was 0.087. Further analysis of the impact of aerosol loading on AOD accuracy in the ecologically fragile study area showed that MAIAC underestimation over barren and savanna sites increased systematically with AOD and followed a strong linear relationship (R 2 > 0.9). Within the AOD range of 0–0.7, each 0.1 increase in AOD increased the MAIAC underestimation by 0.074 over barren surfaces, compared with 0.02 over savanna surfaces, indicating that retrieval bias is more sensitive to aerosol loading over bright surfaces. This pattern has not been reported previously.