Improving aerosol retrieval from FY-3D/MERSI by parameterizing a binary quadratic spectral surface reflectance model with urban percentage and vegetation index

Abstract

The Medium Resolution Spectral Imager (MERSI) Dark Target (DT) algorithm, adapted from the Moderate Resolution Imaging Spectroradiometer (MODIS) DT algorithm, demonstrated aerosol retrieval performance similar to that of MODIS Collection 6.1 (C6.1) DT products. However, discrepancies arose in urban areas, where MERSI retrievals tended to exhibit noticeable positive biases, primarily due to the inadequate representation of urban surface reflectance. To address this issue, this study developed an enhanced algorithm by parameterizing a binary quadratic spectral surface reflectance model using urban percentage (UP) and vegetation index, named the MERSI DT_UP algorithm. Global validation against Aerosol Robotic Network (AERONET) measurements showed that MERSI DT_UP retrievals in 2019 exhibited greater accuracy than MERSI DT retrievals. The correlation coefficient between satellite retrievals and AERONET for MERSI DT_UP (R = 0.882) was higher than that for MERSI DT (R = 0.875), and the mean bias was reduced to −0.001 from 0.018. Additionally, the percentage of matchups falling within the expected error envelope (within EE%) of ± (0.05 + 0.2τ) increased by 1.5 %. Across different urbanization levels, MERSI DT_UP retrievals showed significant reductions in mean biases (ranging from 0.003 to 0.217) and notable increases in within EE% (ranging from 2.2 % to 49.3 %) compared to original results. Moreover, improvements were also observed on moderately vegetated and moderately bright surfaces, as demonstrated by the error dependence analysis. The proposed surface reflectance model supports the algorithm development for MERSI and holds potential for application to other sensors.