Synergistic data fusion of multimodal AOD and air quality data for near real-time full coverage air pollution assessment

Abstract

Data gaps in satellite aerosol optical depth (AOD) retrievals pose a huge challenge in near real-time air quality assessment. Here, we present a multimodal aerosol data fusion approach to integrate multisource AOD and air quality data for the generation of full coverage AOD maps at hourly resolution. Specifically, data gaps in each Himawari-8 AOD snapshot were partially filled by merging all available daytime AOD snapshots, and these partially gap-filled AOD maps were then fused with coarse yet spatially complete numerical AOD simulations to generate full coverage AOD imageries. Ground-based air quality measurements, including concentrations of PM_2.5, PM₁₀, NO₂, and SO_2, were simultaneously assimilated into gridded AOD fields to enhance the overall data accuracy. A practical implementation of the proposed method was illustrated by generating hourly full-coverage AOD maps in China from 2015 to 2020, and the validation results indicate this new AOD dataset agreed well with ground-based AOD measurements (R = 0.83), from which a ubiquitous AOD decreasing trend was revealed, especially during the noontime. Moreover, the hourly resolution and full-coverage advantages of this AOD dataset allow us to better assess spatiotemporal variations of PM₁₀ and PM_2.5 pollution that occurred in China. Overall, the proposed method paves a new way as big data analytics to advance regional air pollution assessment given the full coverage capacity and enhanced accuracy of the resulting AOD and PM concentration data.

Keywords: AOD; Air quality management; Data fusion; Haze pollution; Himawari; Optimal interpolation.