Publications

Zhao, Y; Huang, B; Liu, DS; He, QQ (2021). A sparse representation-based fusion model for improving daily MODIS C6.1 aerosol products on a 3 km grid. INTERNATIONAL JOURNAL OF REMOTE SENSING, 42(3), 1077-1095.

Abstract
The latest MODerate resolution Imaging Spectroradiometer (MODIS) Collection 6.1 (C6.1) Aerosol Optical Depth (AOD) products provide AOD datasets at 3 km and 10 km spatial resolutions retrieved by the Dark Target (DT) and Deep Blue (DB) algorithms, i.e., DT3K and DB10K, respectively. Despite their respective values in aerosol-related studies, the two AOD products compromise between their spatial resolution and spatial coverage, which limits their utilities in studies that require fine-scale resolution AOD over large geographical areas (e.g., continental or global scale). This study proposes a SParse representation-based Gap Filling model (SPGF) to blend the higher spatial resolution of DT3K AOD and the broader spatial coverage of DB10K AOD, i.e., fill the gaps in DT3K AOD data. Specifically, this model downscales DB10K AOD based on a resolved sparse representation matrix of a learned overcomplete AOD dictionary pair (10 km and 3 km), which introduces the spatial information from DT3K AOD and the spatial coverage from DB10K AOD. To validate the fused 3 km AOD products, we conduct accuracy assessments based on two types of criteria that are evaluated qualitatively and quantitatively to demonstrate the contribution of our data fusion method. The results show that the proposed method accurately blends the spatial resolution of DT3K AOD and the spatial coverage of DB10K AOD. The improved 3 km AOD products extend the spatial coverage over the original DT3K AOD and contain more information content and spatial details than the original DB10K AOD. Consequently, the proposed data fusion model can compensate for the deficiency of existing MODIS C6.1 AOD products in a low-cost manner, which is of great significance for increasing the usage and benefit for more detailed studies related to AOD dynamics.

DOI:
10.1080/01431161.2020.1823040

ISSN:
0143-1161