Spatio-temporally seamless daily AMSR-E/AMSR2 vegetation optical depth products (2002-2022)

Vegetation optical depth (VOD) is an important tool for monitoring vegetation characteristics and plays a crucial role in terrestrial ecosystems. The AMSR-E/2 VOD dataset inverted using the Multi-Channel Collaborative Algorithm (MCCA) has a long time series, but complete coverage of the global land is not possible due to the limitation of the satellite orbital scanning gap. This dataset reconstructs the MCCA-AMSR VOD dataset using a 3D partial convolutional neural network. By updating the mask, the model can extract features from valid regions while ignoring invalid regions, thus improving efficiency. The reconstructed VOD was evaluated spatio-temporally by three methods, and the results show that the reconstructed VOD dataset has higher coverage, accuracy, and reliability. In addition, the reconstructed VOD better reflects the diurnal variation of leaf water potential, which is beneficial for various studies related to drought and vegetated ecosystems.