The first 1 km daily L‑VOD dataset over China: A downscaling framework based on a differencing paradigm integrated with physical constraints (PhyDiF)

Abstract

This Zenodo repository provides the daily, 1-km resolution Vegetation Optical Depth (VOD) downscaling product generated for China (April 2015 to March 2021) using the Physics-informed Differencing Downscaling Framework (PhyDiF) introduced in this study. PhyDiF innovatively integrates the differencing paradigm with a physical constraint mechanism based on the Radiative Transfer Model (RTM). Building upon a data-driven model, it introduces the differencing concept from traditional spatiotemporal fusion models to establish a differencing-based downscaling framework. Concurrently, by designing a joint loss function, the model ensures physical consistency with brightness temperature data while maintaining the accuracy of VOD retrieval. PhyDiF effectively overcomes the difficulty of unified VOD modeling at large regional scales and addresses the lack of physical interpretability inherent in purely data-driven downscaling methods. This work was supported by National Key Research and Development Project of China (2025YFE0103300), Natural Science Foundation of China (42471349; 42461144214), Guangdong Basic and Applied Research Foundation (2024A1515030078), Shenzhen Science and Technology Program (JCYJ20240813114013017), Natural Science Foundation of Wuhan (2024040801020279), and Guided Project of Hubei Provincial Department of Education (B2023246). Hongliang Ma was supported by Open Fund of State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University (Grant No. 24R01) and Villum Young Investigator (Grant 53048).