A new article titled “DeepRec: Global Terrestrial Water Storage Reconstruction Since 1941 Using a Spatiotemporal-Aware Deep Learning Model” has been published in the “Journal of Geophysical Research: Machine Learning and Computation”.
The study presents “DeepRec”, a spatiotemporal deep learning framework for reconstructing global terrestrial water storage (TWS) variations from 1941 to the present. The approach addresses gaps in long-term observational records by learning spatial and temporal dependencies from multiple data sources, resulting in a consistent historical TWS dataset.
The research is based on a Master’s thesis by Luis Gentner, who studied aerospace engineering and developed an interest in geodesy during his study period. The work was jointly supervised at the University of Stuttgart and ETH Zurich, reflecting interdisciplinary training across engineering, geodesy, and hydrology.
The reconstructed dataset supports the analysis of long-term hydroclimatic variability and is relevant for studies on droughts, floods, and climate-related changes in water resources.
The full article is available via its DOI: 10.1029/2025JH000889
