Published April 24, 2026 | Version v1

Global gridded (0.5° × 0.5°) groundwater storage from a physics-constrained neural network (PCNN)-based data assimilation framework

Authors/Creators

Description

This dataset presents a global, gridded (0.5° × 0.5°) monthly groundwater product from Apr 2002 to Dec 2023 derived through the integration of WGHM-2.2e and GRACE observations within a physics-constrained neural network (PCNN)-based data assimilation framework. Validation against in situ groundwater observations from 27,602 wells demonstrates that PCNN-GWS exhibits improved correlation relative to WGHM-2.2e. The dataset further demonstrates advantages over existing global groundwater products derived from traditional hydrologic data assimilation techniques and GRACE–land surface model (LSM) residual methods.

Files

Files (236.4 MB)

Name Size
md5:9a63b86709d2851ae1a28948e47de24e
236.4 MB Download

Additional details

References

  • Global gridded (0.5° × 0.5°) groundwater storage from a physics-constrained neural network (PCNN)-based data assimilation framework