A hybrid physics–ML framework for integrating groundwater dynamics into land surface modeling

Abstract. Three-dimensional groundwater dynamics play a critical role in regulating land–atmosphere interactions, yet resolving three-dimensional subsurface flow processes at large scales remains computationally prohibitive. Here we present a hybrid coupling framework that enables the integration of three-dimensional groundwater processes into land surface modeling at substantially reduced computational cost. The framework replaces the physics-based groundwater solver with a deep learning surrogate while preserving the original coupling interface, providing a practical pathway for incorporating groundwater dynamics into Earth system simulations. A key feature of the framework is an error-control strategy based on a free-drainage lower bound, which approximates the treatment of subsurface processes in conventional land surface models where groundwater feedback is largely neglected. The hybrid solution is considered acceptable as long as its deviation remains within this free-drainage bound, with a user-defined threshold providing additional control over acceptable error levels, enabling flexible, application-dependent control of model fidelity. The framework is demonstrated in a ~34,000 km² watershed in the Pearl River Basin, China, achieving an approximately 20× speedup while maintaining strong agreement with the physics-based reference. Over a full-year hourly simulation, the median water table depth error is within 0.5 m and the domain-averaged latent heat flux reaches a Kling–Gupta efficiency of 0.965. This study demonstrates the feasibility of hybrid surrogate–physics coupling for representing groundwater processes and provides a flexible foundation for multi-timescale, on-demand simulations, with potential for extension to diverse hydroclimatic settings and integration into Earth system modeling frameworks.