Author(s): Na Zheng; Zhi Li; Daniel Caviedes-Voullieme; Mario Morales-Hernandez

Linked Author(s): Mario Morales-Hernández, Daniel Caviedes-Voullieme

Keywords: Groundwater model; High-performance computing; Kokkos; Richards solver

Abstract: Groundwater is vital for water security and aquatic ecosystems. Numerical simulation of groundwater flow, especially variably-saturated groundwater flow, is computational expensive due to the nonlinearity of the governing equations. High-performance computing (HPC) enhances the computational efficiency of traditional groundwater models. We propose an efficient variably-saturated groundwater model, named SERGHEI-GW, that solves the 3D Richards equation. SERGHEI-GW is developed based on the SERGHEI model system and under the Kokkos parallelization framework, which allows flexible model deployment on GPUs and CPUs without modifying the source code. The proposed groundwater model could be potentially coupled with SERGHEI shallow water module for catchment-scale surface-subsurface flow simulation. We present the model formulations and the testing results against existing benchmark problems and groundwater models. Notably, on a single desktop GPU, the computational efficiency of SERGHEI-GW exceeds serial CPU execution by over 200 times. SERGHEI-GW offers researchers a reliable, fast, and adaptable tool for simulating variably-saturated groundwater flow. Relevant for rapidly modeling complex hydrological processes and responding to extreme events induced by global climate change.

DOI: https://doi.org/10.3850/iahr-hic2483430201-325

Year: 2024