Bisht et al. (2025) Development of a River Dynamical Core for E3SM to simulate compound flooding on Exascale-class heterogeneous supercomputers

Identification

• Journal: Environmental Modelling & Software
• Year: 2025
• Date: 2025-12-16
• Authors: Gautam Bisht, Donghui Xu, Jeffrey Johnson, Jed Brown, Matthew Knepley, Mark Adams, Dongyu Feng, Dalei Hao, Darren Engwirda, Mukesh Kumar, Zeli Tan
• DOI: 10.1016/j.envsoft.2025.106804

Research Groups

• Atmospheric, Climate, & Earth Sciences Division, Pacific Northwest National Laboratory, Richland, WA, USA
• Cohere Consulting, LLC, Seattle, WA, USA
• Department of Computer Science, University of Colorado Boulder, Boulder, CO, USA
• Computer Science and Engineering, University of Buffalo, Buffalo, NY, USA
• Applied Mathematics & Computational Research Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
• Fluid Dynamics and Solid Mechanics Group (T-3), Los Alamos National Laboratory, Los Alamos, NM, USA
• Environment, Commonwealth Scientific and Industrial Research Organisation, Hobart, TAS, Australia
• Civil, Construction and Environmental Engineering, University of Alabama, Tuscaloosa, AL, USA

Short Summary

This paper introduces the River Dynamical Core (RDycore), an open-source 2D shallow water equation library for the Energy Exascale Earth System Model (E3SM), designed to enable kilometer-scale flood simulations on exascale supercomputers with significant GPU acceleration. It was validated for various problems and demonstrated successful coupling with E3SM for simulating compound flooding events.

Objective

• To develop and validate the River Dynamical Core (RDycore), an open-source 2D shallow water equation library, for the Energy Exascale Earth System Model (E3SM) to enable physically-consistent, kilometer-scale flood simulations on exascale-class heterogeneous supercomputers.

Study Configuration

• Spatial Scale: Kilometer-scale flood simulations; problem with 471 million grid cells.
• Temporal Scale: Multiple 5-day flooding simulations.

Methodology and Data

• Models used: River Dynamical Core (RDycore), Energy Exascale Earth System Model (E3SM), PETSc library, libCEED library.
• Data sources: Analytical solutions, manufactured solutions, a well-studied dam break problem for validation; five precipitation datasets for Hurricane Harvey simulations.

Main Results

• RDycore was successfully validated against analytical, manufactured, and a well-studied dam break problem.
• For a problem involving 471 million grid cells, RDycore achieved a speedup of 6.6 times on GPUs compared to CPUs on the Perlmutter supercomputer.
• For the same problem size, RDycore achieved a speedup of 7.6 times on GPUs compared to CPUs on the Frontier supercomputer.
• One-way E3SM-RDycore coupling was demonstrated by performing multiple 5-day flooding simulations during Hurricane Harvey, driven by five different precipitation datasets.

Contributions

• Development of RDycore, an open-source, 2D shallow water equation library, providing a foundational step for kilometer-scale river dynamics simulations within E3SM.
• Achieved hardware and algorithmic portability for river dynamics on heterogeneous computing architectures (CPUs and GPUs) within E3SM.
• Demonstrated significant computational efficiency gains (up to 7.6x speedup on GPUs) on exascale-class supercomputers.
• Enables more accurate and physically-consistent quantification of compound flood risks in global Earth System Models.

Funding

• Not specified in the provided text.

Citation

@article{Bisht2025Development,
  author = {Bisht, Gautam and Xu, Donghui and Johnson, Jeffrey and Brown, Jed and Knepley, Matthew and Adams, Mark and Feng, Dongyu and Hao, Dalei and Engwirda, Darren and Kumar, Mukesh and Tan, Zeli},
  title = {Development of a River Dynamical Core for E3SM to simulate compound flooding on Exascale-class heterogeneous supercomputers},
  journal = {Environmental Modelling & Software},
  year = {2025},
  doi = {10.1016/j.envsoft.2025.106804},
  url = {https://doi.org/10.1016/j.envsoft.2025.106804}
}