The first Gordon Bell Prize for Climate Modeling was presented at SC23 in Denver. The award went to a team led by Sandia National Laboratories that had developed and run a model of the global atmosphere with unprecedentedly high resolution on Oak Ridge National Laboratory’s Frontier exascale supercomputer.
According to the ACM, the Gordon Bell Prize for Climate Modeling “aims to recognize innovative parallel computing contributions toward solving the global climate crisis.” The Sandia-led model was selected based on its potential to impact climate modeling and related fields.
Click here to listen to our podcast interview with Gordon Bell, discussing the prize last year.
“We have created the first global cloud-resolving model to simulate a world’s year of climate in a day,” said Sandia researcher Mark Taylor, the chief computational scientist of the Energy Exascale Earth Systems Model, or E3SM, an eight-lab project led by Lawrence Livermore National Lab and supported by the DOE’s Office of Science for the development of advanced climate models. “We’re ushering in a new era of accuracy.”
The E3SM model simulates critical aspects of Earth’s climate system, predicting potential impact on U.S. conditions in the coming decades, including extreme temperatures, droughts, floods, and rising sea levels. Taylor, who led the Gordon Bell submission, detailed the team’s record-setting demonstration of SCREAM, the Simple Cloud Resolving E3SM. Atmosphere Model, on Oak Ridge National Laboratory’s Frontier supercomputer.
SCREAM is a full-featured atmospheric general-circulation model developed for very fine-resolution simulations on exascale machines, incorporating state-of-the-art parameterizations for fluid dynamics, microphysics, moist turbulence, and radiation. SCREAM, led by Peter Caldwell of LLNL, achieved these novel results through a close collaboration between atmospheric and computational scientists.
SCREAM consists of combining four key innovations:
- Portable C++ innovations, including using the Kokkos programming model, a C++ library for on-node parallelism
- Algorithmic innovations, including formulations that help the core of the model conserve energy with exact time-stepping
- Subgrid physical parameterizations, a class of models used in climate simulations
- I/O Strategy: SCREAM uses the Software for Caching Output and Reads for Parallel I/O (SCORPIO) library and the Adaptable Input Output System I/O (ADIOS) library
Regarding resources, SCREAM used 32768 graphics-processing units (GPUs) on 8192 Frontier nodes and obtained a record-setting performance of 1.26 simulated years per day for realistic cloud-resolving simulations.
The members of the team are Mark A. Taylor, Luca Bertagna, Conrad Clevenger, James G. Foucar, Oksana Guba, Benjamin R. Hillman, Andrew G. Salinger (all of Sandia National Laboratories); Peter M. Caldwell, Aaron S. Donahue, Noel Keen, Christopher R. Terai, Renata B. McCoy, David C. Bader (all of Lawrence Livermore National Laboratory); Jayesh Krishna, Danqing Wu (both of Argonne National Laboratory); Matthew R. Norman, Sarat Sreepathi (both of Oakridge National Laboratory); James B. White III (Hewlett Packard Enterprise); and L. Ruby Leung (Pacific Northwest National Laboratory).