An international team of climate researchers has been awarded the inaugural 2023 ACM Gordon Bell Prize for Climate Modelling at the SC23 conference. The award recognises their pioneering work on the project titled "The Simple Cloud-Resolving E3SM Atmosphere Model Running on the Frontier Exascale System".
The groundbreaking SCREAM (Simple Cloud Resolving E3SM Atmosphere Model) represents a first in climate modelling, being the first to run on the Frontier Exascale system. With its help, the team achieved a record-setting performance of 1.26 simulated years per day in a realistic cloud resolving simulation. The vast data this yielded is a crucial resource for climate scientists seeking to comprehend the behaviour and effects of deep convective clouds, a vital part of the planet's climate control system.
Designed as a comprehensive tool, SCREAM amalgamates four significant innovations. These include developments in Portable C++, making use of the Kokkos programming model and a C++ library for on-node parallelism. In terms of modelling, algorithmic advancements aid in the core model conserving energy with precise time stepping, while a sub grid physical parameterisation is employed - a class of model commonly used in climate simulations. For Input/Output, SCREAM utilises the SCORPIO and ADIOS libraries.
Launched this year, the ACM Gordon Bell Prize for Climate Modelling seeks to acknowledge the significant contributions of climate scientists and software engineers to our ability to predict climate activity. In an extended role, the prize also hopes to incite further scientific research and endeavours towards addressing the global climate crisis. Funded by Gordon Bell, the award includes a $10,000 prize.
The award-winning modelling tool, SCREAM, allows climate scientists to study convective clouds, formed by warmer air rising in a less dense atmosphere, with unprecedented detail. Understanding these deep convective clouds, which can be kilometers thick, is vital as they drive tropical overturning circulations and influence energy transfer over a substantial portion of the planet.
The winning team of nineteen members derived from various National Laboratories as well as Hewlett Packard Enterprise, demonstrated that through the exceptional power of newly introduced exascale supercomputers and the SCREAM model, it's now possible to efficiently generate accurate simulations of deep convective clouds. This achievement represents a significant leap forward for climate science.
The Frontier’s exascale computer, developed by Oak Ridge National Laboratory in 2022, is a paramount tool in this research. This latest generation of ultra-powerful supercomputers is capable of performing a quintillion operations per second. The winning team utilised the Frontier to set a record, employing 32768 GPUs on 8192 nodes, resulting in an impressive 1.26 simulated years of climate data per day.
This achievement in cloud simulation represents a landmark progress for climate science. The team contend that the data accrued will facilitate reliable forecasting of future climate impacts, which have traditionally been elusive due to their localized occurrence in time and space. The ACM Gordon Bell Prize for Climate Modelling was presented during the International Conference for High Performance Computing, Networking, Storage and Analysis (SC23), held in Denver, Colorado.