Researchers supported by the Exascale Computing Project have developed a Multiscale Modelling Framework (MMF) configuration of E3SM, which involves embedding a limited-area cloud resolving model into each column of the global E3SM model. The novel configuration reduces the computational burden normally associated with an MMF. Their work was published in the January 2020 edition of the Journal of Advances in Earth Modeling Science.
The researchers analyzed the US Department of Energy’s Energy Exascale Earth System Model-MMF (E3SM-MMF), which uses finer vertical resolution than previous MMF model studies and a spectral element dynamical core and incorporates techniques to increase model throughput. The brute force approach to improving cloud simulations is to run models at progressively finer scales. However, interprocess communication costs increase as grid spacing and time steps reduce, which slows down computations and increases cost. To solve this problem, the MMF allows the model to “skip over” intermediate scales, which while important for weather prediction, may not be crucial for climate projections. The end result is a climate model with more realistic cloud behavior that runs as fast as traditional models, which allows for decadal-scale experiments.
The researchers’ work addressed top computational development priorities including grid imprinting and software engineering challenges associated with running the MMF on new hardware. The researchers also validated the model against observation data and other models. They are working to document interesting characteristics of the MMF approach that do not appear in the literature. Future work includes climate sensitivity calculations using the MMF at never-attempted scales made possible by the computational efficiencies developed under the project.
Hannah, W. M., C. R. Jones, B. R. Hillman, M. R. Norman, D. C. Bader, M. A. Taylor, L. R. Leung, et al. 2020. “Initial Results From the Super‐Parameterized E3SM.” Journal of Advances in Modeling Earth Systems 12 (1) (January). Portico. doi:10.1029/2019ms001863. http://dx.doi.org/10.1029/2019MS001863.