Assessing CESM scalability for hierarchical model ensembles

Ryan Sriver, University of Illinois at Urbana-Champaign

Usage Details

We plan to optimize code performance and test the scalability of the Community Earth System Model (CESM) across a wide range of configurations, component sets, and model complexities. The results of this analysis will serve as the foundation for developing new ensemble frameworks to assess the predictability of complex 3-D atmospheric behavior using reduced complexity versions of CESM. The modeling tasks will be performed by co-PI Li, who is a graduate student working with PI Sriver. She has extensive experience running CESM on Blue Waters, and she will be fully-funded to work on this project through an external grant. The anticipated expense of the proposed simulations is considerably outside the capabilities of traditional XSEDE resources, and it is for this reason we are requesting use of Blue Waters to conduct this experiment. The proposed runs will provide critical first steps into developing new cutting-edge research initiatives in petascale hierarchical model ensembles using CESM, which has broad-scale relevance to fields such as: climate science, atmospheric dynamics, machine learning, uncertainty quantification, emulation, and statistical estimation.