High-resolution models with grid spacings in the range between 1–30 km are paramount for capturing scale interactions and for accurately representing local and regional phenomena such as convection, orographically induced precipitation, Mesoscale Convective Systems (MCS), and tropical cyclones (TCs). Such events not only have large regional impacts but also provide feedback mechanisms for the large-scale flow. Over the last 10-15 years, most high-resolution long-term Earth system studies utilized hydrostatic model designs with grid spacings of around 25 km. However, the emerging Earth system modeling frontier is “Global Storm Resolving Models” (GSRMs) or “Convection-Permitting” Earth system models that are suitable for kilometer-scale computational meshes. These require dynamical cores that solve the nonhydrostatic versions of the governing equations and follow a highly optimized computational design philosophy. These two requirements lie at the heart of the NSF StormSPEED project. NSF StormSPEED will enable the CESM community to conduct nonhydrostatic, computationally-efficient, kilometer-scale Earth system simulations with CESM version 3 (CESM3), the upcoming flagship earth system model of the NSF National Center for Atmospheric Research (NCAR).
One of the initial goals of StormSPEED is to remove the hydrostatic restriction of the Spectral Element dynamical core in CESM3 to enable km-scale modeling. For this capability, StormSPEED will be making use of the nonhydrostatic configuration of the Spectral Element dynamical core which has been developed for DoE's 'Energy Exascale Earth System Model (E3SM)' (Taylor et al., 2020) and the 'Simple Cloud Resolving E3SM Atmosphere Model' (SCREAM, Caldwell et al., 2021; Donahue et al., 2024). This dynamical core has excellent GPU performance and enables km-scale simulations on today's exascale computing architectures (Taylor et al., 2023).
StormSPEED will study the evolution of extreme precipitation events with a nonhydrostatic dynamical core. It thereby extends our previous high-resolution CESM studies that have been conducted by TAMU & NCAR for the NSF-funded MESACLIP (understanding the role of MESoscale Atmosphere – ocean interactions in seasonal-to-decadal CLImate Prediction) project. MESACLIP utilizes a hydrostatic CESM configuration with the Spectral Element dynamical core and an atmospheric grid spacing of 25 km. These MESACLIP studies serve as a baseline for StormSPEED.
An example result of the MESACLIP project is shown in the figure above (courtesy of Dan Fu, TAMU) which displays the DJFMAM extreme daily precipitation over the continental U.S. (CONUS). In particular, the figure displays the NOAA CPC precipitation rate observations (upper left) and compares them to high-resolution (HR; 0.25° atmosphere and land, 0.1° ocean and sea-ice) and low-resolution (LR; all 1° components) CESM simulations plus a 4-km CONUS configuration of NCAR's Weather Research and Forecasting (WRF) model. The CESM HR (MESACLIP) and WRF climatologies closely match the observations which is also highlighted in the probability density diagrams over the two selected regions.