Examining Extreme Precipitation Features and Future Changes with Non-hydrostatic CESM at Storm Resolving Scale over the Western US Region

Global Storm Resolving Models (GSRMs) capture the multiscale aspects of weather and climate phenomena, and may therefore provide an improved representation of climate impacts. The primary goal of this research is to assess how GSRM capabilities can enhance our comprehension of hydrological cycle processes and advance the accuracy of precipitation predictions in the scope of western US heavy precipitation events. The proposed work uses a version of the Community Earth System Model (CESM2) that uses the Model for Prediction Across Scales (MPAS) as the non-hydrostatic dynamical core for the atmosphere. This configuration is designed for storm-resolving resolution and has regional refinement capability. We conduct seasonal simulations in current and future climate states designed to probe the hydrological impacts of heavy precipitation over a refined study region at 3 km within a global domain.

Precipitation extremes mainly in the form of atmospheric rivers (ARs) over WUS have important impacts over mountainous areas, which are critical sources of water and play a crucial role in regulating the hydrological cycle. We focus on the process-level understanding of the large-scale hydrological cycle processes and regional hydroclimate impacts based on multi-scale modeling of intense precipitation events. We configure CESM2 with the MPAS non-hydrostatic dynamical core and Community Atmosphere Model (CAM6) physics, and the Community Land Model Version 5.0 (CLM5) and prescribed SST (sea surface temperature) and ice. To achieve the highest resolution (as fine as 3km) at an achievable computational cost, we use the regional refinement capability of CAM-MPAS wherein the grid spacing is 60km for most of the globe, but is reduced to 3km along the west coast of the US (WUS). Experiments at each resolution are initialized from the same climate state remapped to each target resolution.

For the historical simulations, we use the existing five wet-season (October to March) data over WUS from year 1999 to 2004 to accommodate the annual variability. For future SSP370 simulations, the mean monthly SST climatology difference between two time periods (1991-2010 vs. 2071-2080) is applied to the historical SST forcing file without changing other forcings and initial conditions. The computation cost for those future simulations is about 5 million core hours (including all five wet-season simulations for future 60-3km simulations) based on NCAR’s new supercomputer Derecho. We examine precipitation extremes and the changes in AR events' intensity, frequency, and duration. With that, we also investigate the dynamic versus thermodynamic contributions to the intensity of precipitation extremes under different scenario conditions. We will also investigate the mesoscale and orographic features of heavy precipitation events and their impact on local-scale hydrological responses to extreme AR storms captured by the storm-resolving modeling at 3 km. We compare the simulations to look at whether and how warming SST has altered any of those behaviors. By targeting future seasonal simulations vs. historical ones, we aim to better understand heavy precipitation prediction and impact assessments under an accelerated hydrological cycle by applying storm-resolving modeling.