Climate simulation with a higher spatial resolution is a long-time goal of climate model development to capture smaller scale physical processes, which are otherwise unresolved in a lower resolution simulation.  Limited by computational resources, increasing grid spacing globally is often not feasible, especially for multi-century simulation campaigns, e.g., Coupled Model Intercomparison Project 6 (CMIP6).  In E3SM version 2 (E3SMv2), we overcame this limit by leveraging the regionally refined model (RRM) configuration, i.e., E3SMv2 North American RRM (NARRM), which has finer horizontal grids centered over North America and consists of 25->100 km atmosphere and land, a 0.125^o river-routing model, and 14->60 km ocean and sea ice.  Combined with other computational performance improvements, such as atmospheric physics grids and a new semi-Lagrangian scheme for passive tracer transport, historical NARRM simulations run at a throughput (12+ simulated years per day) as efficient as the E3SMv1 low-resolution (LR, 100 km) model.  As a result, we managed to deliver a first set of climate production simulations using a fully coupled RRM, including CMIP6 DECK (Diagnosis, Evaluation, and Characterization of Klima) and historical simulations.

In this study we will show that NARRM simulates improved climate within the high-resolution mesh without sacrificing the global performance relative to the LR counterpart by our novel hybrid time step strategy for the atmosphere.  Over the refined NA area, NARRM is generally superior to LR, including for precipitation and clouds over the contiguous US (CONUS), summertime marine stratocumulus clouds off the coast of California, liquid and ice phase clouds near the North Pole region, and spatial variability in land hydrological processes.  Two main reasons for these improvements are: 1) better resolved topography and 2) RRM meshes in multiple components.

One of the primary motivations for pushing higher resolution climate simulation is to potentially better capture extremes.  Compared to LR, NARRM shows a particularly improved skill when simulating the extratropical cyclone activities along the oceanic storm tracks and over the mountain range to the east of the Rocky Mountains, and suggests a higher contribution of human-induced warming in extreme heat events over western NA.  Lastly, we take advantage of E3SM’s non-hydrostatic dynamical core to extend RRM to refine California at ~3 km.  We will demonstrate the model sensitivity in simulating pyrocumulonimbus during the record-high 2020 Creek fire.  Such applications indicate great potential of RRM for studying extremes at local to regional scales.