Across most of the continental United States, winter has been the fastest warming season over the past 50 years. The warming is more pronounced in seasonally snow-covered regions, especially the northeastern United States and the upper Midwest. With warmer winters, snowy regions have seen a reduction in snow cover duration, reduced snow water storage, more winter precipitation falling as rain instead of snow, and more frequent mid-winter thaw events. The rapid pace of winter climate change is hypothesized to have large impacts on the natural functioning of ecosystems and corresponding ecosystem services. Next-generation climate modeling provides the framework to run global coupled land–atmosphere simulations with refined regions at higher spatial resolutions critical to predict snowpack development at a fraction of the computational cost of globally uniform high-resolution grids. Here, we present results from the global coupled land-atmosphere Variable Resolution Community Earth System Model (VR-CESM2) under historical (1984-2014) conditions and for future climate pathways, with a refined high-resolution mesh (0.125^o, or 14 km) over the Contiguous United States (CONUS).

In comparison with gridded observational products, the VR-CESM broadly captures historical trends and patterns in snow cover duration, timing of snow onset and disappearance, temperature and precipitation. Under moderate warming, the largest declines in snow season duration occur in the upper Midwest (20-40% decline), with more resilient snowpacks in the northern Rocky Mountains (0-20% decline). With higher warming levels, the mid-Atlantic and southern New England are snow-free under higher warming levels. These findings are consistent with previous results for the northeastern US produced by forcing a regional hydrologic model with the CMIP5 Localized Climate Analogs (LOCA) statistically downscaled (0.0625º, or ~7 km) gridded temperature and precipitation. As the snow-covered season contracts and snow water storage declines, we explore the impacts on regional hydrology in the VR-CESM2 simulations and compare them to a scalable gridded hydrologic model forced with CMIP6 LOCA2 data for lower, mid-range, and higher climate scenarios.