Development and validation of a dynamically downscaled historical dataset over the Western United States

Abstract:

Dynamical downscaling is an important tool for producing realistic estimates of the regional impacts of climate change. As part of an ongoing project to downscale several CMIP6 GCMs through the end of the century over the Western United States, here we share the results from validating our downscaling process with observational data over a historical period. Using the Weather Research and Forecasting (WRF) model, we produced a 40-yr historical baseline driven by ERA5 reanalysis at 9km resolution over the Western United States, and at 3km resolution over California and a portion of the Northern Rockies. The results were evaluated against gridded observational data using PRISM, DAYMET, and TRMM, as well as station data from SNOTEL. Our choice of WRF parameterizations is informed by a series of sensitivity tests, and in particular we found that model performance was greatly affected by changes to the nudged parameters and configuration of the nested grids. Additionally, we evaluate the performance of downscaling to a convective-permitting 3km scale at reproducing extreme precipitation events and accurate snowpack in regions of complex topography. These results give us confidence in the performance of our ongoing downscaling of CMIP6 GCMs, and provide a validated high-resolution historical baseline that enables the study of a multitude of climate impacts.