Improving CFSv2 Snow Water Equivalent Forecasts in the Colorado River Basin with Generalized Analog Regression Downscaling

Abstract Snowpack is the most important water resource for the Colorado River basin (CRB), along with the Great Salt Lake (GSL) basin. Predicting snowpack in this region months ahead presents grand challenges to scientific and operational communities due to a lack of methods and products for snowpack forecasts in the Colorado River basin. Here, we apply the generalized analog regression downscaling (GARD) method to snow water equivalent in the Climate Forecast System, version 2 (CFSv2), model of NOAA (∼100-km atmosphere component) along with the benchmarking 4-km National Snow and Ice Data Center (NSIDC) snow water equivalent (SWE) data. We train GARD with CFSv2 hindcasts, followed by testing the model using the most recent water years 2020–21. The forecasting results based on GARD outperform CFSv2 in terms of common evaluation metrics such as seasonal bias and root-mean-square error across lead months 1–4, especially in the high-elevation areas of the CRB including the Wasatch, Uinta, and Rocky Mountain ranges. The research findings are valuable for decision-makers in these regions for managing water resources and protecting the Colorado River and the Great Salt Lake.