Abstract. Seasonal snowpack dynamics shape the biophysical and societal characteristics of many global regions. However, snowpack accumulation and duration have generally declined in recent decades, largely due to anthropogenic climate change. Mechanistic understanding of snowpack spatiotemporal heterogeneity and climate change impacts will benefit from snow data products that are based on physical principles, simulated at high spatial resolution, and cover large geographic domains. Most existing datasets do not meet these requirements, hindering our ability to understand both contemporary and changing snow regimes and to develop adaptation strategies in regions where snowpack patterns and processes are important components of Earth systems. We developed a computationally efficient process-based snow model, SnowClim, that can be run in the cloud. The model was evaluated and calibrated at Snowpack Telemetry (SNOTEL) sites across the western United States (US), achieving a site-median root-mean-squared error for daily snow water equivalent (SWE) of 64 mm, bias in peak SWE of −2.6 mm, and bias in snow duration of −4.5 d when run hourly. Positive biases were found at sites with mean winter temperature above freezing where the estimation of precipitation phase is prone to errors. The model was applied to the western US (a domain covering 3.1 million square kilometers) using newly developed forcing data created by statistically downscaling pre-industrial, historical, and pseudo-global warming climate data from the Weather Research and Forecasting (WRF) model. The resulting product is the SnowClim dataset, a suite of summary climate and snow metrics, including monthly SWE and snow depth, as well as annual maximum SWE and snow cover duration, for the western US at 210 m spatial resolution (Lute et al., 2021). The physical basis, large extent, and high spatial resolution of this dataset enable novel analyses of changing hydroclimate and its implications for natural and human systems.