Accurate winter storm forecasting is critical in mountainous regions to reduce road maintenance costs, minimize traffic delays, protect lives and property, and promote tourism. The fine-scale nature of the complex terrain in these regions, however, makes snow challenging to forecast. In this study, we validate snow-to-liquid ratio (SLR) and snowfall amount forecasts derived from the Global Forecast System (GFS), one of the leading weather prediction models used by the National Weather Service, for Snoqualmie Pass, WA where Interstate-90 traverses the Cascade Mountains. Machine learning is used to predict SLR, which is used to derive a snowfall forecast from the GFS forecast of liquid precipitation. After evaluating multiple machine-learning techniques for SLR, this study focuses on random forest (RF) and support vector regression (SVR). The GFS slightly overpredicts precipitation, and application of the RF and SVR models for SLR results in an overprediction of snowfall. A cause for error that may impact these trends is the GFS model biases, which may bias SLR calculations or incorrectly detect the melting layer Future work could evaluate the sources of error in the GFS forecasts and machine-learning SLR models to improve snowfall forecasts at this important mountain location.