Abstract. A challenge in establishing new ground-based stations for monitoring snowpack accumulation and ablation is to locate the sites in areas that represent the key processes affecting snow accumulation and ablation. This is especially challenging in forested montane watersheds where the combined effects of terrain, climate, and land cover affect seasonal snowpack. We present a coupled modeling approach used to objectively identify representative snow-monitoring locations in a forested watershed in the western Oregon Cascades mountain range. We used a binary regression tree (BRT) non-parametric statistical model to classify peak snow water equivalent (SWE) based on physiographic landscape characteristics in an average snow year, an above-average snow year, and a below-average snow year. Training data for the BRT classification were derived using spatially distributed estimates of SWE from a validated physically based model of snow evolution. The optimal BRT model showed that elevation and land cover type were the most significant drivers of spatial variability in peak SWE across the watershed (R²  =  0.93, p value  <  0.01). Geospatial elevation and land cover data were used to map the BRT-derived snow classes across the watershed. Specific snow-monitoring sites were selected randomly within the dominant BRT-derived snow classes to capture the range of spatial variability in snowpack conditions in the McKenzie River basin. The Forest Elevational Snow Transect (ForEST) is a result of this coupled modeling approach and represents combinations of forested and open land cover types at low, mid-, and high elevations. After 5 years of snowpack monitoring, the ForEST network provides a valuable and detailed dataset of snow accumulation, snow ablation, and snowpack energy balance in forested and open sites from the rain–snow transition zone to the upper seasonal snow zone in the western Oregon Cascades.