Developing a representative snow-monitoring network in a forested mountain watershed

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