Modeling Spatially Heterogeneous Nonstationary Watershed Response in the Snow-Dominated Rio Grande Headwaters

Hydrologic models commonly assume stationary watershed behavior, although evolving environmental conditions can reduce parameter transferability. This study investigated temporal and spatial variability in the snow-dominated Rio Grande headwaters near Del Norte, Colorado, during 1980–2020 using the Soil and Water Assessment Tool (SWAT). A stationary calibration was compared with 31 overlapping 11-year moving-window calibrations incorporating updated land-cover maps. Calibrated snow and reservoir parameters were held fixed, while runoff-related parameters varied among windows. Moving-window parameterizations improved out-of-sample streamflow representation across 10 matched, overlapping three-year validation periods during 2009–2020, with mean Kling–Gupta efficiency (KGE) gains of 0.18–0.25 across the evaluated gauges. From 1985 to 2020, forest area declined by approximately 18%, while shrubland, herbaceous cover, and developed land increased by about 47%, 20%, and 23%, respectively. At the RIOWAGCO gauge, the calibrated relative adjustment to the Soil Conservation Service runoff curve number for antecedent moisture condition II (CN2) showed an increasing tendency that coincided with persistent, directional land-cover transitions. In contrast, the contributing catchment at Del Norte exhibited no comparable monotonic CN2-adjustment tendency despite substantial landscape change. These contrasting trajectories indicate spatial heterogeneity in effective parameter behavior. Findings indicate that stationary parameter transferability cannot be assumed across evolving headwater systems. Time-varying parameter trajectories, interpreted alongside land-cover changes, provide diagnostic evidence consistent with spatially heterogeneous hydrologic nonstationarity.