Shifting hydroclimate sensitivity of streamflow under warming climate

Climate-driven shifts in snow accumulation and melt are reshaping the streamflow regime in the Sangre De Cristo (SDC) Mountains of the Upper Rio Grande Basin (URGB). Declining snowpack has reduced runoff contributions with altered timing, yet the evolution of streamflow sensitivities across hydroclimate regimes remains unclear. Studies often overlook the gradual changes in the sensitivity of factors driving streamflow. Moreover, existing research typically focuses on either linear or nonlinear relationships between streamflow and hydroclimate drivers in isolation, assuming these relationships are static rather than dynamic. This study investigates how streamflow sensitivity to hydroclimate factors evolves during the baseflow and runoff seasons over four decades (1980–2019) in eight SDC watersheds. Using Multiple Linear Regression (MLR) and Generalized Additive Models (GAM) within a tailored Generalized Information Criterion (GIC), the study contrasts linear and nonlinear hydroclimate controls over the streamflow. Nonlinear GAMs outperformed MLRs during snowmelt-runoff seasons, achieving accuracy up to 66% cross‐validated (CV) R2, whereas linear MLRs adequately described baseflow, predicting up to 62%. Sensitivities are derived from MLR coefficients and GAM partial derivatives. Although the minimum temperature remained a significant predictor, thermal controls reversed over the decades, while heterogeneous influences, such as soil moisture and sublimation, became increasingly important in mediating streamflow. Other factors exhibited notable but sporadic effects, and the variations were modulated by snow fraction, elevation, and latitude. All predictors generally declined in their explanatory power over time. This study provided an adaptive, seasonally resolved modeling framework that uncovered critical insights into dynamic hydroclimatic sensitivities, thereby supporting more effective water resource management.