Uncovering Key Drivers of Colorado River Flow Uncertainty with High-Resolution Models

Abstract The Colorado River is a critical water source for the arid and semi-arid American Southwest, yet its future remains highly uncertain because climate models struggle to represent the complex hydroclimate of its mountainous headwaters. By analyzing a suite of global climate models with grid resolutions from ~250 km to ~25 km, we demonstrate that high-resolution models with improved representation of fine-scale topography provide a much more realistic simulation of upper-elevation snowpack. This snowpack functions as a hydrological buffer, storing precipitation that would otherwise be lost to evapotranspiration and releasing it months later as runoff. In high-elevation terrain, snowpack and runoff are primarily modulated by precipitation rather than temperature. However, as climate warming pushes winter temperatures close to and above the melting threshold, snowpack becomes increasingly sensitive to temperature, shifting precipitation partitioning toward evapotranspiration at the expense of runoff. Coarse-resolution models fail to simulate these key processes. Consequently, the dominant source of uncertainty in future runoff is not the wide inter-model spread in precipitation projections but rather the misrepresentation of snowpack and the resulting biases in runoff sensitivity to warming. Our findings suggest that future declines in Colorado River flow might be more severe than current projections indicate.