Lawrence Berkeley National Laboratory

Abstract. Projecting the spatio-temporal changes to water resources under a no-analog future climate requires physically-based integrated hydrologic models, which simulate the transfer of water and energy across the earth's surface. These models show promise in the context of unprecedented climate extremes given their reliance on the underlying physics of the system as opposed to empirical relationships. However, these techniques are plagued by several sources of uncertainty, including the inaccuracy of input datasets such as meteorological forcing. These datasets, usually derived from climate models or satellite-based products, typically have a resolution of several kilometers, while hydrologic metrics of interest (e.g. discharge, groundwater levels) require a resolution at much smaller scales. In this work, a high-resolution watershed model is forced with various resolutions (0.5 to 40.5 km) of meteorological forcing generated by a dynamical downscaling analysis based on a regional climate model (WRF) to assess how the uncertainties associated with the spatial resolution of meteorological forcing affect the simulated hydrology. The Cosumnes watershed, which spans the Sierra Nevada and Central Valley interface of California (USA), exhibits semi-natural flow conditions due to its rare un-dammed river basin and is used here as a testbed to illustrate potential impacts on snow accumulation and snowmelt, surface runoff, infiltration, evapotranspiration, and groundwater levels. Results show that localized biases in groundwater levels can be as large as 5–10 m and that other metric biases (e.g. ET and snowpack dynamics) are seasonally and spatially-dependent, but can have serious implications for model calibration and ultimately water management decisions.