Inter-Model Diagnostics for Two Snow Models Across Multiple Western U.S. Locations and Implications for Management

Abstract:

In the western United States, water resource management is increasingly reliant on numerical modeling of hydrological processes, namely snow accumulation and ablation. We seek to advance a framework for providing model diagnostics for such systems by combining an improved understanding of model structural differences (i.e., conceptual vs. physically based) and parameter sensitivities. The two snow models used in this study are SNOW-17, a conceptual degree-day model, and the Variable Infiltration Capacity (VIC) snow model, which is physically based and solves the full water and energy balances. To better understand the performance of these models, several approaches will be used. For the conceptual model, global sensitivity analysis methods (e.g., Sobol’ and Method of Morris), and a multi-objective calibration will be applied to identify important parameters and show calibrated parameter values. For the physically based model, we will contribute a novel exploration of some parameters that can be adjusted within the model, including the liquid water holding capacity, the density of newly fallen snow, and the snow roughness. Additionally, the VIC model will be run with explicit radiation inputs at selected sites. For each model run, snow sensitivities and errors (i.e., snow water equivalent results) will be translated into estimated changes in annual water yield for the study areas. Accurately predicting water yield is essential for water management, and it is used here as a practical measure to determine the importance of model parameter sensitivity and calibration. The analysis will be conducted across a range of snow-dominated locations representing a variety of climates across the western United States (e.g. continental, maritime, alpine).