Modeling Seasonal Evolution of Liquid Water Percolation in Maritime and Continental Snowpacks

Abstract

Accurate model representation of liquid water percolation through snow is crucial for snowpack runoff forecasts. Here, we investigate model representation of liquid water percolation with both parameter-based and physically-based routines. Routines for these two approaches were implemented within the SnowModel snow evolution framework and evaluated against point measurements at two maritime sites in Washington State and one continental site in Colorado, USA. Physically-based simulations improved peak snow depth over 3 meters (79%) with respect to the parameter-based routine in the Olympic Mountains of Washington State. Further, improvements in snowpack root-mean-squared-error (RMSE) and snow disappearance timing were also observed in all domains. The physically-based percolation routine also outperformed the parameter-based percolation routine when using parameter-sets from other climates and snow seasons. However, the Olympic Mountains of Washington, USA contained unique liquid water fluxes that were not fully described by the physically-based routine. We hypothesize that flow in this region exhibits evidence of preferential flow paths instead of uniform, matric flow and is therefore not transferable between domains given the modeling architecture.