A New Multiple-layer Snow Model for Land Surface Models

Snow is an important component of the land surface energy and water budgets. In the western U.S., mountain snowmelt accounts for 70-90% of the annual streamflow. The ability to prescribe and predict mountain snowpack is critically important to hydrologic and atmospheric predictions. Recent studies have shown that significant underestimations of snow water equivalent (SWE) by land surface models exist for mountainous regions, when compared to observations. Our analysis has shown that uncertainties in the forcings due to scale incompatibility and possibly other factors cannot explain the majority of the large SWE underestimation found between the model and SNOpack TELemetry (SNOTEL) for mountainous regions, especially for moderate and deep snowpacks. Motivated by the significant underestimations of SWE, especially during the snow melt process, we develop a new snow model which includes a more complete description of the snow physics of the heat transfer processes (e.g., heat conduction, heat transfer by vapor diffusion, liquid water retention, extinction of short wave radiation, melting and refreezing, etc.) within a snow pack. A flexible time-varying multi-layer approach is employed for the new snow model. The new layer structure and physical processes of the new snow model allow the snow properties of each of the multiple snow layers to be simulated more accurately, especially for deep snowpacks, where current models do not perform well. This new model is coupled with the Variable Infiltration Capacity (VIC) land surface model and will be tested using 18-year observed data available in the Valdai water-balance research site in Russia for shallow snowpacks, and data from the California SNOTEL sites for deep snowpacks.