Snow Redistribution Modelling in Alpine Norway : Validation of SnowModel for a wet, high mountain climate

The redistribution of seasonal snow is an integral part of the processes controlling soil temperature, permafrost, soil moisture and vegetation distribution, and plays an important role in the planning of infrastructure and hydropower production. Models exist that are able to simulate these snow distributions by using available meteorological data. In this study, an extensive dataset of snow distributions has been collected to evaluate the performance of the snow evolution and distribution model SnowModel. Snow distributions are observed at Finse, a high mountain plateau in Norway, at monthly intervals over 2 winter seasons. Ground Penetrating Radar (GPR) has been used to determine snow depth distribution within a 1 km^2 area. The observations show a heterogeneous snow cover in rough terrain, due to wind redistribution. Additionally, 3 meteorological weather stations were installed and have recorded wind speed, wind direction and temperature. SnowModel is implemented for the study area with a spatial resolution of 4 m and 1 h time steps, and model results are compared with the collected validation data. Initial results suggest that SnowModel is unable to reproduce the observed snow distributions under the given climatic conditions. The issues are accredited to 4 sources: (1) The modelled wind distribution does not show sufficient variation to recreate the observed wind distributions, with a mismatch of 42 %, 50 % and -28 % of the observed wind speed from the validation stations. (2) SnowModel recreates hard, immovable snow layers if temperatures exceed 3 ℃, but does not handle immovable layers created by wind-induced mechanical metamorphism. (3) Snow surface density is reset to the new snow density at any solid precipitation event, regardless of snow surface density evolution up until that point. And (4) simulations show that snow is transported out of the model domain without any snow being introduced upwind in the model domain, leading to a loss of snow. The issues may be due to the climate in alpine Southern Norway, when compared with previous implementations. Methods for improving model performance are discussed and implemented, and manage to rectify the loss of snow out of the domain at the cost of the spatial variation in snow depth.