A seasonal snowpack model forced with dynamically downscaled forcing data resolves hydrologically relevant accumulation patterns Provisionally Accepted

Justine Berg¹ Dylan Reynolds^{2, 3} Louis Quéno² Tobias Jonas² Michael Lehning^{2, 3} Rebecca Mott^2*

• ¹University of Bern, Switzerland
• ²WSL Institute for Snow and Avalanche Research SLF, Switzerland
• ³Swiss Federal Institute of Technology Lausanne, Switzerland

The Mountain snowpack stores months of winter precipitation at high elevations, supplying snowmelt to lowland areas in drier seasons for agriculture and human consumption worldwide.Accurate seasonal predictions of the snowpack are thus of great importance, but such forecasts suffer from major challenges such as resolving interactions between forcing variables at high spatial resolutions. To test novel approaches to resolve these processes, seasonal snowpack simulations are run at different grid resolutions (50 m, 100 m, 250 m) and with variable forcing data for the water year 2016/2017. COSMO-1E data is either dynamically downscaled with the High-resolution Intermediate Complexity Atmospheric Research (HICAR) model or statistically downscaled to provide forcing data for snowpack simulations with the Flexible Snowpack Model (FSM2oshd). Simulations covering complex terrain in the Swiss Alps are carried out with the operational settings of the FSM2oshd model or with a model extension including wind-and gravitational-induced snow transport (FSM2trans). The simulated snow height is evaluated against observed snow height collected during LiDAR flights in spring 2017. Spatial snow accumulation patterns and snow height distribution are best matched with simulations using dynamically downscaled data and the FSM2trans model extension, indicating the importance of both accurate meteorological forcing data and snow transport schemes. This study demonstrates for the first time the effects of applying dynamical downscaling schemes to snowpack simulations at the seasonal and catchment scale.