We present physics-based snowpack simulations for four snow seasons with detailed wet snow avalanche activity records. The distributed, spatially explicit simulations using the Alpine3D and SNOWPACK model show that the simulated snowpack in the release areas of documented wet snow avalanches often exhibits its first wetting of the season on the release day. This first wetting is accompanied in the simulations by liquid water accumulating on capillary barriers, often formed by depth hoar layers. The strongest water accumulations and largest increases in percolation depth are found on the day of avalanche release. For individual avalanche paths, however, this only holds in 25%-30% of the cases. Assuming that the depth of the strongest water accumulation corresponds to the avalanche fracture depth, the avalanche dynamics model RAMMS-Extended was run using simulated snowpack properties as initial conditions in the release area and boundary conditions along the avalanche path. On average, the simulated affected area by the avalanche and runout distance for the release day are statistically significant in closer agreement with the observations than two days before the release. This does not hold for the simulations of 1day before and 1 and 2days after the release. This suggests that fracture depths and the temporal evolution of percolation depths are adequately simulated within a 1-day period. The results show a large potential for distributed snow cover and avalanche dynamics simulations to assess wet snow avalanche hazards, although predictions for individual avalanche paths remain challenging.

Plain Language Summary Recent years have shown an increasing interest in methods to forecast wet snow avalanches. Wet snow avalanche activity is closely related to water flow processes in snow. It is often considered that water accumulating on layer transitions inside the snowpack (e.g., layers with fine grains on top of coarse grains), may locally reduce snow strength and trigger an avalanche. We show results from detailed simulations of liquid water flow in snow in a mountainous area surrounding Davos, Switzerland, for which also detailed avalanche records are available. We find that periods with strong local water accumulations inside the snowpack correspond well with periods of avalanche activity. On average, the strongest water accumulations and the strongest increases in percolations depth in the simulations occur within 1day of the observed avalanche activity. Assuming that the depth of the water accumulation is the fracture depth, avalanche simulations were carried out to assess the resulting avalanche size. Also, here a good correspondence was found between the average simulated and observed runout distances, suggesting that fracture depths are on average adequately simulated. However, for individual avalanche paths, errors in runout distance were regularly considerable, indicating that predictions for individual avalanche paths remain challenging.