Combining snowpack modeling and terrestrial laser scanner observations improves the simulation of small scale snow dynamics

Abstract

Summary Accurately determining the snowpack distribution in mountain areas is complex because of the difficulty of establishing over large areas the spatial distribution of all variables that define the state of the snowpack at any particular time. In this study we used distributed snowpack simulations that were corrected throughout the snow season using snow depth distributions measured using a terrestrial laser scanner (TLS). This enabled us to obtain a more realistic view of the small scale spatial evolution of the mass and depth of the snowpack. Several TLS snow depth data acquisitions were made during the 2012, 2013 and 2014 snow seasons in a small catchment (55 ha) located in the central Spanish Pyrenees. The Crocus snowpack model was used to simulate the snowpack evolution on a 5-m grid, based on downscaled meteorological variables obtained using SAFRAN reanalysis. The simulation was stopped when a snow depth distribution map measured using the TLS was available, and the modeled snow depth distribution was adjusted to match the observed snow depth. The snow simulation was then restarted and run until the next TLS acquisition was available. Prior to matching the simulation and observational data, both snowpack distributions were compared. The results for the three snow seasons showed an improvement in the snowpack simulation, especially with respect to simulating the influence of small scale topographic effects on the observed snowpack distribution, and also the timing of snow melt dynamics.