High-resolution and large scale modelling of seasonal snow in forests over the European Alps

Mid-elevation alpine regions are currently undergoing profound changes, with snow cover regimes shifting from seasonal to ephemeral. At the same time, forests around the world are also undergoing large changes due both natural and human-induced disturbances. Quantifying the impact of these environmental changes on seasonal snow requires physics based models that incorporate the relevant processes, as well as sufficiently detailed datasets of forest structure.

In the last decade, a new generation of snow models have been developed that explicitly represent interactions between forests, snow and meteorology. These models build on airborne lidar data  incorporating the effect of individual tree crowns on radiation transfer and snow interception processes, replacing the use of the leaf-area index and the “big-leaf” approach. However, these new models rely on airborne lidar data with limited spatial extents defined by arbitrary boundaries such as state, municipal and/or isolated hydrological catchments. Large spatial scale and global forest snow modelling is therefore still reliant on the “big-leaf” approach, which is known to have limited performance especially in heterogeneous forest environments. 

This study presents a modelling chain to predict seasonal snow accumulation and ablation in forests based on satellite forest products and global climate forcings (ERA5) applied across the European Alps as a first large-scale use case. The motivation to develop this modelling chain is to facilitate modelling forest snow processes across large spatial scales, especially in previously unstudied remote forested regions around the globe.

The model chain begins with a 10m canopy height model (CHM) derived from Sentinel-2 imagery. The CHM is used with Copernicus Land Monitoring Service forest products as input to the Canopy Radiation Model (CanRad) to calculate the forest structure and radiation transfer input variables for the Flexible Snow Model (FSM2). Forest variables are calculated at 25m sub-grid scale and averaged to run FSM2 at 250 m resolution over the European Alps. The ERA5 meteorological input for FSM2 are downscaled and aggregated at the hillslope scale using the climate downscaling toolkit TopoPyScale (TPS). Throughout the modelling chain, the model outputs are validated using airborne lidar data of both forest structure and snow cover in both the French and Swiss Alps. 

This model chain overcomes large-scale forest snow modelling challenges with 1) an explicit description of snow-canopy interactions, 2) a method compensating for the lack of a global canopy dataset, and 3) reduced computational cost of running large scale simulations. The main advantage of this approach is the ease of use and availability to run over much smaller domains as well as its relevance for global applications in fields such as permafrost, snow and hydrological research.