Multi-scale snowpack modeling in the Pyrenees using the Canadian Hydrological Model

In mountain ecosystems, snow is a critical resource that regulates hydrological processes, ecosystem dynamics, economical activities and downstream water availability. Accurately estimating snow at these highly heterogeneus environments remains a challenge, due to the strong spatial and temporal variability. The combination of snowdrift-permitting models and snowpack remote sensing observations can improve the accuracy of snowpack estimations across scales. The Canadian Hydrological Model (CHM) is a novel snow modeling framework that explicitly represents lateral snow transport processes over an irregular mesh. This study analyzes the impact of modelling spatial scales over three domains in the Pyrenees between 2019 and 2025 using CHM: the Izas Experimental Cathment (~10 km²), a portion of the Tena Valley (~100 km²), and a larger section of the mountain range (~1200 km²) using a snowdrift permitting model. Each domain is modelled using a different horizontal resolution, relative to the domain area, and driven by downscaled meteorological forcings. We analyze several snowpack properties, including snow covered area and snow depth, across the spatial scales, using point-scale snow survey stations, UAV-derived snow depth distribution maps at the catchment scale, Pléiades-derived snow depth maps at the valley scale and Sentinel 2 imagery at the mountain range scale. Error statistics, spatial efficiency metrics and scale breaks derived from semi variograms are used to evaluate the model performance. Preliminary results show that higher resolution simulations have a better representation of snow depth variograms and their scale breaks, and lower mean snow depth biases over the Izas catchment. However, snow depth is overestimated during the accumulation period and underestimated during the ablation season, and differences between the observed and simulated spatial snow distribution can be seen. This study improves our understanding of snowpack dynamics across spatial scales and of the horizontal resolution required for accurate snow simulations. Finally, this study enables the development of a remote sensing–based monitoring framework for the Pyrenees to improve snowpack simulation, which open new insights and allow more reliable forecasts.