A Physics-Based Regionalized Snow Modeling Framework for the Data-Sparse Pan-Siberian Domain

Quantifying snow water equivalent (SWE) and melt dynamics across the Pan-Siberian domain is critical for understanding the hydro-climatological conditions of the entire Northern Hemisphere. However, due to the scarcity of data, the strong interaction between vegetation and climate, and representativeness biases of sparsely distributed measurement stations, there is a high degree of uncertainty in current estimations. While traditional monitoring networks provide essential points of reference, their limited spatial coverage and site-selection biases—often favoring open clearings—hinder the accurate assessment of regional snow storage across diverse and complex landscapes.
In this study, we develop and apply a physics-based snow process model designed for data-sparse cold regions, combined with a corresponding regionalization strategy to bridge the gap between sparse point-scale observations and regional snow dynamics. The model was first validated at the Sodankylä site in Finland, demonstrating high performance for both Snow Depth (NSE > 0.78) and SWE (NSE > 0.83), indicating a physically consistent representation of snow density, compaction, and melt processes. The model was then applied across Pan-Siberia by grouping 85 stations into hydro-climatic regimes based on wind, precipitation characteristics, and forest cover. Model parameters were calibrated simultaneously across stations within each regime to derive robust zonal parameter sets, thereby ensuring physical consistency and overcoming parameter equifinality.
The resulting regionalized model achieves robust performance across the majority of the domain (median KGE > 0.75), substantially outperforming global default parameterizations. The results reveal a key physical insight in forest-dominated Taiga regions, where the optimized wind correction factor converges toward zero, confirming the strong canopy sheltering effect and indicating that standard WMO wind corrections systematically overestimate snowfall under forest cover. In contrast, the Cold Continental regime (Yakutia) exhibits a high rain–snow temperature threshold (~+3.7°C), reflecting sublimation-driven cooling under extremely dry atmospheric conditions. 
This approach enables the reconstruction of spatially consistent, multi-year snow dynamics across Pan-Siberia, providing a scalable strategy for hydrological modeling in ungauged, cryosphere-dominated regions and offering new insights into the spatiotemporal evolution of Eurasian snow resources.