On the performance and uncertainty of regional snow depth modelling across landscapes in northern latitudes

Abstract:

Snowmelt is one of the major drivers of hydrological cycles in cold regions, and its accurate representation in regional hydrological models is key for regional snow depth and streamflow predictions. The choice of snowmelt modules in hydrological models is yet subjected to debate, and thorough characterization of the uncertainty of this process, particularly in the context of climate change, has remained essential. This study conducted the performance and uncertainty associated with using two general snowmelt approaches, namely Energy-Balance Modules (EBMs) and Temperature-Index Modules (TIMs) in analysis of snow depth and streamflow. Furthermore, two common Snow Density formulations (SNDs) were used to map snow water equivalent (SWE) to snow depth. We used (1) the Soil and Water Assessment Tool (SWAT) model to evaluate performances of different snowmelt modules in reproducing snow depth and streamflow; and (2) the Analysis of Variance (ANOVA) method to quantify the contribution of model parameters, global climate models, emission scenarios and downscaling methods to the cascade of uncertainty of snow depth projections within SWAT model. These analyses are implemented on monthly basis in mountainous, foothill, and plain regions in a large snow-dominated watershed in western Canada. Results show that the choice of SNDs has a major effect on the performance and uncertainty of snowmelt modules, rather than the choice between TIMs and EBMs. The results further show that the model parameter uncertainty holds the largest uncertainty share in snow depth projections, particularly in mountainous and foothill regions. In plain regions, however, this share becomes less dominant compared to other sources of uncertainty. Overall, it is demonstrated that the hydro-climatic variability and topographic conditions of different regions play an important part in the contribution of uncertainty sources in terms of regional projections of snow depth.