Evaluating snow models for hydrological applications

Abstract:

Much effort has been invested in developing snow models over several decades, resulting in a wide variety of empirical and physically-based snow models. Within the two categories, models are built on the same principles but mainly differ in choices of model simplifications and parameterizations describing individual processes. In this study, we demonstrate an informative method for evaluating a large range of snow model structures for hydrological applications using an existing multi-model energy-balance framework and data from two well-instrumented sites with a seasonal snow cover. We also include two temperature-index snow models and one physically-based multi-layer snow model in our analyses. Our results show that the ability of models to predict snowpack runoff is strongly related to the agreement of observed and modelled snow water equivalent whereas such relationship is not present for snow depth or snow surface temperature measurements. For snow water equivalent and runoff, the models seem transferable between our two study sites, a behaviour which is not observed for snow surface temperature predictions due to site-specificity of turbulent heat transfer formulations. Uncertainties in the input and validation data, rather than model formulation, appear to contribute most to low model performances in some winters. More importantly, we find that model complexity is not a determinant for predicting daily snow water equivalent and runoff reliably, but choosing an appropriate model structure is. Our study shows the usefulness of the multi-model framework for identifying appropriate models under given constraints such as data availability, properties of interest and computational cost.