Including Parameter Uncertainty in an Intercomparison of Physically-Based Snow Models

Snow models that solve coupled energy and mass balances require model parameters to be set, just like their conceptual counterparts. Despite the physical basis of these models, appropriate choices of the parameter values entail a rather high degree of uncertainty as some of them are not directly measurable, observations are lacking, or values are not adaptable from literature. In this study, we test whether it is possible to reach the same performance with energy balance snow models of varying complexity by means of parameter optimization. We utilize a multi-physics snow model which enables the exploration of a multitude of model structures and model complexities with respect to their performance against point-scale observations of snow water equivalent and snowpack runoff observations, and catchment-scale observations of snow cover fraction and spring water balance. We find that parameter uncertainty can compensate structural model deficiencies to a large degree, so that model structures cannot be reliably differentiated within a calibration period. Even with deliberately biased forcing data, comparable calibration performances can be achieved. Our results also show that parameter values need to be chosen very carefully, as no model structure guarantees acceptable simulation results with random (but still physically meaningful) parameters.