Abstract
This article investigates the degrees of liberty and complexity warranted in temperature-index models to jointly simulate local snow measurements, remotely-sensed snow cover and runoff in mountainous areas. To address this issue, the snow- and ice-accounting routine (SIAR) on top of the GR4J model was developed on a dataset covering 17 mountainous catchments (45 to 3 580 km²) in the French Alps and Pyrenees. Model calibration and control were based on streamflow series, fractional snow cover (FSC) computed from MODIS snow products and at least one chronicle of local measurements of snow water equivalent (SWE) acquired in each catchment for the period 2004−2016. SIAR was applied according to a flexible number of equal-area elevation zones and different parametrizations ranging from 11 free parameters (precipitation orographic correction, temperature lapse rate, variation in the temperature lapse rate, correction for snowfall gauge under-catch, rainfall lapse rate, thermal inertia of the snow pack, constant and variable part of the degree-day snow melt factor, degree-day ice melt factor, 2-parameter hysteresis between SWE and FSC), to only fixed parameters. Results showed that the one-free-parameter SIAR is as efficient as more flexible structures in simulating both local and distributed snow dynamics as well as runoff. Interestingly, using SIAR without any free parameters by fixing the snowfall gauge under-catch correction to a median value of 160% only led to slight deterioration of local SWE dynamics. SIAR was then compared with alternative versions to justify internal processes (glacier-component, sublimation, energy balance, snowpack cold-content, variable melt factor) and modes of distribution that were retained in the final version. Finally, SIAR was compared with the Cemaneige snow routine, which showed its modeling performance was better. These findings suggest that it is possible, and even advisable, to limit the number of free parameters in temperature-index models in order to reduce problems of over-parameterization and equifinality.