This study describes the application of the coupled SAFRAN (meteorological variables), ISBA (soil–vegetation–atmosphere transfer) and CROCUS (snow cover evolution) models to simulate daily discharges for the upper Durance catchment (French Alps) from 1981 to 1994. The results are validated by comparison with measurements at three gauging stations located in the watershed. Previous investigations have shown a remarkable overestimation of the spring flood peak generated by the modelled snowmelt. It could be significantly improved by increasing the model resolution from 8 km to 1 km, thus more precisely considering the elevation-dependent snowmelt process. However, it is also possible to use subgrid parameterisations at the coarse grid resolution to improve simulations. This paper investigates the influence of a subgrid parameterisation for topography, a subgrid parameterisation for the snow cover in a forest canopy and a combination of the two on the simulated spring flood peak. Results show a significant improvement in the simulations by both subgrid parameterisations, in particular by their combination: the Nash–Sutcliffe efficiency of the daily discharges is improved from 0-73 (original experiment) to 0-77 (subgrid topography), to 0-75 (forest) and to 0-78 (combination of subgrid topography and forest).
