| description abstract | ydrological modeling in mountainous regions, where catchment hydrology is heavily influenced by snow (and possibly ice) processes, is a challenging task. The intrinsic complexity of local processes is added to the difficulty of estimating spatially distributed inputs such as precipitation and temperature, which often exhibit a high spatial heterogeneity that cannot be fully captured by measurement networks. Hence, an interpolation step is often required prior to the hydrological modeling step. Usually, the reconstruction of meteorological forcings and the calibration of the hydrological model are done sequentially. The outputs of the hydrological model (discharge estimates) may give some insight into the quality of the forcings used to feed it, but in this two-step independent analysis, it is not possible to easily feed the interpolation scheme back with the discrepancies between observed and simulated discharges. Yet, despite having undergone the rainfall?runoff (or snow?runoff) transformation, discharge at the outlet of a (sub)catchment is still an interesting integrator (spatial low-pass filter) of the forcing fields and is ancillary areal information complementing the direct, point-scale data collected at gauges. In this perspective, choosing the best interpolation scheme partly becomes an inverse hydrological problem. Here, a joint calibration strategy is presented where the parameters of both the interpolation model (i.e., reconstruction procedure of meteorological forcings) and the hydrological model (snow cover, soil moisture accounting, and flow-routing schemes) are jointly inferred in a multisite and multivariable approach. Interpolated fields are daily rainfall and temperature, whereas hydrological variables consist of discharge and snow water equivalent time series at several locations in the Durance River catchment. | |
