On the potential of variational calibration for a fully distributed hydrological model: application on a Mediterranean catchment

摘要: Abstract. Flash flood alerts in metropolitan France are provided by SCHAPI (Service Central Hydrometeorologique et d’Appui a la Prevision des Inondations) through the Vigicrues service, which is designed to work ungauged catchments. The AIGA method implemented for forecasting on small- and medium-scale watersheds. It based distributed hydrological model accounting spatial variability of rainfall catchment properties, radar observation inputs. Calibration parameters describing these properties with high resolution difficult, both technically (in terms estimation method), because identifiability issues. Indeed, number be calibrated much greater than locations where discharge observations usually available. However, propagation dynamic process, so have also temporal dimension. This must larger enough comprise representative set events. In order fully benefit from using method, we consider its (GRD) combination variational (data assimilation) method. this optimal found minimizing objective function includes misfit between observed predicted values some additional constraints. minimization process requires gradient cost respect all control parameters, efficiently computed adjoint model. scalable, fast converging, offers convenient framework introducing constraints relevant hydrology. can used calibrating estimating initial state system short range manner weather forecasting). study area Gardon d’Anduze watershed four gauging stations numerical experiments, benefits against uniform calibration analysed predictive performance. Distributed shows encouraging results better prediction at gauged locations.