Comparative global sensitivity analysis on the hydrologic models GRP and MORDOR-TS

Today, forecast uncertainties are often provided together with meteorological or hydrologic forecasts. Generating ensemble forecasts covers this objective. Hydrologic forecast uncertainties can be classified into two classes: (i) input uncertainties, and (ii) hydrologic uncertainties, among which are the uncertainties on initial conditions, on the model structure and its parameters. The weight of the various sources of uncertainty can be assessed through sensitivity analysis. This article investigates hydrologic forecast uncertainties with a Global Sensitivity Analysis for the two hydrological models GRP (lumped) and MORDOR-TS (semi-distributed) applied on three sub-catchments of the Odet catchment (Brittany, France). Uncertainty sources taken into account are rainfall, with the use of the ensemble rainfall forecast PEAROME, and the models parameters. The Sobol indices time-series are computed over the event from the 26th to the 30th January 2016. Results highlight a similar behaviour in the response of the two models. It appears that the major uncertainty source comes from the water balance model component. It also appears that the response time for the Treodet sub-catchment differs between both hydrological models.