Multi-objective regional modelling

Regional rainfall runoff models are used to estimate the hydrological dynamics, the water balance, and the statistics of hydrological variables at ungauged sites. Most hydrological models are calibrated in order to get a good fit between observed and simulated variables. One or several objectives might be used for the calibration, and the modelling uncertainty is seen in at least two phenomena: (1) The surface defined by a objective function is often non-smooth, multi-modal and even discontinuous. (2) Different objectives prefer different parameter sets. The aim of this study is to obtain regional parameter sets for the Ecomag model applied to the Saone catchment, and to assess the modelling uncertainty based on the trade-offs between the R-eff criterion for daily streamflows in different catchments. This aim was achieved by applying a multi-objective calibration method with streamflow data from seven catchments as the objectives. A Pareto-set with 2330 parameter values were obtained. The 2330 parameter sets were used to obtain envelopes for the simulated streamflow. The results show that all the parameters might be considered as sensitive to the output results, that the trade-off uncertainty between the different catchments is important, and that this uncertainty can only explain parts of the complex uncertainties in catchment modelling. There is a need for finding new concepts describing the complex uncertainties in catchment models operating on daily time step. (c) 2005 Elsevier B.V. All rights reserved.