Multiscale calibration and validation of a conceptual rainfall-runoff model

Model calibration and validation is usually limited to comparing streamflows at the basin outlet. In this study observed runoff series from nested basins were used for calibration and validation on different spatial scales. A conceptual rainfall-runoff model was applied to nested basins of different sizes (15.2, 40 and 257 km(2)) located in the Black Forest in south-west Germany. The first step was to calibrate the model individually for each of the three basins and to use the runoff series from the other two basins for validation. Optimised parameter values were related to sizes and other properties of the basins. In the second step, the model was calibrated simultaneously to the runoff series from all three catchments using a genetic algorithm and a fuzzy combination of the individual objective function values. It was not possible to obtain as good fits as those achieved by separately calibrating the model to each subbasin. However, the fit between measured and observed streamflow for the individual basins were acceptable (model efficiency values around 0.8) and significantly better than those obtained using a parameter set optimised in just one of the other basins. (C) 2000 Elsevier Science Ltd. All rights reserved.