Abstract: Attributing a physical meaning to the calibration of a conceptual hydrological model is a risk due to equifinality, i.e., the existence of multiple optimal parameterizations that might or might not represent the actual behavior of a catchment; a risk that propagates to posterior studies that use the outputs of the hydrological model as an input.This study proposes and analyses sequential procedures for calibrating conceptual hydrological models aimed at reducing equifinality. These procedures force the model to reproduce a flow separation of the observed streamflow into quick flow and base flow, which we assumed representative of the run-off generating processes in the catchment. The sequential calibration of the model parameters that control quick flow and base flow forces the model to reproduce the flow separation, and introduces additional constraints in the calibration process that reduce equifinality and improve the overall calibration procedure. We applied this procedure to two mesoscale catchments in the ?Picos de Europa? National Park (northern Spain). We compared the performance of the different calibration procedures both in the real scenario and in hypothetical scenarios of land use and soil permeability, to provide a sounder assessment of the ability of the procedures under diverse conditions. Results show that a calibration method applying hydrograph separation ensures models with a better discharge partition, whereas methods that do not apply separation failed in a considerable number of cases. In terms of performance (NSE and bias), the method applying hydrograph separation outperformed the reference method (without separation) for the real scenario, even for total streamflow; in the hypothetical scenarios though, the improvement in process representativeness came at the expense of a slight loss in performance. The sequential methods here developed were more computationally efficient; since they explore the parameter space in subsets, the number of iterations until convergence was a third of that needed with the reference method. In summary, we have developed a simple calibration procedure that ensures a better model behavior (more in line with the underlying conceptualization) with a similar, and even better, performance and a shorter calibration time than the reference method.

Autoría

Descargar referencia