Representative parameter estimation for hydrological models using a lexicographic calibration strategy

Autor: Günter Meon, Phillip Kreye, Marlene Gelleszun
Rok vydání: 2017
Předmět:
Zdroj: Journal of Hydrology. 553:722-734
ISSN: 0022-1694
DOI: 10.1016/j.jhydrol.2017.08.015
Popis: We introduce the developed lexicographic calibration strategy to circumvent the imbalance between sophisticated hydrological models in combination with complex optimisation algorithms. The criteria for the evaluation of the approach were (i) robustness and transferability of the resulting parameters, (ii) goodness-of-fit criteria in calibration and validation and (iii) time-efficiency. An order of preference was determined prior to the calibration and the parameters were separated into groups for a stepwise calibration to reduce the search space. A comparison with the global optimisation method SCE-UA showed that only 6% of the calculation time was needed; the conditions total volume, seasonality and shape of the hydrograph were successfully achieved for the calibration and for the cross-validation periods. Furthermore, the parameter sets obtained by the lexicographic calibration strategy for different time periods were much more similar to each other than the parameters obtained by SCE-UA. Besides the similarities of the parameter sets, the goodness-of-fit criteria for the cross-validation were better for the lexicographic approach and the water balance components were also more similar. Thus, we concluded that the resulting parameters were more representative for the corresponding catchments and therefore more suitable for transferability. Time-efficient approximate methods were used to account for parameter uncertainty, confidence intervals and the stability of the solution in the optimum.
Databáze: OpenAIRE
Externí odkaz: