Many-Criterion Calibration of Hydrologic Models Using Hydrological Signatures

Monday, 15 December 2014
Mahyar Shafii Hassanabadi and Bryan Tolson, University of Waterloo, Waterloo, ON, Canada
Experts have used a variety of techniques for calibration of hydrologic models including optimization-based and Bayesian methods. The simulated outcome of calibrated hydrologic models should be consistent with both the measured response data and the catchment’s hydrological behaviour. The literature indicates that, due to the computational issues associated with many-criterion optimization-based calibration, calibration experiments start by the optimization against performance metrics, e.g., the Nash-Sutcliffe Efficiency measure, and are followed by the evaluation of the results with respect to hydrological signatures, e.g., the flow duration curve indices, to verify the hydrological consistency of the obtained parameter sets. The proposed research develops a full multi-criterion calibration framework that is capable of utilizing many goodness-of-fit and hydrological signature-based measures to locate the hydrologically consistent parameter sets. This calibration approach applies optimization concepts to facilitate the parameters search, and moreover, implements experts’ priorities in terms of the level of acceptability for each calibration measure. The preliminary results demonstrate that the proposed many-criterion calibration approach yields a higher proportion of hydrologically consistent parameter sets in comparison to the traditional approaches in the literature. The proposed approach can be extended to other calibration studies where the modeller has specific targets for individual calibration criteria.