On the Choice of a Performance Metric for Model Calibration Scheme Using Discharge-Age Information

The robustness of a hydrological model is mainly dependent on how well the simulations resemble a quantity of interest, typically discharge, which is mainly achieved by successful model calibration. Performance metrics are used to quantify to what degree the system of interest is represented by a model. However, as there is not a general procedure and a rule for the selection of a performance metric, it is ultimately the modeller’s subjective decision. This study explores the contribution of discharge young water fractions—derived from stable water isotopes—to the selection of a hydrologically appropriate performance metric for the model calibration. For our analysis, we examined different metric combinations by a multi-variable calibration scheme on the Pareto-optimal frontier. By searching for a trade-off between the model performance on discharge and the model performance on discharge young water age, we not only satisfy the minimization of model simulation bias but also bring the process-based discharge-age information content into the model parametrization. To test our hypothesis, we applied our approach to one of the karstic water sources in Austria. Our finding indicates that the information content of young water age supports the proper choice of a performance metric for the model calibration scheme while reducing the modeller’s subjectivity on performance metric selection, thereby ensuring physically more plausible parameter sets.