Multi-period and multi-criteria model conditioning to reduce prediction uncertainty in an application of TOPMODEL within the GLUE framework

A new approach to multi-criteria model evaluation is presented. The approach is consistent with the equifinality thesis and is developed within the Generalised Likelihood Uncertainty Estimation (GLUE) framework. The predictions of Monte Carlo realisations of TOPMODEL parameter sets are evaluated using a number of performance measures calibrated for both global (annual) and seasonal (30 day) periods. The seasonal periods were clustered using a Fuzzy C-means algorithm, into 15 types representing different hydrological conditions. The model shows good performance on a classical efficiency measure at the global level, but no model realizations were found that were behavioural over all multi-period clusters and all performance measures, raising questions about what should be considered as an acceptable model performance. Prediction uncertainties can still be calculated by allowing that different clusters require different parameter sets. Variations in parameter distributions between clusters, as well as examination of where observed discharges depart from model prediction bounds, give some indication of model structure deficiencies.