Process-based strategies for model structural improvement and reduction of model prediction uncertainty

This paper examines how the combination of simulated discharge and mean catchment residence time (MRT) may be used to subsume flow path process complexity and provide a simple, scalable evaluative data source for water quantity-quality based conceptual models at the catchment scale. A simple Monte Carlo framework is used to evaluate the identifiability of parameters, and how values of mean residence time contribute to the evaluative process and ultimate level of model complexity warranted in the model structure. Our results show that models that might otherwise be acceptable for flow may be wholly rejected for an inability to capture residence time dynamics. The incorporation of soft, or highly uncertain and potentially qualitative data in model evaluation is a useful rejectionist-based mechanism to bring experimental evidence into the process of model evaluation and selection. This may provide a way to reconcile hillslope complexity with catchment scale simplicity and to help define the degree of process complexity needed in a given model application.