Evaluating Different Model Structures for Representing Watershed Functions through the use of Signature Measures

The increasing availability of hydrometeorological data and computational resources has allowed the evolution of hydrological models from lumped, conceptual to fully distributed. However, principal catchment behavioral functions are poorly understood, mainly because model evaluation has been typically based on the comparison of simulated and observed time series of model outputs (e.g., streamflow), ignoring the possibility that we may be getting the right results because of a compensation of errors in model structure, parameters and data. In recent years the hydrological community has redirected its efforts to look for a better understanding of hydrological models from a functional point of view (e.g. water balance, vertical redistribution of soil moisture and redistribution of runoff in time, among others). In this research, we evaluate the ability of three hydrological models (PRMS, VIC and Noah-MP) to skillfully reproduce relevant watershed processes in the Animas River basin, which is a sub-basin of the Colorado River Basin. A suite of signature measures that have diagnostic power of model behaviors is developed and analyzed in order to diagnose the model deficiency. All model simulations were run with the same spatial discretization and forcing data to enable fair comparison of model structures