The proper role of automatic methods in parameter estimation

Model parameter estimation involves the selection of values for the model parameters so that the model output matches the behavior of the watershed system as closely as possible. Automatic parameter estimation methods for hydrologic models have been under development for at least three and a half decades, with the degree of sophistication generally paralleling the increases in computing power. The goal has been to develop an objective strategy for parameter estimation that provides consistent performance, by eliminating the kinds of subjective human judgements involved in manual parameter estimation techniques. Note that the manual approach uses the highly subjective process of visual inspection and comparison of the model output and the observed data to implicitly measure the ability of the model to simulate specific aspects of the hydrologic behavior. Dissatisfaction with the results obtained using the traditional automatic approach have led to recent suggestions that a automatic "multi-criteria" approach which emulates (rather than replacing) the evaluation procedures used in the manual approach may be required to provide model calibrations that are more consistent and reliable. With this approach, the model performance is evaluated in terms of several carefully selected measures that reflect different observable characteristics of the system behavior. The region of the parameter space associated with simultaneous minimization of these measures is called the Pareto region. The rationale for this method, and the power of the approach are illustrated through a case study using the SACramento Soil Moisture Accounting model (SAC-SMA). The results indicate that parameter sets selected within the Pareto region tend to provide more consistent and reliable model forecasts.