Selection of catchment descriptors for the physical similarity approach. Part I: Theory

This paper focuses on a description of the method used for the identification of optimal catchment descriptors for the physical similarity approach consisting of a scheme for the identification of optimal catchment descriptors and the procedure for finding hydrologically homogeneous regions using inverse clustering. Andrews' curves are used as the basis for homogeneity checking. The identification of an optimum catchment descriptor is based on the assumption that the addition of an optimal catchment descriptor to a predefined set of catchment descriptors improves the accuracy of model parameter estimation within a set of tested catchments. Two criteria are proposed for the selection of optimal catchment descriptors - a criterion evaluating estimates of model parameters on the basis of different potentially optimal groups of catchment descriptors, MIN, and a criterion evaluating the improvement in model parameter estimation after the addition of a potentially optimal catchment descriptor into the group of preliminarily identified optimal catchment descriptors, MAX. The proposed method provides an alternative to the trial-and-error method for the identification of optimal catchment descriptors.