A simple and useful modification of the index flood method.

Abstract

The index flood method is a hydrologic regionalization technique widely used in the world. The method, which was originally proposed by the U.S. Geological Survey, was developed based on the application of the Langbein homogeneity test. Its general application steps are the definition of the homogeneus region, the determination of an index flood, the generation of a regional frequency curve, and the use of various relationships between physiographic characteristics of the basin and the so called index flood. Over time, many advantages have been recognized for the method, but some deficiencies have also been identified. It is common, for example, that the inclusion of a certain station in the analysis of an initial group of stations causes some of the other stations to fall out of the homogeneous region, or vice versa, not including a certain station can produce some others to become part of the homogeneous region. This anomalous behaviour of the method is the main issue of this work. In the paper, a simple but useful modification of the original method is proposed. The use of a "filtering" procedure avoids the inclusion, from the outset, of those stations that impose serious deviations to the method. The use of this filter not only generates bigger homogeneous regions, but also regions with a better homogeneity. Furthermore, through a cross validation technique, it was shown that the discharge estimates of ungauged basins within the region are much better than those obtained with the original version of the method.