Maximum flood regionalization in large basins: study case of the Alto São Francisco region--Minas Gerais, Brazil

This paper describes a regionalization method, based on the TCEV (two-component extreme value) distribution function, and its application to annual maximum flood series for the São Francisco River, in Minas Gerais State (Brazil). This statistical model is equivalent to the product of two exponentials, both representing a Poisson process: the first corresponding to the most frequent floods and the second to the rare floods. The theoretical aspects of the model and the regionalization method are briefly presented. The study region is a 50 870 km² river basin, located in the southeast region of Brazil, characterized by a tropical climate. In applying the technique, two homogeneous regions were identified, dividing the basin into southern and northern parts with respect to the annual maximum floods. To validate the TCEV regionalization model, a stream gauge with a short of maximum floods that was not used in the fitting procedure was selected. The TCEV regionalized model fitted well, showing that the method can be used for prediction across the entire basin. Copyright © 2009 John Wiley & Sons, Ltd.