Flood Frequency Analysis Through Nonparametric Methods

Abstract

A detailed description is given of two nonparametric methods that are used in annual flood frequency analysis. This new approach does not select a priori a probability density function (pdf) model but based on the available register or sample values, computes one. It is appropriate for multimodal or heavy-tailed pdf. The procedures put forward correspond to kernel function methods (circular; rectangular and Gaussian) and the direct partial estimator of the accumulated probability distribution. The computer programs that were used and their test applications are discussed. Finally, two numerical applications are described and analysed for etwo hydrometric stations with very different characteristics