Witold Struipczewski

Abstract

The assumption of stationarity made in flood frequency analysis has been relaxed by introducing the possibility of the existence of a tread in the first two moments of a probability distribution function. The problem of the estimation of the time-variable moments and probability distribution function has been analysed. The basic assump-tions and structure of the IDT (Identification of Distribution and Trend) soft package as well as a plan for its further development have been presented and discussed. The maximum likelihood method is used to estimate model parameters from a time series and the Akaike Information Criterion goodness-of-fit test to identify an optimum model, i.e. of distribution and trend functions. Preliminary results of a flood frequency analysis for Polish rivers obtained with the IDT package are described briefly.

Keywords: Time series, trends, annual flood peaks, probability distribution, time variable parameters, time-dependent moments, likelihood function, Akaike Infor-mation Criterion