JOINT USE OF L-MOMENT DIAGRAM AND GOODNESS-OF-FIT TEST - A CASE-STUDYOF DIVERSE SERIES

Selection of a probability distribution for annual maximum discharges was carried through exploratory and confirmatory stages. The L-moment diagram was applied first in order to screen out inappropriate candida te distributions. The Anderson-Darling test was then applied in order to examine the descriptive performance of screened distributions. The effectiveness of this two-stage procedure was studied on an ensemble o f diverse annual maximum discharge series recorded in Israel. The Gene ralized Pareto distribution was the only one of a number of candidate distributions whose theoretical relationship between L-moments followe d closely enough the actual relationships in the case study. This dist ribution was well fitted to about one half of the discharge series, an d satisfactorily fitted to the entire ensemble. It was concluded that such a two-stage procedure, which applies quantitative measures in bot h stages, would reduce the subjectivity involved with the selection of a probability distribution, thus improving the credibility of predict ed high discharges. (C) 1997 Elsevier Science B.V.