Testing for nonzero skew in maximum discharge runoff data

Abstract

Water Resource Council Bulletin 17B (Interagency Advisory Committee on Water Data, 1982) recommends finding T‐year flood values by fitting stream gauge data with a log Pearson III distribution. A problem arising in using this method is that, as is well known, the estimator for the skew of this distribution is highly variable and the computed T‐year values are sensitive to the value of skew used. The variability in estimating the skew is studied in this note by computing operating characteristic curves which display the probability that a sample from a site with a nonzero skew will actually fall in a specific interval. In rough summary it is not usually possible to reject, at a high level of significance the hypothesis that the skew at a site is zero by using only the data available for that site.