ESTIMATION OF ERRORS IN CALCULATED HYDROLOGICAL CHARACTERISTICS IN CASE OF INHOMOGENEOUS OBSERVATION SERIES

Abstract

The determination of errors in the runoff probability distribution parameters when estimating them on a heterogeneous sample is considered. For the case that allows using composite curves (or a mixture of distributions), some expressions for errors of the estimates of the mean and the variation coefficient are proposed. It has been found that accounting of the sample heterogeneity based on a probabilistic model of the time series, which allows its examination in the form of two stationary sections without accounting the averaged autocorrelation, magnifies the error of the mean value estimate. It is shown that the overestimation of the first autocorrelation coefficient assessed on the entire series without taking into account inhomogeneity should be considered.