摘要
Hydrological frequency analysis is the foundation for hydraulic engineering design and water resources management. However, the existence of uncertainty in sample representation usually causes the uncertainty in quantile or design value estimation. Standard deviation (SD) of the design value with a given non-exceedance probability, as an extremely valuable indicator, is suggested to quantify the uncertainties of hydrological frequency analysis. In order to assess the impact of the sampling uncertainty on design value, in China’s national standard for design flood calculation, an empirical formula for estimating SD of the design value was suggested; however, it is applicable only to the case that Pearson type three probability distribution function (PE3) and method of moment are used to analyze the hydrological samples. In principle, for other types of probability distribution functions and parameter estimation methods, the suggested empirical formula is not suitable. The aim of this article is to propose a general approach to estimate the SD of a design value by using the Bootstrap resampling technique. In order to testify the applicability of the approach, under the condition of PE3 distribution, three kinds of schemes were designed to calculate SD, i.e. the suggested empirical formula in China’s national standard for design flood calculation (SEF), Bootstrap technique with method of moment for parameter estimation and Bootstrap with linear moment for parameter estimation. The annual precipitation observations from 50 gauges around China were analyzed using these three approaches. Results show that the SD values of quantile estimates are significantly different among these three approaches, which means that when parameter estimation methods rather than method of moment are employed, the SD provided by SEF is inapplicable. It also indicates that, in terms of modified design value, i.e., original design value adding security correction value, the proposed method using the Bootstrap method, as a general method for SD estimation, is effective and feasible.