Nonstationary hydrological frequency analysis in light of model parameters and climate projections uncertainty

Abstract

Nonstationary hydrological frequency analysis (HFA) has been a commonly used approach for design flood estimation under nonstationary conditions, which underlies deep uncertainty. Most of the previous studies focus on the influence of model parameter (MP) uncertainty on design flood, little is known about quantitative assessment on the commonly-used precipitation projection (PP) uncertainty impact on it. To address this concern, this study proposed an analysis framework to quantify the design flood estimation uncertainty in light of MP uncertainty, PP uncertainty and both of them. Firstly, the equidistant cumulative distribution function method was used to correct the PP provided by twenty-eight GCMs to improve their quality; secondly, the optimal nonstationary distribution model, coupling with the Equivalent Reliability method, was applied to estimate the design flood; then the uncertainty analysis framework composed of Bayesian theory and poor-man ensemble strategy was employed to analyze the impact of the MP uncertainty, PP uncertainty and both of them on the design flood estimation. Moreover, the investigation on which one between MP and PP uncertainty was dominated in the total uncertainty on design flood estimation was conducted based on a case study. Results showed that MP and PP uncertainty has noticeable impacts on the design flood uncertainty. The MP uncertainty impact on the design flood was greater than the PP uncertainty.