Hydrologic risk analysis in the Yangtze River basin through coupling Gaussian mixtures into copulas. (February 2016)

• Record Type: Journal Article
• Authors: Fan, Y.R.; Huang, W.W.; Huang, G.H.; Li, Y.P.; Huang, K.; Li, Z.
• Is Part Of: Advances in water resources. Volume 88(2016)
• Page Start: 170
• Page End: 185
• Publication Date: 2016-02
• Subjects: Flood risk -- Copula -- Flood frequency analysis -- Distribution -- Conditional distribution -- Gaussian mixture model; Hydrology -- Periodicals; Hydrodynamics -- Periodicals; Hydraulic engineering -- Periodicals; 551.48
• DOI: 10.1016/j.advwatres.2015.12.017
• Languages: English
• ISSNs: 0309-1708

Abstract

Highlights: Coupled GMM-Copula method is proposed for hydrologic risk analysis. GMM is applied for marginal distribution estimation of flood variables. The impacts of flood volume and duration on hydrologic risk are revealed. The implications of bivariate hydrologic risk on flood control are explored. Abstract: In this study, a bivariate hydrologic risk framework is proposed through coupling Gaussian mixtures into copulas, leading to a coupled GMM-copula method. In the coupled GMM-Copula method, the marginal distributions of flood peak, volume and duration are quantified through Gaussian mixture models and the joint probability distributions of flood peak-volume, peak-duration and volume-duration are established through copulas. The bivariate hydrologic risk is then derived based on the joint return period of flood variable pairs. The proposed method is applied to the risk analysis for the Yichang station on the main stream of the Yangtze River, China. The results indicate that (i) the bivariate risk for flood peak-volume would keep constant for the flood volume less than 1.0 × 10 5 m 3 /s day, but present a significant decreasing trend for the flood volume larger than 1.7 × 10 5 m 3 /s day; and (ii) the bivariate risk for flood peak-duration would not change significantly for the flood duration less than 8 days, and then decrease significantly as duration value become larger. The probability density functions (pdfs) of the flood volume and duration conditional on flood peak can also be generated through the fitted copulas. The results indicate that the conditional pdfs of flood volume and duration follow bimodal distributions, with the occurrence frequency of the first vertex decreasing and the latter one increasing as the increase of flood peak. The obtained conclusions from the bivariate hydrologic analysis can provide decision support for flood control and mitigation.