Abstract: Flood overtopping risk rate is a key factor in reservoir operation and dam safety management. This paper estimates this factor based on Copula functions and analyzes its dependency on the selection of these functions and the correlation between flood peak and volume. We apply the Gumbel-Hougaard Copula function to an analysis of the Geheyan reservoir in the Qing River basin, and demonstrate that its optimal form can better describe the upper tail dependence between flood peak and flood volume, while the Frank and Clayton Copula functions underestimate the risk rate significantly. For this reservoir, the estimated flood overtopping risk rate is 3.79×10-5, but it increases as the correlation between flood peak and volume increases. In the cases of full correlation it is overestimated while underestimated in the independence cases. Our bivariate statistical approach based on Copula functions, through effectively capturing the dependence between flood peak and volume, improves the estimation of overtopping risk rate and would be useful for overtopping risk analysis.

Key words: dam flood overtopping, risk analysis, Copula function, Monte Carlo simulation, Geheyan reservoir