Abstract: Generally, flood events have several characteristic attributes, and the time variations in these attributes can be comprehensively analyzed using multivariate flood frequency analysis. This study develops a peak discharge-flood volume joint frequency analysis model for discontinuous sequences based on the Copula function, considering both quantitative and non-quantitative historical floods and using a genetic algorithm to estimate the model parameters. And the influence of uncertain changes in historical floods on parameter estimation and design floods is analyzed. This model is applied in a case study of the Yichang hydrologic station that has a systematic gauge record and a historical flood record. Results show that parameter estimation is more reasonable when the peak discharges of a historical flood event are described using confidence intervals with lower and upper bounds, and in this case the estimates of flood frequency analysis agree better with the field observed relationship of peak discharge versus flood volume. The peak discharge of the design flood is decreased with an increase in its confidence interval, and the decrease becomes larger at a longer recurrence period.

Key words: copula function, multivariate joint frequency distribution, flood frequency analysis, genetic algorithm, historical flood, Yichang hydrologic station