Comparative study of flood regional composition methods for design flood estimation in cascade reservoir system

Abstract

The equivalent frequency regional composition (EFRC) method and most likely regional composition (MLRC) method have been widely used for design flood estimation in cascade reservoir system. The EFRC method assumes an identical exceedance probability of design flood quantile between the upstream reservoir site (EFRC-1) or intermediate basin (EFRC-2) and the downstream reservoir site. The MLRC method statistically considers the actual inter-correlation between floods of different sub-basins and determines the flood regional composition by selecting the one with maximum occurrence likelihood. The advances, limitations, inherent connections and differences of these two methods have never been theoretically investigated in literature. In this study, the EFRC and MLRC methods are comprehensively compared through theoretical derivation, statistical experiment and applicability assessment. The main findings are: (1) The difference between EFRC-1 and MLRC methods tends to be smaller with the increase of correlation of floods across different sub-basins, and these two methods are mathematically equivalent in terms of totally correlated floods. (2) The difference between EFRC-2 and MLRC methods are impacted by both the correlation and the variances of annual maximum flood series of upstream and downstream reservoirs. (3) The EFRC method is reasonable when the correlation coefficient exceeds 0.7 for one or two reservoirs; while the MLRC method tends to be more reasonable and practical for complex cascade reservoirs. (4) Compared with original design flood results, the design flood of downstream reservoirs in the Wu River decrease significantly due to the regulation of upstream reservoirs. The flood limited water levels (FLWLs) of downstream reservoirs can be raised without increasing flood control risks, which can increase 174.3 million kW·h annual hydropower generation in flood season.