Original version
Journal of Hydrology. 2019, 577:124003, DOI: https://doi.org/10.1016/j.jhydrol.2019.124003
Abstract
The hydrological regimes of downstream reservoirs have been significantly altered due to the operation and regulation of upstream cascade reservoirs. The original design flood quantiles, namely “design flood in construction period”, do not consider anthropogenic impacts in reservoir operation period, and have led to enormous conflicts between flood control and conservation. In this study, the “design flood and flood limited water level in operation period” are defined for practical application. We establish a general framework to measure the spatiotemporal pattern of streamflow and to estimate design floods of cascade reservoirs in operation period. The multivariate t-copula and a genetic algorithm strategy are proposed to solve the curse of dimensionality encountered in the derivation of most likely regional composition. The Jinsha River and Yalong River cascade reservoir system in China, which consists of 13 large reservoirs with the total storage capacity of 74.06 billion m3 and hydropower capacity of 71.47 GW, is selected as a case study. Results indicate that: (1) The curse of dimensionality can be well addressed by applying multivariate t-copula to build high dimensional joint distribution and using the genetic algorithm to achieve the most likely regional composition. (2) Compared with the design floods in construction period, the design floods of downstream reservoirs in operation period have been significantly reduced due to the upstream reservoir regulation. The 1000-year design peak flood discharge, 3-day, 7-day and 30-day flood volumes of Xiangjiaba reservoir decrease by 38.7%, 37.4%, 34.2% and 13.8%, respectively. (3) The flood limited water level of these reservoirs can be raised without increasing flood control risks in operation period. The cascade reservoirs in the Jinsha River and Yalong River can generate 3.28 billion kW h more hydropower (or increase 4.3%) annually during flood season.