Adaptive assessment of reservoir scheduling to hydrometeorological comprehensive dry and wet condition evolution in a multi-reservoir region of southeastern China

Abstract

The role of reservoirs in water resource management is becoming crucial for flood control and drought mitigation in any basin because of the frequent occurrence of extreme weather events attributed to global climate change and human activities. Therefore, evaluating the relationship between reservoir storage (discharge) and wet (dry) evolution is crucial. This study explores the time-delay effect and spatial heterogeneity of reservoir discharge and storage on dry and wet conditions in several basins of Lin’an District (LAD) in southeastern China. An integrated methodology is developed in this study to assess the relationship by a monthly streamflow simulation model, the meteorological and hydrological comprehensive drought index (CDI) using a Frank Copula function, and an eXtreme Gradient Boosting (XGBoost) model and Shapley Additive exPlanations (SHAP) framework were used to develop a model to forecast dry and wet conditions and to evaluate the key factors affecting their changes. Results from the study indicate that the monthly water balance model can simulate the monthly hydrological processes with relatively high accuracy in the LAD region. The CDI reflects the intensity of wet and dry events more precisely, thoroughly, sensitively, and consistently by combining the benefits of hydrological and meteorological drought indicators. Precipitation, evaporation, streamflow, the Pacific Decadal Oscillation (PDO), and the Indian Ocean Dipole (IOD) were the main contributing factors influencing the above 80% accuracy of the wet and dry forecast models. The average correlation between the outflow of each reservoir in LAD and CDI is 0.47, which is higher than the Standardized Precipitation Index (SPI) and Standardized Runoff Index (SRI). Moreover, the delay in months of dry (wet) events based on SPI, SRI, and CDI are 0.45 (0.41), 1.07 (0.65), and 0.87 (0.60), respectively. It suggests reservoirs are less capable of adaptive scheduling for drought events than for wet events, and they respond most quickly to SPI defined events. The results can provide scientific and technological support for water safety and security in the study area.