Integration and Evaluation of Forecast-Informed Multiobjective Reservoir Operations

Incorporating streamflow forecasts into reservoir operations can improve water resources management efficiency, yet the forecast value in multipurpose reservoir systems is rarely investigated, let alone the relationship between forecast accuracy and value in multiobjective reservoir operation. Here, we propose a forecast-informed framework to derive multiobjective operating rules based on radial basis functions and the Pareto archived dynamically dimensioned search optimization algorithm and subsequently develop indicators reflective of Pareto fronts with and without forecast information to characterize forecast value. Based on a case study of the Hanjiang cascade of reservoirs in the Yangtze River Basin, China, the optimal inclusion of streamflow forecasts notably improves the performance of multiobjective reservoir operations, mainly by significantly increasing the hydropower generation. The relationship between forecast accuracy and value is explored by comparing four accuracy indicators (Nash–Sutcliffe efficiency, mutual information, correlation coefficient, and Kullback–Leibler distance) and forecast value. The correlation coefficient is found to be the most suitable forecast indicator given its high correlation with forecast value and stability in the regression. For multiobjective forecast-informed reservoir systems, it is critical to understand and define the relationship between forecast accuracy and forecast value; if improvements in accuracy lead to steep gains in value, investing in further forecast model development may be warranted.