Exploring the value of seasonal flow forecasts for drought management in South Korea

Drought poses significant challenges across various water-dependent sectors. In the past few decades, numerous devastating droughts have been reported worldwide including in South Korea. A recent drought in South Korea, which lasted from 2014 to 2016, led to significant consequences including water restrictions and nationwide crop failures. Historically, reservoirs have played a crucial role in mitigating hydrological droughts by increasing water supply stability. With exacerbating intensity and frequency of droughts, enhancing the operational efficiency of existing reservoirs becomes increasingly important. This study examines the value of Seasonal Flow Forecasts (SFFs) in informing reservoir operations during three historical drought events, with a focus on two key reservoir systems in South Korea. For these events, we simulate what would have happened if the reservoir managers had optimised operations using SFFs. For comparison, we also simulate the effect of reservoir operations optimised using two deterministic scenarios (worst-case and 20-year return period drought) and another ensemble forecasts product (Ensemble Streamflow Prediction, ESP). We repeat our simulation experiments by varying the key choices in the experimental set-up, i.e. the forecast lead time, decision-making time step, and the method for selecting a compromise solution between conflicting objectives. We then propose a new, simple and intuitive method for measuring the value of the different scenarios/forecasts, based on the frequency of outperforming (in a Pareto-dominance sense) the historical operation across such experiments. Our findings indicate that while deterministic scenarios show higher accuracy, forecast-informed operations with ensemble forecasts tend to yield greater value. This highlights the importance of considering forecast uncertainty in optimising reservoir operations. Although SFFs generally show higher accuracy than ESP, the difference in value is small. Last, sensitivity analysis shows that the method used to select a compromise release schedule between competing operational objectives is a key control of forecast value, implying that the benefits of using seasonal forecasts may vary widely depending on how priorities between objectives are established.