Assessing the value of seasonal flow forecasts in reservoir operations for drought management in South Korea

Abstract

Given the escalating uncertainty in water resources management attributed to climate change, the significance of reliable flow forecasts becomes increasingly crucial. Recent technological advancements have brought attention to seasonal weather forecasts, which provide predictions of weather variables for the next several months. Accordingly, numerous studies have investigated the skill of Seasonal Flow Forecasts (SFFs) forced by seasonal weather forecasts, in various regions and countries. Our previous work on the skill of SFFs across South Korea demonstrated that SFFs generally outperform Ensemble Streamflow Prediction (ESP) up to 3 months ahead and exhibit notably higher skill during the wet season in abnormally dry years. This study builds upon our earlier research with the objective of evaluating the value of SFFs for reservoir operations for drought management. This analysis is conducted for two pivotal reservoirs, Soyanggang and Chungju, which serve as the major water sources for the metropolitan area, including the capital city, Seoul. For the severe drought period from July 2014 to June 2016, we used model simulation to compare different reservoir operation models: the Simple Conjunctive Operation (SCO) model, forced by either a worst-case or low inflow scenario (similarly to what currently done in reality) and the Forecast-informed Conjunctive Operation (FCO) model, forced by either ESP or SFFs. Multi-objective evolutionary algorithm is utilised to optimise release scheduling with two objective functions: securing storage volume at the end of the hydrological year and minimizing supply deficit over the entire year. We explore the impact of using different reservoir operation models, different forecasting lead times, as well as different ways to determine a ‘best compromise’ solution between the two competing objectives.