Quantification of the Risks on Dam Preliminary Release Based on Ensemble Rainfall Forecasts and Determination of Operation

Preliminary release conducted at multipurpose dams is the operation which aims to temporarily increase their flood control capacity by releasing water stored for water utilization ahead of a flood based on rainfall forecasts. However, if rainfall forecasts overestimate actual rainfall, there is a risk that the volume needed for water utilization may not fully recover after preliminary release. Consequently, because of apprehension of such a risk, preliminary release has been seldom operated in Japan. This paper introduces the determination of preliminary release based on ensemble rainfall forecasts. First, the two risks related to preliminary release were defined at first. One is the flood control operation for extreme floods, which is conducted when preliminary release has not been conducted or inadequately conducted, whereas the other is the water level may not fully recover to the level needed for water utilization if preliminary release is decided based on rainfall forecasts overestimating actual rainfall. Next, these two risks were quantified using the results from ensemble rainfall forecasting and hydrological simulation. Finally, based on the quantified risks, the operation of preliminary release was decided. The experimental simulation of preliminary release based on ensemble rainfall forecasts was conducted. The flood caused by Typhoon Man-yi in 2013 is selected as the experimental flood, and the Hiyoshi Dam as the experimental dam, where the flood control operation for extreme floods was conducted during the flood. The simulation results of the experimental flood showed that flood control operation for extreme floods would still be required notwithstanding the increased flood control capacity generated by preliminary release. However, the results also showed the possibility that preliminary release can delay flood control operation for extreme floods and considerably reduce the maximum outflow compared with the maximum outflow that would be required when no preliminary release is conducted. In addition, the study found that ensemble rainfall forecasting can be an effective tool to support the formation of a consensus among stakeholders such as dam managers and water users by quantifying and visualizing the risks associated with preliminary release, which cannot be accomplished with the conventional deterministic rainfall forecasting.