This study proposes the framework of real-time “probabilistic” storm water level prediction and then examines the prediction for a model typhoon, which crosses or passes by target ports in Japan's Seto Inland Sea. The prediction includes the storm surge simulation for 109 typhoon tracks and provides the probability of a disastrous storm water level for the port facility and coastal defense.
Every year, several typhoons approach the Japanese Islands and a few make landfall. Among recent typhoons, Typhoon Bart in 1999 brought a storm surge of approximately 3.5 m into Yatsushiro Bay, Kyushu Island, near the time of a spring high tide. The storm surge triggered flooding water levels, reaching the roofs of one-story houses on low-lying land, and took 12 lives in these houses. Fig. 1 shows the variation in the storm water level with time at the affected place. It was difficult to identify the difference between the actual storm water level and the usual astronomical tide level until 5:00 a.m., as well as to imagine that the storm water level had risen by 4 m until 6:30 a.m. owing to a significant storm surge and the usual rise in astronomical tide level. From such events, we should recognize the necessity of storm water level prediction for disaster prevention work, such as tide gate operation, sand bag piling and evacuation orders, before the storm water level reaches the point of disaster. Meteorological organizations not only in Japan but also in the other countries in the world issue the most probable typhoon track with the range of the standard prediction error. The persons in charge of disaster prevention work carry out the work if they feel there is a certain high risk of flooding; however, they do not know the probability that a higher-than allowable-water level will manifest.