アンサンブル予報を用いた台風性降雨シナリオに関する検討

Abstract

Forecast circle has been used in Japan Meteorological Agency as a standard content to describe forecast uncertainty on typhoon central position. Although circle is a simple and familiar content for the general public use, it is not easy either to imagine high probability area inside circle or to suppose worst-case scenarios of typhoon-related rainfall in detail. We propose more beneficial method which represent typhoon track and typhoon-related rainfall using ensemble forecast dataset. Ensemble model of ECMWF consists of 50 perturbed and one control is used. Extracted tracks are classified to multiple groups using cluster analysis in consideration of the direction as well as the speed of typhoon. After bias correction considering frequency distribution, it is possible to obtain worst-case or mean scenarios of rainfall from each groups. We applied this method to Kanto-Tohoku heavy rainfall by typhoon No.18 in 2015. The result shows that the temporal and spatial characteristic of maximum rainfall extracted from most similar group to actual track is good agreement with the observed data in Kinugawa, and the water level exceeds flood danger level for a long time in Tamagawa in the case that typhoon track slightly shifted to the west.