유사태풍 통계기반 한반도 영향태풍의 누적강수량 예측

Abstract

This study aims to develop a statistical model for typhoon-induced accumulated rainfall (TAR) prediction over the Korean Peninsular (KP) using historical track, intensity, and rainfall data for 91 KP-influence typhoons during 1977-2014. Tracks of typhoon are obtained from the best track data archived by the RSMC Tokyo and rainfall data are from the 56 meteorological stations over the KP. The procedure for the statistical estimation of TAR consists of three steps: (i) estimating the TAR at 56 stations for the KP-influence typhoons during 1977-2014, (ii) selecting historical typhoons similar to the track of the target typhoon within the regions of 32-40°N and 120-138°E using a Fuzzy c-Mean clustering method, (iii) averaging the TAR for the top-16 typhoons with the highest track similarity after a bias correction of TAR based on a linear regression between TAR and intensity (maximum wind speed of typhoon) bias. Simulation results for 91 KP-influence typhoons revealed that the present model is capable to simulate TAR with an accuracy of 56mm mean root mean square error (RMSE) over 56 stations. For the validation of model, real-case predictions for Typhoon Chanhom and Goni in 2015 are performed. The result shows that the model predicted TAR with the RMSE of 33mm and 29mm for Chanhom and Goni, respectively, when the best track data are used. In the predictions using the predicted track and intensity data for Typhoon Goni, out model also showed an overall good performance with the RMSE of 28~50mm, varying according to the accuracy of the predicted track and intensity.