Author(s): Seoyeong Ku; Jong-Suk Kim; Jongyun Byun; Hoyoung Cha; Jongjin Baik; Changhyun Jun
Linked Author(s):
Keywords: Statistical Prediction; Typhoon Track Similarity; Dynamic Time Warping
Abstract: Typhoons and tropical cyclones are among the most severe natural disasters, bringing torrential rains in a short period and causing devastating consequences such as loss of life and destruction of property. In response to this hazard, there has been interest in accurately predicting Typhoon-induced Accumulated Rainfall (TAR), particularly by exploiting the topographically induced congruence between typhoon tracks and resulting rainfall patterns. In previous studies, various methods (i. e., Fuzzy c-mean clustering and polygon indices) were used to assess the similarity of typhoon tracks. In this study, we utilized to measure the similarity of typhoon tracks based on Dynamic Time Warping (DTW), which evaluates the similarity of time series data. Using the best typhoon track data from the Regional Specialized Meteorology Center, Tokyo, and precipitation data from the National Oceanic and Atmospheric Administrations Climate Prediction Center, a total of 1122 typhoons were used from 1979 to 2022. To evaluate the similarity of the typhoon track based on the latitude and longitude of the center of the typhoon and the time-series characteristics of various meteorological datasets (e. g., daily precipitation, pressure, translation speed, etc. ), the DTW method was considered. By considering selected 6 typhoons, the typhoon tracks were grouped as highly similar, and the improved TAR based on the linear relationship between observed TAR and translation speed was used, reducing the uncertainty of the predicted TAR. Finally, the Optimal Ensemble Number (OEN) based on the minimum root mean square error was applied. These results showed reasonable performance better than the result of previous studies that applied Fuzzy c-mean and polygon indices. Based on these results, the DTW method can help improve TAR prediction because it can take into account not only the track of typhoons but also different time series data from typhoons.
Year: 2024