| Summary: | 碩士 === 國立臺灣海洋大學 === 海洋環境資訊學系 === 95 === Swells generated by a distant typhoon can be felt prior to the landing of the typhoon. The issuance of typhoon warnings by the Central Weather Bureau are based on the distance of the typhoon from the coast, rather than the arrival of typhoon waves. For this reason, the arrival time of typhoon waves may not coincide with the initial warning issued by the weather bureau. This in turn may pose danger for people participate in coastal activities.
This study examined the wave records measured at LongDong and KueiSan Island, and determined the time of arrival of typhoon waves and the time of the sea out of the influence of the typhoons after they left Taiwan. Based on the typhoon data reported by the Central Weather Bureau and Joint Typhoon Warning Center of the US navy, the distances of the typhoon at the times of swell arrival and the ending of the influence of typhoon waves were determined. From the obtained meteorological typhoon data, the center pressure, the typhoon radius and the typhoon speed were also determined. With these parameters, data mining techniques, linear regression and decision tree, using Microsoft SQL Server 2005, were used to analyze the relationship between the typhoon distance and these calculated typhoon parameters for five typhoons with type I route.
Parameters from the Typhoon Herb, which was also a type I route typhoon but not included in the learning process, were used to verify the relationship determined by the data mining technique. It was found that the relationship using decision tree better predict the distances of the starting and ending influence of the typhoon waves than that obtained by linear regression, and the prediction of the distance with the decision tree came within 35km. Since there were only five typhoon used in the learning process, it is suggested further investigation should be carried out once more typhoon data become available.
|
|---|
