碩士 / 國立臺灣海洋大學 / 海洋環境資訊系 / 103 / Rapidly deteriorating sea conditions pose threat to the sea traffic. When the wave height grew more than two times and grew to higher than 1.5 m in 6 hours, this event was defined in this study as a rapid growth wave event. This paper investigated two models based on data mining techniques to study the occurrence of rapid growth wave events occurred off the coast of Hualien. One technique is using artificial neural networks for predicting future wave height and then to determine whether the rapid growth waves occurred. The other data mining method investigated is to examine the metrological factors associated with these events by cluster analysis.
In the artificial neural networks model two output factors were established: the future wave height in two hours as well as the maximum wave height within the next six hours. Comparing the predicted wave height with the measured data, it was found that the correlation coefficient (CC) between the future 2-hour wave height and the measured was as high as 0.94 with the root mean square error (RMSE) about 22cm. The CC and RMSE for the 6-hour predicted wave height were 0.88 and about 38 cm, respectively. However, it was found that both the predicted wave heights cannot be used for predicting the rapid growth wave event. The reason is that either predicted wave height rarely larger than 3.5m, since high wave height and thus rapid growth events were rare. For this reason, the two output wave height in the neural network models cannot predict the occurrence of rapid growth event satisfactorily.
In the cluster analysis of the occurrence of rapid growth events, this study artificially increased the number of rapid growth event, by resampling the event data in the training stage. It was found that when the data ratio between the non-rapid growth wave and rapid growth wave was as high as 1 to 1 provided best result. The cluster analysis showed that the occurrence of rapid growth waves was heavily affected by the meteorological factors, such as the wind speed, the northeast component of the wind and the period. The characteristics of the rapid growth waves along the coast of Hualien can be observed as follows: one, the direction of the wind is mainly either from the northeast or southwest as the wind speed continued to increase; second, with no dominant wind direction but the wave period and wind speed continued to increase. The cluster analysis also showed that when the wind speed and wave height have already been high, the rapid growth wave did not appear easily, for it is difficult for the wave to further grow twice in height in 6-hours.
| Identifer | oai:union.ndltd.org:TW/103NTOU5276013 |
| Date | January 2015 |
| Creators | Chou, Cheng-Hung, 周正弘 |
| Contributors | Tsai, Chen-Han, 蔡政翰 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 54 |
Page generated in 0.0017 seconds