DC 欄位 值 語言 DC.contributor 土木工程學系 zh_TW DC.creator 馮馨柔 zh_TW DC.creator Shan-Non Feng en_US dc.date.accessioned 2023-7-5T07:39:07Z dc.date.available 2023-7-5T07:39:07Z dc.date.issued 2023 dc.identifier.uri http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=109322090 dc.contributor.department 土木工程學系 zh_TW DC.description 國立中央大學 zh_TW DC.description National Central University en_US dc.description.abstract 2020 年，位於澎湖南方海域的澎湖南方四島發生嚴重的珊瑚白化現象且可 能造成大量珊瑚死亡，珊瑚為海洋中重要的生態庫，豐富海中的生物多樣性，但 珊瑚可以存活的溫度範圍狹窄，暖化造成的海溫(Sea Surface Temperature，SST) 上升影響不容小覷，珊瑚的復原時間也長達 10 年，若極端氣候持續的發生，珊 瑚將會更難生存，故能準確預測未來之海溫變化極為重要。 除了海溫外，澎湖附近海域之珊瑚白化指數 DHW (Degree Heating Weeks) 也在 2020 年達到史上最高，且近 15 年珊瑚白化風險透過分析結果發現有著明顯 的上升。本研究以深度學習方法，長短期記憶模型(Long Short-Term Memory, LSTM)結合卷積神經網路(Convolutional Neural Network, CNN)來預測海溫，並可 以利用預測之海溫反推算 DHW，海溫歷史資料之準確度高達 99.22%，對於預測 未來海溫及 DHW 可以提供良好貢獻。 澎湖在 2020 年也並未直接受到颱風的侵襲，少了颱風攪動海水，冷卻海溫 的效果，連續的高溫也是其中原因導致大面積的珊瑚白化，而颱風的強度、風速 及海洋混和層厚度都會造成不同程度的海溫冷卻，本研究分析了過去颱風事件得 到的不同因子對於海溫冷卻之相關性及貢獻程度外，也建立了機器學習模型隨機 森林分類器(Random Forest Classifier)預測並分類不同因子之颱風事件是否造成 顯著海溫冷卻效應，透過歷史資料預測也有 92.9%之準確率，其中以經驗正交函 數(Empirical Orthogonal Function, EOF)對於海溫的分組提供最顯著之海溫冷卻因 子。 除了分析數據資料外，本研究也到澎湖南方四島進行了 2 次現地調查，在 2021 年澎湖南方四島之珊瑚並未受到明顯珊瑚白化之影響。本研究建立海溫及 DHW 預測模型與颱風冷卻分類模型，希望未來可以提供相關機構預測之海溫 及珊瑚白化資訊，為澎湖南方四島之珊瑚保育盡一份心力。 zh_TW dc.description.abstract In 2020, the South Penghu Marine National Park had experienced sever coral bleaching and may cause coral deaths. It’s an important creature that enrich the biodiversity in the ocean. However, the SST that corals can survive is narrow, and the recovery time is about 10 years. The impact of extreme weather can’t be underestimate, and it is critical to accurately predict SST variations. In addition to SST, DHW near Penghu achieved the highest level in 2020, the risk of coral bleaching has increased significantly in the past. In this study, the deep learning method, LSTM and CNN, is applied to predict SST, and is already reached 99.22% accuracy. Penghu was not directly attacked by any typhoon in 2020, so lack of typhoon induced SST cooling effect, may be a reason that large area of coral bleaching. The intensity, wind speed, and the thickness of the ocean mixed layer would cause different degrees of SST cooling. A machine learning model, Random Forest, is established to classify whether this typhoon would cause significant cooling effect. The accuracy of the model is 92.9%, and the EOF analysis provides the most obvious cooling factor. In addition to analyzing the data, this study also conducted twice site survey in Penghu Marine National Park in 2021, and corals were not affected by obvious coral bleaching. This research establishes SST, DHW prediction model and a typhoon induced SST cooling model, hope to provide forecast information for relevant institutions and contribute to the coral conservation of Penghu Marine National Park. en_US DC.subject 珊瑚白化 zh_TW DC.subject 海溫 zh_TW DC.subject 深度學習 zh_TW DC.subject 颱風特徵 zh_TW DC.subject DHW珊瑚白化閾值 zh_TW DC.subject Coral bleaching en_US DC.subject Sea surface temperature en_US DC.subject Deep learning en_US DC.subject Typhoon characteristics en_US DC.subject Degree Heating Weeks en_US DC.title 以機器學習預測海溫及熱帶氣旋特徵對於珊瑚白化之影響 – 以澎湖南方四島為例 zh_TW dc.language.iso zh-TW zh-TW DC.title Applying Machine Learning to Predict the SST Variation and Tropical Cyclone Patterns on Influence of Coral Bleaching – Using South Penghu Marine National Park as an example en_US DC.type 博碩士論文 zh_TW DC.type thesis en_US DC.publisher National Central University en_US