中華白海豚哨叫聲偵測、模擬與定位之研究

Abstract

基於2025年非核家園的目標與國際間減碳的承諾，台灣政府如火如荼的發展自主性的綠色能源產業。伴隨著台灣海峽地理環境的優勢，台灣西部海域乃非常優良之風場，政府也把握此良機，積極投入建設西部海域離岸風場。然而其比鄰中華白海豚重要棲息環境，打樁建造過程與之後的營運勢必造成對生態環境的影響，一套完整的鯨豚監測系統將是未來的發展重點。被動式聲學監測(Passive Acoustic Monitoring, PAM)即為廣泛被使用且效率極高之監測方法，不僅彌補目視法的缺陷，還可同時記錄海洋背景噪音與鯨豚叫聲，理解兩者的關係。 本研究於台中港北堤外海佈放四組水下麥克風監測站，為了測試此架構的效能，使用人造哨叫聲訊號模擬真實中華白海豚哨叫聲，經過臺大水聲實驗室自行開發之哨叫聲偵測器抓取訊號到達時間片段，最後計算訊號之到達時間差，完成(Time Difference of Arrival, TDOA)定位。由於整體PAM架構尺度達數公里，定位結果呈現出離監測站愈遠之聲源，定位效果愈差，平均定位誤差為105公尺。另一方面，為了瞭解訊號在水面下的傳遞，以增進水下麥克風站的監測效能，本研究使用Bellhop聲線追蹤模型模擬人造哨叫聲在水中的傳遞。Bellhop可以輸出多種結果，包含特徵聲線、傳輸損耗、到達時間與接收端時序列。透過將原始訊號與Bellhop輸出結果計算而得之脈衝響應進行摺積，即可得接收端訊號時序列。將模擬結果與台中港實驗資料進行比較，在波形與頻譜圖分析上，相關係數均大約為0.5，結果表明此模擬方法具備一定精度與準確性；此外，更可利用模擬訊號之到達時間回推出訊號源位置。整體而言，本研究未將來PAM關鍵技術提供了初步的參考方向。

For a nuclear-free homeland in 2025 and considering their international carbon reduction commitment, Taiwan’s government is developing an independent green industry. The development of offshore wind farms along the western coast currently is the government's main energy policy. Nevertheless, the wind farm is adjacent to the Chinese White Dolphin reservation zone. A passive acoustic monitoring system for cetaceans will be established as a development priority in the future wind farm operation. PAM (Passive Acoustic Monitoring) is used widely and its efficiency is much higher, so we use PAM to monitor the ocean background noise and the cetacean call. In addition to making up for the defects of the visual method, it also can understand the relationship between the two. This study deployed four hydrophone stations near Taichung Harbor. To testify this system’s capability, this study used a human-made dolphin whistle signal to imitate a real dolphin’s call, and used a self-developed whistle detector to capture the signal. Moreover, we used the time difference of arrivals (TDOA) method to compute the localization of signal sources. This TDOA experiment was conducted near Taichung Harbor. The result showed that the average of the positioning errors was 150m due to the long range hydrophone array. On the other hand, to improve this algorithm, it’s important to understand the transmission of the whistle signal. This study also used Bellhop ray tracing program to simulate the transmission of the human-made whistle signal. Bellhop can produce several useful outputs including eigenrays, transmission loss, arrival time and received time series. By convolving the source signal with the Bellhop output impulse response function, we can produce a receiver time series. After comparing the simulation result with the Taichung Harbor experiment result, it shows that this simulate method possess certain accuracy and similarity, due to the 0.5 correlation coefficient on waveform and spectrum. And we can also derive the source signal location from the arrival time produced by the Bellhop. Over all, this study provides initial reference directions for key technologies of the PAM used in Taiwan.