Earthworm平台應用於臺灣地震速報預警研究

Abstract

地震預警系統在中央氣象局運作超過十年，過去僅有少數特定機構接收此系統產生的預警訊息。從2012年起中央氣象局使用Earthworm平台整合來自不同種類的即時觀測資料，並且以此系統進行臺灣地震活動監測。本研究發展的Earthworm Based Earthquake Alarm Reporting (eBEAR) 系統與過去系統相比，能夠縮短資料處理時間並且提升預警資訊的精準度。eBEAR系統中包含三個於Earthworm環境下新開發的模組。這些新的模組可以處理：P波到時挑選、波相組合、地震定位、規模計算及預警訊息發布。本論文主要內容在闡述eBEAR系統的方法與成效。為了調整此系統，選取154個規模4.0到6.5地震進行離線測試。測試結果顯示，平均地震定位誤差為4.2公里，規模誤差為0.3，系統發佈預警訊息時間為地震發生之後14.7秒。另外，實際系統運作情形顯示對於島內及島外地震，平均處理時效分別為地震發生之後15秒及26秒，比起過去的預警系統平均島內地震快3.2秒，島外地震快5.5秒。目前eBEAR系統已經將預警訊息於地震發生之後，即時地傳遞到全國中小學，以爭取在強烈地震波抵達學校前的數秒到數十秒時間發出警報。為持續提升地震監測能力，強化地震預警系統，本研究將543個低價位地震儀與中央氣象局地震觀測網整合，建置更高密度地震觀測網。選取46個規模4.5到6.5地震進行離線測試。測試結果顯示，系統發佈預警訊息時間可從地震發生之後14.7秒，進步到13.1秒。 目前地震預警系統仍存在許多待突破的困難，本文建議如下:對於測站覆蓋的空缺角(GAP)過大時，造成地震定位精確度不良的問題，透過分析現有測站覆蓋度，建議在台灣東部地區增加測站密度；對於規模大於7以上地震可能造成的規模低估情形，建議延長P波時間窗，以利完整記錄斷層錯動時所釋放的能量；對於短時間內發生數個地震導致預警系統可能漏報的情形，建議以即時震度分布圖估算地震大略位置與規模。 由於即時地震資料無論是來自不同的地震觀測儀器，或是來自不同的地震觀測機構，都能夠經由Earthworm軟體整合至同一個作業平台。因此在Earthworm環境下所發展的地震預警系統(eBEAR)能夠有效地在不同的觀測環境下運作。目前eBEAR系統已經在印度、韓國及太平洋海嘯警報中心等機構測試。

For more than 10 years, the Central Weather Bureau of Taiwan has operated an earthquake early warning (EEW) system and has issued warnings for specific agencies. Since 2012 the Earthworm platform in Taiwan has been used to integrate real-time seismic data streams from different types of seismic stations and to monitor seismicity. Using the Earthworm platform, the Earthworm Based Earthquake Alarm Reporting (eBEAR) system is currently in development for shortening reporting times and improving the accuracy of warnings for EEW purposes. The eBEAR system consists of new Earthworm modules for managing P-wave phase picking, trigger associations, hypocenter locations, magnitude estimations, and alert filtering prior to broadcasting. Here, we outline the methodology and performance of the eBEAR system. To calibrate the eBEAR system, an offline test was implemented using 154 earthquakes with magnitudes ranging from ML 4.0 to 6.5. Comparing between the eBEAR and the CWB catalog the results from the offline test show that the epicenter error is about 4.2 km, the standard deviation of magnitude is about 0.3, and the reporting time is about 14.7 s. Additionally, in a comparison of online performance using the current EEW system, the eBEAR system reduced reporting times and improved the accuracy of offshore earthquake locations and magnitudes. Online performance of the eBEAR system indicated that the average reporting times afforded by the system are approximately 15 and 26 s for inland and offshore earthquakes, respectively. The eBEAR system in average can provide more warning time than the current EEW system (3.2 s and 5.5 s for inland and offshore earthquakes, respectively). The eBEAR system now delivers warnings to elementary and junior high schools in Taiwan. For further improving the capabilities of monitoring earthquakes, an EEW system with dense seismic network is constructed by deploying a total of 543 low-cost sensors in Taiwan and incorporating with the official seismic network of Taiwan’s Central Weather Bureau (CWB). The experiment results show that the integrated system can have stable results of source parameters and issue alarms faster (from 14.7 s to 13.1s) than the current system run by only the CWB seismic network (CWBSN).