| 摘要: | 地震於斷層上的滑移空間分布並不均勻，中大型地震於斷層上最大錯動的位置不必然位於地震震央，有限斷層逆推技術(finite-fault inversion)所得之地震滑移模型已證實此特性。而最大錯動位置附近往往為災損嚴重之區域，且隱含孕震構造的訊息。若能即時計算出此位置，可為地震後續危害評估爭取寶貴時間。然而，利用有限斷層逆推技術分析地震破裂空間分布需要較長運算時間，無法於地震發生後立即獲得結果，為了填補此空窗期，本研究運用前人所研發的震源掃描法(source-scanning algorithm)，配合台灣地震預警系統 P-alert，試圖在地震發生後數分鐘決定地震最大錯動量位置。本研究的目標為台灣地區2013年以來，有有限斷層模型的地震共7個及沒有模型的4個，包含2016年M6.6美濃地震等災害型地震，使用P-alert加速度地震資料。我們透過震源掃描法搜索最大滑移量可能發生之區域，其概念與波束成形(beamforming)相似，理論到時計算使用台灣三維速度構造，最後將本研究結果與有限斷層解進行比較。本研究結果發現規模大於5.5解析力較好，原因為地震具有較明顯的S波，接收到地震波的測站固然也較多，數據多分析也更為精確；而測站數量及距離震央的範圍也甚關重要，經多方測試後得知，範圍越小測站數目不夠導致結果失準，範圍過大S波又易受表面波干擾，因此最佳結果為震央方圓50~70公里結果最優良，考量運算效率，本研究選擇50公里作為分析參數；除了範圍，測站包覆性也會影響判斷結果，因花蓮地震時而發生於外海且測站皆位在陸地，使水平方向上解析能力受限，故得知發生於陸地之地震解析會較為準確，總歸研究得證大多地震水平向解析力皆良好，然而我們發現部分地震最大滑移帶深度解析力較為不佳，增加內圈測站資料權重似能解決此問題。本研究結果顯示，利用地震預警地震網能於地震發生幾分鐘內得到最大錯動量位置，未來計畫整合至P-alert地震預警系統中提供學術與防災單位參考。;According to the slip models from the finite-fault inversion technique, we understand that spatial distribution of slips on faults for a large earthquake may be heterogeneous. It causes that the largest ground shaking (and damage) on the surface may not be related to the epicenter but to an area with the largest slip. This area also indicates information of the seismic source (e.g., which fault sliding). It is important to efficiently determine the area for further seismic hazard assessments. However, since determing a stable finite-fault model requires a long computing time (couple days), we cannot have details of the source characteristic immediately when an earthquake occurring. To fill this window gap, we use the source-scanning algorithm (SSA) technique and consider data from a Taiwan earthquake early warning system, called P-alert. In this study, we analyze 7 earthquakes with the finite-fault models, including the 2016 M6.6 Meinong Earthquake and the 2018 M6.3 Hualien Earthquake, and 4 events with no fault models in Taiwan since 2013. We use the SSA approach, whose concept is similar to beamforming, to search for the area where the maximum slip occurred. The Taiwan 3D velocity model is applied. We then compare the determined maximum slip from SSA with it from the finite-fault model for each event. Our results show that the resolution of the earthquakes with a magnitude greater than 5.5 is better due to their stronger S waves and more available seismic records. Furthermore, number of seismic stations and their epicentral distance may be important. After a test, we find that applying stations with an epicenter distance ranging from 50 to 70 km is the best choice. Considering the calculation efficiency, we choose 50 km in the present study. In addition, the coverage of the seismic stations also affects the SSA results. The events occurred offshore show worse results. In addition, we discover that the results have the good spatial resolution in horizontal location, but the slightly poor resolution in vertical location. To solve this problem, increasing data weightening of the stations near the epicenter seems to a potential method. The results of this research show that the earthquake early warning seismic network can be used to obtain the maximum ground motion within a few minutes after the earthquake. We plan to combine SSA into the P-alert earthquake early warning system to provide reference for academic and disaster prevention units. |
