台灣非火山長微震半自動化偵測系統

Abstract

非火山長微震 (Non-volcanic tremor, NVT) 為常見於世界上主要板塊邊界之特殊地震訊號。其特徵為︰(1)持續時間長，可從數分鐘至數十分鐘；(2)振幅微弱；(3)無明顯體波波相；(4)主頻介於1-10 Hz之間。由於人工辨識長微震困難度高，傳統地震學利用S-P波走時差之定位方式難以適用，因此許多專為長微震設計的偵測方法紛紛被提出。長微震依形成機制可分為觸發型長微震(triggered tremor)以及自發型長微震(ambient tremor)。觸發型長微震多由遠震表面波所觸發；自發型長微震則被解釋成與流體存在有關，或是板塊交界面上剪力滑動所造成。台灣過去曾於1998-2010年間對觸發型長微震進行系統性的搜尋，並發現數個由遠震表面波所觸發的長微震事件 (Peng and Chao, 2008；葉庭禎, 2011；Chao et al., 2012)。然而對於自發型長微震之研究仍相當缺乏，其發生位置、時間特徵、發震機制仍未有充分討論。本研究針對台灣中央山脈南段之自發型長微震，設計半自動化偵測系統，分別針對(1)不同測站具一致波形；(2)振幅大於背景雜訊；(3)持續時間夠長，設計三階段的門檻篩選，並利用此〝似長微震目錄〞進行人工判識。分析2009-2010年間連續資料，結果共定義121個明顯的自發型長微震事件，持續時間由5至30分鐘不等。其震源位於中央山脈南段孕震帶以下，深度多集中於16-34 km間，水平空間跨距約23 km × 55.5 km。分析長微震之累積持續時間，我們發現長微震活動度於2個區域M>6.0地震發生後明顯變高，但此現象無法以簡單的靜態庫倫應力改變來解釋。我們亦比較觸發型與自發型長微震之時空分布與訊號特徵，結果顯示觸發型長微震與自發型長微震之頻譜曲線類似，且振幅與持續時間關係均遵循指數關係，因此可對應至類似的震源過程。

In a transition zone between stable sliding and locked faults, occurrence of noise-like, non-volcanic tremors (NVT) are found to play an important role in full spectrum of strain release processes. The triggered and ambient tremor episodes previously discovered in Taiwan illustrates the generation potential of non-volcanic tremors underneath Central Range. However, small tremor bursts (several to tens of minutes) with high noise level and spare seismic station coverage makes the automatic detection in this region problematic and challenging. We propose an automatic detection scheme to identify tremors in this particular region. Applying this detection method in the two years data starting from January, 2009, we found 121 ambient tremor episodes with duration ranging from 5 to 30 minutes. The detected tremor events are confined in a small area of 23 x 55.5 km2 with the depth range of 16-34 km. We also obs erve larger number of tremors at the time of two local M6 earthquakes. Such acceleration lasted 70 and 90 days after the mainshocks. Given the uncertainty of assumed tremor source model, the static stress triggering relationship between M6 event and tremor is inconclusive. The scaling relationship between the duration and amplitude of ambient tremors fits an exponential model, suggesting a scale-bound source process.