臺灣中央山脈南段長微震的定位誤差分析及改進
No Thumbnail Available
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
非火山長微震( Non-volcanic tremor )為常見於世界上主要的板塊邊界、對應著慢速滑移的一種特殊地震訊號，主要分布於環太平洋隱沒帶，多呈現深度集中、平行隱沒板塊的帶狀分布，大多發生於孕震帶更深處，其特徵主要為：(1)持續時間長，可以持續數分鐘至數小時；(2)振幅較小，無明顯的P波和S波；(3)主要頻段介於2-8 Hz之間。由於無明顯的體波波相，難以用傳統的P波與S波的到時來做定位，因此許多不同於傳統的地震定位方法紛紛被提出，常見的主要有Hypoecc ( Ide et al., 2010 )以及WECC ( Wech et al., 2008 )等，主要手段為利用波形包絡化後的相似度以決定各測站的相對走時差。在台灣，長微震大多發生在中央山脈南段下方，由於定位誤差大，對應的孕震構造仍有諸多爭論。本研究以WECC的定位方式為基礎，改變其網格搜索的方式，並使其能套用三維速度模型( Huang et al., 2014 )， 期能獲取更佳定位結果。我們利用 2016年1月1日到2016年9月15日BATS、CWB (寬頻及短周期) 以及在中央山脈南段新架設的三個測站的連續資料，本研究開發新的定位方法，並和傳統的WECC和Hypoecc定位結果比較。為了釐清定位誤差來來，我們更利用不同震源機制、不同子震源組成所得之合成波形，擬合長微震之波形特徵和持續時間，以討論(1)包絡化 (或平方)之波形形貌 (2)濾波頻段的選擇 (3)測站的包覆性與幾何形貌 (4)速度模型幾種因素對定位結果的影響。本研究發現，以新方法得到最小的定位誤差，水平和垂直方向分別為1.5-3.5 km以及7-14 km之間，比傳統的長微震定位方法少了約1 km的誤差，此定位結果配合最小平方差之回歸線性，顯示了長微震震源區為一東北-西南走向(70-80°)及向東南傾沒約40-55°之構造，和前人研究相比，在走向的定義上更為明確，這一個深部構造和過去Huang and Byne (2014) 提出的地質邊界-土隴灣斷層的走向一致，顯示此斷層可能為長微震在地表的地形地貌表現。
Non-volcanic tremor is a slow earthquake phenomenon commonly found in the plate boundary zones. They usually occur below the seismogenic zones with seismic characteristics of long duration of several minutes to hours, small amplitude without obvious P wave and S wave, and the main frequency band of 2-8 Hz. Given no obvious body wave phase, it is difficult to apply the traditional location scheme using P-wave and S-wave arrivals. Many relocation methods were proposed including Hypoecc (Ide et al., 2010) and WECC (Wech et al., 2008). These two methods consider envelope cross-correlation to determine the time lagses between stations. Previous studies indicate that tremors in Taiwan occurred at the depth of 25-45 km below the southern Central Range. The tectonic origin of the tremors, however, are still under debate due to the large locating errors > 7 km based on Chunag et al. (2014). The goal of this study is to develop a location scheme that allows the implement of three-dimensional velocity model and grid search method, to better understand the fault geometry of tremor sources. Instead of cross-correlation coeffificent, the new location method is weighted by the envelope cross-correlation derived time lapses. We simulate synthetic tremor waveforms with a variety of focal mechanisms and initial locations, to further discuss how the following factors influence the location error: (1) the shape of the envelope (or square) waveform (2) the selection of filtering frequency bands (3) the station coverage (4) choice of velocity model. The resulting location uncertainty is found to be smallest using the new location method, as 1.5-3.5 km and 7-14 km in horizontal and vertical, respectively. Applying this time-lapse-weighted location method on the continuous data from January 1st to 15th September 2016 (recorded at three different seismic networks including a small array deployed above the tremor zone), a linear structure with northeast-southwest trending and southeasting dripping is obtained. This structure is consistent with the Tulungwan-Chaochou fault system that was recognized as thegeological boundary between a slate belt and unmetamorphosed fold-and-thrust belt by Huang and Byne (2014).
Non-volcanic tremor is a slow earthquake phenomenon commonly found in the plate boundary zones. They usually occur below the seismogenic zones with seismic characteristics of long duration of several minutes to hours, small amplitude without obvious P wave and S wave, and the main frequency band of 2-8 Hz. Given no obvious body wave phase, it is difficult to apply the traditional location scheme using P-wave and S-wave arrivals. Many relocation methods were proposed including Hypoecc (Ide et al., 2010) and WECC (Wech et al., 2008). These two methods consider envelope cross-correlation to determine the time lagses between stations. Previous studies indicate that tremors in Taiwan occurred at the depth of 25-45 km below the southern Central Range. The tectonic origin of the tremors, however, are still under debate due to the large locating errors > 7 km based on Chunag et al. (2014). The goal of this study is to develop a location scheme that allows the implement of three-dimensional velocity model and grid search method, to better understand the fault geometry of tremor sources. Instead of cross-correlation coeffificent, the new location method is weighted by the envelope cross-correlation derived time lapses. We simulate synthetic tremor waveforms with a variety of focal mechanisms and initial locations, to further discuss how the following factors influence the location error: (1) the shape of the envelope (or square) waveform (2) the selection of filtering frequency bands (3) the station coverage (4) choice of velocity model. The resulting location uncertainty is found to be smallest using the new location method, as 1.5-3.5 km and 7-14 km in horizontal and vertical, respectively. Applying this time-lapse-weighted location method on the continuous data from January 1st to 15th September 2016 (recorded at three different seismic networks including a small array deployed above the tremor zone), a linear structure with northeast-southwest trending and southeasting dripping is obtained. This structure is consistent with the Tulungwan-Chaochou fault system that was recognized as thegeological boundary between a slate belt and unmetamorphosed fold-and-thrust belt by Huang and Byne (2014).
Description
Keywords
中央山脈南段, 長微震, 定位, 誤差分析, Southern Central Range, Tremor, Location, Error Analysis