臺灣西部井下應變儀觀測研究-環境因子之影響與處理

為了解臺灣褶皺逆衝帶斷層活動引致的地殼變形，本研究收集中央地質調查所設置的13座井下應變資料(2005 ~2017)進行分析，旨在建立臺灣西部地區GTSM井下應變儀觀測資料的處理技術，以及各觀測站因地形與地質因素差異的區域化修正。在此研究中，我們建立了井下應變儀資料的處理流程，利用雙指數曲線擬合的方式有效地去除井孔回復與混凝土膨脹效應，並針對不同環境因素諸如氣壓、潮汐與降雨等影響提出修正之方法與參數。在利用潮汐分潮進行觀測數值校正後，各站的氣壓響應應變約在1 ~ 10 nanostrain/mbar，與降雨產生的最大應變響應大小基本一致，呈現良好的線性關係。此外，我們也針對均質耦合(isotropic coupling)與垂直耦合(vertical coupling)兩種不同的校正方式進行對比，發現在對應垂直應力產生的應變，兩者的修正差距可達30%。同時降雨所產的額外荷重對於觀測結果也會產生一定影響，降雨落在地面後，短時間內無法進入地下的雨水化為地表逕流，因其累積重量的影響，應變儀迅速地達到最大響應量，然後呈現指數性衰減，影響時長可達200小時以上。由於地下水對於井下應變觀測的影響非常顯著，我們嘗試利用獨立成分分析的方法來解析地下水變化在應變觀測資料中的影響。透過模擬訊號測試發現，當較少訊號來源(兩個)時，獨立成分分析的方法可有效且良好地分離出個別訊號，但當訊號來源較多時(三個以上) ，則分離的效果就會逐步下降。從對比結果來看，若提取比重最高的主要訊號進行成分比對，獲得分離成分的效果能有所提升。我們也收集2010年至2014年底5年期的應變資料，與規模6以上的島內範圍地震之間進行比對，發現地震發生時除較近距離的觀測站有大小不一瞬時的同震變形現象外，一些較遠測站在地震發生過後也發生緩慢且持續的變形，這些變形可能為岩層為配合斷層滑動後應力變化的調適，或可能為表面波震動帶來的短暫變形行為，亦或者是因為表面波通過觸發當地構造活動所產生的非震性變形。根據我們的觀察結果，在2010/03/04甲仙地震發生前約2個禮拜，嘉南地區多個測站都可觀察到震前異常的發生，時間點也與長微震發生率與持續時間與GPS觀測發生異常時間接近，顯示出發生地震前岩層整體確實有些變化，但是並無直接證據顯示此為甲仙地震的前兆現象，仍有待後續進一步的研究。井下應變儀的暫態應變改變，反映不同的不同環境因子和構造活動，如何分離不同因子的訊號特性，是將來研究重點。更多對這些環境因子的研究(尤以地下水最甚)除了幫助我們能更了解井下應變儀這個儀器的特性外，也可為後續進行相關研究者提供更多之參考依據。

關鍵字

井下應變儀；環境因子；應變變化；地震前兆；獨立成分分析

並列摘要

In order to understand the crustal strain in response to fault activity in fold-and-thrust belt in Taiwan, this study collects and analyzes the borehole strainmeter (BSM) data of Central Geological Survey (2005 ~ 2017) to establish the technology of processing BSM data and the method to diminish the environmental response caused by topographical and geological changes. We established the GTSM borehole strainmeter processing flow, using double exponential curve fitting to effectively remove the effects of borehole relaxation and grout curing, and proposed corrections and parameters for environmental responses such as air pressure, tides, and rainfall. After calibration using tidal constituents, the pressure response of each station ranges from 1 to 10 nanostrain/mbar, which is basically the same as the maximum rainfall response, and both show a good linear relationship. In addition, we also compared two different calibration methods, isotropic and vertical coupling, and found that the correction difference between both can be up to 30% in the strain corresponding to the vertical stress. The extra load produced by rainfall also plays a role on strain transient of the observations. In most cases, the rainfall response will reach the maximum in half an hour after heavy rainfall, and then show an exponential decay, which might persist more than 200 h depending on the hydrogeological condition around the station. In addition, groundwater also has a considerable impact on strain transient of observations. We try to use independent component analysis (ICA) to analyze groundwater changes in the strain data. When there are fewer signal sources (two), the ICA method can effectively separate out the individual signals, but when there are more signal sources (more than three), the results of separation will not be good enough. Since groundwater response accounts for a relatively high proportion of strain data, the use of principal component comparisons of different azimuth data can improve the results of hydrological data separation. We collected and compared the strain data since 2010 to 2014 and earthquakes with a magnitude of 6 or more in Taiwan. We found that not only instantaneous co-seismic deformations were recorded at some closer stations, but also some slow and persistent deformations were detected at relatively distant stations. These deformations may be the response to the adjustment of the stress changes after the triggering of far-field fault activity, or the dynamic transient deformation caused by the surface wave propagation, or some aseismic activities of the local structure triggered by the surface wave. According to our observations, the precursory anomalies can be observed at multiple stations in Jianan area about 2 weeks prior to the 2010/03/04 Jiasian earthquake, which some anomalies of tremor and GPS observation were reported in the same period of observation. This infers that the pre-seismic transient deformation could be detected with different observation. However, such precursory signals need to be analyzed in detail to characterize mechanical behavior of pre-seismic deformation. The data of the borehole strainmeter provides the valuable information for strain transient due to environmental change and tectonic activity, how to separate these factors is a crucial issue for future study. More research on these environmental factors (especially groundwater) will help us to better understand the characteristics of the borehole strainmeter, and also provide a good reference for those in follow-up study in the future.