2005年宜蘭雙主震之震源機制解: 應用近場波形反演法探討區域地震機制

Abstract

宜蘭雙主震發生於2005年3月5日(東經121.84度, 北緯24.66度), 第一主震深度為10.54公里，在19:06分發震，1.08分鐘後第二主震又在深度6.2公里處發震。兩主震發生時間及位置都十分相近，故一般合稱為雙主震。此雙主震由中研院台灣寬頻地震(BATS)所發布之波形反演地震斷層機制解，兩者均判定為左移之走向滑移斷層，但在氣象局(CWBSN)由P 波初動解出的斷層機制解，第一主震則為東北-西南走向之正斷層。此處震源之爭議點在於其可能的構造意涵而引發學界關注，由於宜蘭平原向來被認知為台灣東部構造連接至沖繩海槽張裂帶的轉換，且為沖繩海槽的向西延伸，目前論述與模型繁多。而清楚底定雙主震之破裂機制，因其特殊位置，正可助於釐清關於構造轉換點的問題，又單就其一主震，有機會進一步探討制定地震斷層機制解在方法上的差異。 本研究中我們針對宜蘭雙主震的第一主震，以兩種方法結果做比較。由於加上在宜蘭平原上密布之台灣強地動網陣列(TSMIP)，一方面有更精確的 P波初動解，另一方面可對加速度積分之地表位移做區域性的波形反演測試。據詳細理論方法的推斷，以及配合餘震叢集的結果，我們認為兩種方法呈現不同的地震破裂意涵。P波初動只使用地震波中最初波相所具有的訊息，表現了斷層初始破裂的運動，而波形反演則使用完整波形的訊息，代表了整個破裂過程平均的運動。因此，兩種不同方法所得地震斷層面解有所差異，是印證理論及反應此地震的區域特性的結果。 另外，在區域波形反演方面，我們延用R. B. Herrmann(1985)，並由黃柏壽(1994)和張道明(2005)修改之波數積分法產生合成波形。首先以平原上23個測站單站波形擬合，發現單站無法配合此區域側向上速度模型的不確定性，使得反演結果並不穩定，因而改以15個測站共同反演，並且引用吳逸民(2007)最新台灣區域的地震層析成像結果，截取宜蘭地區平均的一維速度模型，並加上宜蘭平原下低速層的區域構造。多站反演結果顯示，共同反演確實可以得到較為可信的斷層機制解。即使降低了各站波形的擬合度，多站反演出的單一解其穩定性、及與餘震之斷層機制解對比，仍表現出較合理的特性。根據本研究的結果，此一主震應為帶有正斷層分量之左移走向斷層。由這個證據應可推估，張裂與走向滑移的轉換點，最可能發生在宜蘭平原此一位置。

On 5 March 2005, an earthquake pair occurred at Ilan (121.84°E, 24.66°N), northeastern Taiwan. Both mainshocks have similar magnitude (ML = 5.9 and ML = 6.0). The first one was determined as a normal solution based on first-motion polarities of P wave from Central Weather Bureau Seismic Network (CWBSN) and Taiwan Strong Motion Instrumentation Program (TSMIP), while it was solved as a strike-slip solution by Broadband Array in Taiwan for Seismology (BATS) Centroid Moment Tensor (CMT) inversion. Normal faulting would be consistent with the opening and extension of Okinawa Trough on tectonic basis. On the other hand, a strike-slip solution would provide us one possible answer to the slip accommodation of compression. Thus, obtaining one reliable solution to this focal mechanism has become one important issue. To resolve the focal mechanism mentioned above, we target the first mainshock for its solution inconsistency in two methods. Firstly we applied Moment Tensor Inversions using much denser TSMIP strong motion array in Ilan region. The kernel computer code was adopted from R. B. Herrmann (1985) and later modified by Huang (1994) and Chang (2005). Synthetic waveform is generated with wavenumber integration method to simulate the three-component time history from hypothesized point source. Results from 23 stations using single station approach suggest that lateral variance of velocity structures underneath Ilan plain are dramatic. Multi-station inversion helped us to model this event at threshold magnitude(ML = 5.9) as point source. It serves to cancel out velocity variances from each station. We further applied multi-station approach to conceal velocity variation along different paths. The inversion result from fifteen stations provided a more convincing, averaged slip solution. This event was an oblique event in between normal component and strike-slip. The first motion solution simply presents initial rupture motion of an event, which could be irrelevant to subsequent slip motion of the whole cluster. By our result, we can conclude the initial rupture motion to be normal, and later strike-slip motion dominated the rest of this event. The multi-station approach is applicable for near-field TSMIP records to obtain reliable CMT solutions.