| 摘要: | 台灣存在著許多斷層分佈包括已知的地表斷層線以及盲斷層，因此了解基於何種物理條件下會造成多重斷層的破裂是極為重要的課題。斷層動態破裂模擬有助於了解斷層破裂面上的物理機制，隨著計算資源以及理論方法的優化，多斷層的破裂模擬也得以實現。 芮氏規模7.1之新竹台中地震發生在1935年4月21號，為台灣歷史上造成大規模傷亡的嚴重災害之一，同時也是存在著複雜斷層幾何構造的經典案例之一。了解1935地震的多重斷層破裂是基於何種物理條件有助於日後的地震危害度分析。根據現地的野外調查，地震發生後，地表有兩條主要的斷層破裂，分別為逆衝機制的獅潭斷層以及右移機制的屯子腳斷層，但震央並不在這兩條地表破裂線上，這也暗示著有盲斷層的存在。顏銘萱等人（2016）利用1935年前後的三角測量所得到之地表位移作有限斷層逆推得到最佳的斷層模型解顯示總共有四個斷層：獅潭斷層、屯子腳斷層以及兩條斷層中間的兩個不相連的盲斷層分別傾向東邊以及西邊。本研究中，我們針對兩種可能的斷層模型，並使用異質性之初始應力設置來計算錯動量分佈以及破裂速度並與過去的運動學逆推結果做比對。第一個模型A為參考顏等人提出的最佳斷層模型；第二個模型B則為加入一個連接斷層在兩個不相連的盲斷層中間。結果顯示兩種模型在錯動量分佈以及理論規模都可以符合觀測紀錄，但在應力設置時模型A在北邊的盲斷層需給定一個極高的應力值，才能使破裂從初始斷層跳躍到北邊的斷層，而模型B則可以利用較合理的應力值來使所有斷層破裂，然而在地表變形模擬方面，模型B明顯因為多了一塊連接斷層而造成整體的誤差值變大。這也暗示著模型A可能較為合理但同時也得滿足在部分區域需累積極高的應力。除此之外，兩者的震源時間函數有明顯的差異，本研究中也進一步利用波形模擬來解析兩個模型的差異並且根據模擬的結果來討論動態破裂模擬中參數的調整。結果顯示，兩個模型結果在體波的部分較相符觀測波形，暗示著在動態破裂模擬的震源參數結果是符合實際的情況。
;For a tectonically active collision environment, the island of Taiwan is populated with identified active faults and seismogenic structure within deeper depth without surface rupture. Historical severe damage earthquakes suggest the possibility of the multi-fault (seismogenic structure) ruptures. Understanding the mechanism of multi-fault rupture is an imperative target. Earthquake dynamic rupture modeling provides key element on understanding the physics of earthquake initiation and propagation. Also, through the advanced computational resources and methodology, the multi-fault rupture simulation becomes possible. The severe Hsinchu-Taichung Earthquake with local magnitude of 7.1 occurred on April 21st, 1935 from multi-fault rupture. Understanding the physical rupture process of this event might shed light on the future identification on fault system from multiple faults, and hints for seismic hazard potential evaluation. Historical data in geological investigation suggests this event involved two major faults: the Shihtan reverse fault and the Tuntzuchiao right-lateral strike-slip fault, both with ruptures to the surface. The hypocenter, however, did not appear to have been on either of these structures, suggesting the existence of unobserved blind fault. Yen (NCU, MS thesis, 2016) performed a finite-fault inversion that constructed a possible fault model by comparing the synthetic and observed surface displacement, and leveling data. The optimum model suggested the event was resulted from a fault system, which consists of four segments, including both reverse and strike slip fault, with the rupture-jumping between two blind faults that have opposite direction of dip, and thus increasing separation with depth. After construction of dynamic fault models with several trials of reasonable dynamic parameters, in this study, we propose two possible dynamic fault models, both associated with a heterogeneous pre- stress designed to produce slip similar to that of the inversion result. Model A is from geometries modified from Yen’s study as two faults with surface ruptures and two blind faults (jumping model). Model B is assuming a linking fault between two blind faults (linking model). Through further refinement of dynamic parameters, the correlation between the slip pattern derived from the dynamic model and inversion model, along with the jumping phenomenon, are used as critical factors in evaluating the reasonableness of our dynamic rupture model. We further examine several synthetic results with observation including surface displacement and seismogram to help us to understand the source characteristics of 1935 earthquake. |
