| 摘要: | 歐亞大陸板塊跟菲律賓海板塊兩板塊彼此間的相對碰撞運動是造成台灣複雜地體構造的主要成因。地球物理學的相關研究多以震波速度構造來探討地下構造的形貌。其中地震波到時為反應震波速度最直接的第一手資料。本論文主軸為透過震波到時的觀點來探討有關台灣速度構造的相關問題，主要分為參部份：(1)現有已知台灣三維速度模型的評估與改善；(2)依走時資料庫發展簡易的三維震源定位法；(3)地利地區的寬角度震測資料處理與由走時資料進行走時逆推，並由震測剖面建構台灣中部的二維速度構造。 過去十五年來利用地震到時資料與透過層析成像法所建構的台灣三維速度模型，包括Roecker et al.（1987）、Rau and Wu（1995）、Ma et al.（1996）、Cheng et al.（1999）、Kim et al.（2005）和Wu et al.（2007）。其中Rau與Ma兩速度模型的適用性已於早期相關研究中評估過(李坤松, 2002)。近期較具代表性的則為Kim與Wu兩速度模型。本論文針對此兩模型，同時結合陳美玲(2006)所建構的台灣西部沈積盆地速度構造，進一步修改了台灣西部深度五公里以內的三維速度模型。利用三維有限差分法及已知的921與331歷史地震，進行台灣三維參考速度模型的評估。由比較理論到時與觀測到時間的差值為準則，可判斷Kim的原始速度模型有相對較小的到時差。當考慮納入淺部沈積盆地的速度修正(陳美玲, 2007)後，亦證實Kim的淺層速度修正量也較小。由此評估經進一步修改過後之Kim的速度模型在淺層速度的三維分佈方面有較大的可性度。 2008、09年間執行的陸上高能量寬角度炸測實驗(TAIGER)其研究目的為探討深部地殼與上部地函的地下構造。於2006年10月19日在南投縣地利鄉先行實施一較小型的炸測實驗，其目的之一為利用該次實驗來瞭解五百公斤炸藥使用的安全性及估算其能量在台灣地質與岩性構造影響下的有效傳遞與隨距離增加的振幅衰減特性。 本論文亦透過一簡單、有效的思維來瞭解與解決傳統快速地震定位的問題。結合定位問題與資料庫的概念，定位問題可以快速且有效的找到最佳解。在演算法設計上，首先以三維有限差分法建構各地震觀測站的全台灣理論初達波到時與三維波前空間分佈資料庫，由各測站所接收到的觀測初達波到時，經由此資料庫可迅速有效的搜尋出所有可能的震源位置。由這些非唯一的震源位置再按已知的三維參考速度模型與各測站的理論到時，求取理論到時與觀測到時間的最小方均根到時差來推估最有可能的震源位置。透過此次地利人工炸測實驗與已知的震源位置GPS資訊，同時比較Kim與Wu兩種不同參考速度模型的計算結果來推估震源位置。由自行研發的三維定位結果亦得知Kim的三維速度模型有較佳的解。 TAIGER寬角度炸測可以有效的探討台灣深部地下構造。本論文利用地利炸測的震測走時資料，透過二維走時逆推建構出台灣中部二維速度構造。原始震測資料經過帶通濾波保留10?40 Hz的訊號，接著使用預測解迴旋法，壓縮訊號波形並降低震源鈴盪的效應。進行逆推時其在初始模型的建構上則參考修改後之Kim的三維速度模型為基準，逆推過程中首先以初達波到時決定淺層速度值與淺層速度不連續面位置，進一步由遠支距資料中標定反射波相到時，推估該反射面深度約為24公里。而由測站間距為兩百米間距的CMP剖面中，約11秒及15秒兩處的反射訊號進行逆推，其結果產生收斂不佳且走時誤差相對較大的解，由此結果無法完全排除該區域具有較複雜的三維構造所造成之反射資料特性。 就目前研究成果而言，(1)Kim的台灣三維模型經修改後有較佳的可性度，可依此做為目前台灣地區的主要三維參考速度模型。(2)而三維定位方面，於地震觀測網內的地震皆能有不錯的解，但觀測網外的地震仍需持續改良所發展的演算法來求得更精確的解。(3)地利地區的淺層速度構造，經由走時逆推可建構出符合目前已知相關地質背景的二維速度模型。(4)而深部構造反射波探察方面，因受限於資料數量與品質的不足，無法獲得更精確的解釋。未來針對TAIGER炸測實驗所接收到大量且詳細的震波走時資料，將可更進一步研究震波走時相關的研究課題。 Complex tectonic structure in Taiwan is strongly affected by the arc-continent collision between Philippine Sea Plate and Eurasia Plate. Geophysical studies involving wave propagation within complex media requires the fundamental travel-time information for structure imaging and related studies. The goal of this thesis have three folds, mainly on travel-time information contents, involving research efforts (1) to evaluate Taiwan 3-D reference models, (2) to propose a simple and effective methodology for absolute earthquake location and (3) to investigate a 2-D velocity structure profile across Western Central Taiwan through travel-time inversion. For the past 15 years, various 3-D models were proposed including Roecker et al., (1987), Rau and Wu (1995), Ma et al. (1996), Cheng et al. (1999), Kim et al. (2005) and Wu et al. (2007). Previous study on Rau and Wu (1995) and Ma et al. (1996) has been intensively studies (Li, K. S., 2002). For most recently published 3-D velocity models, including both Kim3D and Wu3D, the corresponding 3-D travel-time trajectories that correlate with wave-fronts were compared to evaluate its feasibility on various possible applications shown in this thesis. The travel-time computation is base on the modified expanding wave-front tracking approach proposed by Vidale (1988, 1990) through finite-difference approximation. By incorporating neo-tectonic basin structure in the Taiwan Strait and western foreland basin (Chen, 2006), the newly established 3-D reference models (MKim3D and MWu3D) can be individually identified and evaluated through the comparison on the differences in travel-time values, earthquake location and initial model setup for DiLi test shot profile. The results from both synthetic and real data applications all indicate that both Kim3D and MKim3D have smaller travel-time differences than Wu3D and MWu3D and thus have more reliable applications in there related research topics. By comparing the efforts to incorporate basin structure into existing models also indicate that spatial averaging of velocity variations in Wu3D model has to be taken care of in order to produce reasonable model. Such efforts can be visualization through comparison of travel-time differences before and after the modification. The purpose of TAIGER experiment during 2006 to 2009 is to investigate main seismogenic structure features of the crust and preferable even up to upper mantle through both active and passive source experiments. The primary goals for DiLi test shot experiment is to understand the wave propagation and attenuation characteristics across western foreland basin and the safety concerns while utilize 500 Kg explosive as effective energy source. One of the seismic dataset recorded by Texans recording array in the DiLi experiment is used for this study. A simple and effective approach through 3-D travel-time calculations for absolute earthquake location problem is proposed. By combing efficient 3-D calculation of travel-times and the construction of necessary database, non-linear location problem can be solved through recursive forward simulation of travel-times and the minimization of root-mean-square travel-time errors. Both synthetic tests on errors involved in finite-difference computation, consideration of spatial coverage of stations and influence on the existing velocity model were performed in order to understand the fundamental issues involved in absolute earthquake location problem. Study results show that the proposed algorithm is effective and the efficiency is subject to how detail the travel-time databank is demanded, the capacity of the hard disk and the efficiency of data fetching mechanism. Both synthetic and real data tests all show that the proposed methodology is workable for in-land earthquakes but not for the offshore earthquake as demonstrate from PingTung Earthquake synthetic case study. Further modification of the proposed algorithm can be pursuit by using S-P time and databases produced from the currently developed methodology. Wide angle refraction and reflection (WARR) study of DiLi test shot data involving data processing indicated that minimum phase and predictive de-convolution procedures are necessary in order to reduce strong oscillating source wavelet signature. The de-convolved data has the advantage for travel-time picking and phase identification. Band-passed filtering (10-40 Hz) of the recorded data enhanced the possible reflection (?) and/or diffraction (?) arrivals. Shallow basin structure and faulted block determined from first-arrival travel-times can be well defined both from near-source CMP and WARR sections. Four major faults with clear velocity change are clearly identified. However, the questionable arrivals at 11 and 15 seconds are not well constrained through 2-D travel-time forward modeling and inversion. The final model obtained from identified phases show two distinctive reflectors at 12 km and a clear reflector patch located at depth of 24 km. The interpretation of final velocity structure can be correlated well with results from seismic survey and gravity studies and the background geology. The apparent 3-D propagation effects and/or possible diffractions (?) for 11 and 15 sec phases that were caused by the complex wave propagation phenomena can be verified through explicit consideration of 3-D effects. The quality and quantity of the DaLi test shot dataset can be improved for future detailed velocity structure investigation in central Taiwan. |
