聯合多重資料反演地殼-上部地函地震速度構造之成像：以越南及台灣為例

Abstract

地震層析成像技術已被廣泛應用於各地不同的構造環境並實證為一能有效擷取地球內部訊息之工具。我們對複雜活動構造的了解往往取決於我們能夠解析出多寬廣、精確的構造成像。因此，擴展並提升地震層析成像的解析力一直是很重要的一項工作。本論文即試圖利用聯合反演多重資料(包括多種地震波波相與地質工程資料)來提升對地殼-上部地函速度構造之成像能力，進而幫助越南與台灣地區做更精細的構造判讀與解釋。 論文通篇主要由三個研究計畫所組成。在第一個計畫中，利用近期在北越所佈置的寬頻測站陣列與越南當地測站網所接收到之地殼Pg與上部地函Pn波資料，我們求得該區從過去以來的第一個三維P波速度模型，及其莫荷面深度變化。反演之速度模型大致與地質岩性和主要斷層構造相吻合，綜合P波速度與莫荷面形貌之結果也顯示紅河剪切帶可能為一岩石圈尺度的斷層至少切穿地殼至上部地函，上部地函沿著剪切帶分布的低速異常亦暗示著相關的熱活動異常。整體而言，北越地區因為長期的熱構造活動歷史有著相對弱的地殼與熱地函的特徵，可能有過下地殼剝離作用(Delamination)的發生。 第二個計畫是利用高密度測站網與地震和地質資料的聯合演算來解析台灣造山帶之速度結構。雖然台灣過去許多的地震層析成像研究已有相當豐碩的成果，但這些研究大多以反演P波速度(Vp)模型，或同時解算P波與S波之速度比值(Vp/Vs)為主，對於S波速度(Vs)模型則較少著墨。然而，為更加了解複雜的造山構造，在推論時同時交相驗證比對三種模型是至關重要的。但是不同研究中的模型因反演參數、模型建置、及資料多寡分布都不盡相同的情形下，難以做細部的參考對照。因此，我們期望藉由特別設計的聯合反演方法來求取彼此在解析度與一致性上都相匹配地Vp、Vs與Vp/Vs模型。在此方法中，我們也利用了遍佈全台的445處井測資料來加以約束一般傳統上地震資料所無法解析地近地表的速度構造。反演之模型能夠表現出近地表平原區真實的極低速值；透過同時參照Vp、Vs與Vp/Vs的模型也提供了豐富的構造資訊。藉此，我們能夠清楚的描繪出歐亞大陸岩石圈與隱沒的菲律賓海板塊西緣間接觸面的西傾形貌，以及南向演育的海岸山脈與弧前陸塊之進程。為了更進一步向下擴展我們模型的解析力來探討隱沒板塊間的幾何形貌，在第三個計畫裡我們接著量測並結合了遠震的資料。並且採取兩階段式的反演：先利用近震求取精確的三維本土模型(藉此修正複雜地殼構造對遠震走時的影響)，再以此三維模型為初始模型來進行非線性的近、遠震聯合反演(疊代進行波線追跡且不固定模型底層入射點)。結果顯示位於台灣中北部向東隱沒的歐亞大陸板塊能向下追溯至深度約200公里，並呈現複雜的形貌，在約北緯23.2度處有一西北東南走向的撓曲現像。根據層析影像我們亦對台灣地區的板塊幾何進行了三維的描繪。 我們最後彙整與討論了各個計劃的成果，以及其未來工作的展望。另外，也將兩個在博士期間完成的子計畫一併簡介及附錄於後：其一是對甲仙地震之主餘震序列進行分析與構造探討，二是利用地震與應力之空間分析來建立北台灣地區之構造運動模型。

Seismic tomography technique has been widely applied to different tectonic environments around the world and proved as a powerful tool to retrieving the information of Earth’s interior. The understanding of complex tectonic processes largely relies on how extensive and accurate structural imaging we can resolve, and is therefore of primary importance to proceed. In this dissertation, we attempted to improve our imaging ability in a scale of crust-to-upper mantle by the joint inversion with multiple datasets, including various phases of seismic waves and geotechnical data, to better clarify the tectonic interpretations for the Vietnam and the Taiwan region. The dissertation mainly consists of three projects. In the first, with a newly deployed portable broadband array and the local seismic network in northern Vietnam, we used the crustal P-wave (Pg) and uppermost-mantle head-wave (Pn) data to obtain the first local 3-D P-wave velocity model and the related Moho depth variation. Our results show a good correlation with the surface geology and major structures, and reveal that the RRSZ is likely a lithospheric-scale structure penetrating to the uppermost mantle with mantle thermal anomalies. The northern Vietnam in general appears to possess a weak crust and hot upper mantle with a long and complex thermo-tectonic history probably induced by the past delamination. In the second project, we studied the velocity structures of Taiwan orogen by jointly using the seismological and geotechnical data with a highly dense seismic network. Although the previous tomographic achievements in Taiwan region have been fruitful, most of them were P-wave velocities (Vp), or with Vp/Vs ratio jointly, but few for S-wave velocities (Vs). However, to better unravel the tectonic complexity of Taiwan, interpreting with all three indicators could be crucial. But models from different studies are hard to compare and verify to each other due to different datasets and inversion settings. Therefore, we attempted to provide a new set of Vp, Vs, and Vp/Vs models in comparable resolution and internal consistency by an elaborate joint-inversion scheme; in which a special constraint by the borehole logging data was also imposed on the near-surface part of model where is usually poor-resolved by ordinary inversion. Derived models show tremendous changes of shallow velocities than previous studies and copious information with mutual verifications of Vp, Vs, and Vp/Vs, rendering us to clearly delineate the west-dipping interaction between the Eurasian lithosphere and the subducting Philippine Sea plate, and the southward evolution of the Coastal Range and forearc basement. To further expand the model resolution to the deeper depth to explore the slab geometry beneath Taiwan region (especially for north of 23°N latitude), in the third project we measured and combined the teleseismic data into tomographic inversion. Rather than the direct joint inversion, we adopted a two-step strategy to invert an accurate local model first by local data only, and then implement a nonlinear joint inversion with the teleseismic data together (without fixing the ray-incident points at the model bottom). Results show that the deep extension of the eastward subducting Eurasian plate can be retrieved readily to ca. 200 km deep with a plausible slab deflection around the latitude 23.2°N. We then constructed a 3-D schematic model accordingly for Taiwan region. We ultimately summarized the achievements for respective projects, and their prospections for future works. In addition, two side projects done in my Ph.D period are also briefly introduced and supplemented: One is studying the out-of-ordinary Jiasian, Taiwan earthquake (Mw=6.3) for its tectonic implications, and the other is using the spatial seismicity and stress pattern to explore and construct the northern-Taiwan kinematic model.