台灣宜蘭平原使用背景噪聲波場求取淺部剪力波速度構造: 使用相位交相關、模擬和反演的研究;Ambient Wavefield Studies by Phase Cross-Correlation, Simulation and Inversion of Ilan Plain, Taiwan

NCU Institutional Repository > 地球科學學院 > 地球物理研究所 > 博碩士論文 > Item 987654321/92352

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/92352

題名:	台灣宜蘭平原使用背景噪聲波場求取淺部剪力波速度構造: 使用相位交相關、模擬和反演的研究;Ambient Wavefield Studies by Phase Cross-Correlation, Simulation and Inversion of Ilan Plain, Taiwan
作者:	費安妲;Fitrianditha, Fanda
貢獻者:	地球科學學系
關鍵詞:	環境雜訊;相位交叉相關;相位加權疊加;雜訊波場模擬;頻散分析;淺層地表構造;表面波反演;剪力波層析成像;Ambient Noise;Phase Cross Correlation;Phase-weighted Stack;Ambient Wavefield Simulation;Near-surface Velocity Structure;Dispersion Analysis;Surface-wave Inversion;Tomography
日期:	2023-07-24
上傳時間:	2024-09-19 15:47:13 (UTC+8)
出版者:	國立中央大學
摘要:	宜蘭平原是沖繩海槽最西端開裂的產物。地殼構造上的活動影響了宜蘭平原內盆地的沉積盆地形貌，形成了一個三角形狀的盆地。蘭陽溪將該盆地分為北側和南側兩部分。ILAN2014 的主(兩天)、被動式(約五個月)大型震測實驗計畫於2014 年8 月14 日至2015 年 1 月14 日期間執行。在宜蘭平原共部署了163 個臨時測站，以覆蓋宜蘭平原的整個區域，站間距從1 公里到4 公里不等。在建立所有兩測站間的經驗格林函數(Empirical Green′s Functions，EGFs) 的過程時，資料處理的第一步驟先將兩個測站間的長時記錄先利用相位交互相關 (Phase Cross Correlation，PCC) 技術提取可能包含地震事件的背景噪聲波形資料。測站間的相位相關信號的提取特別強調以瞬時相位資訊為主要的檢測機制，PCC 對波形的相位相似性很敏感，即使對於小的地震或背景噪聲事件都能偵測得到，但此種保有原振幅的資料處理方法不會因含有強振幅度事件而造成對相位相似的弱信號的提取產生偏差。故而可以提取含極多地震或背景噪聲的事件訊號。此技術可簡稱為「保幅但因對振幅大小影響不敏感而不產生提取偏差」的特殊處理手段。兩站間的長時PCC 時間記錄，再透過時間尺度相位加權疊加（ts‐PWS）演算法進行資料的進一步疊加，由於 Ts‐PWS 在檢測幅度和相位方面具有不受時變影響的特性(近似平移不變性)，且計算成本較低，因此 Ts‐PWS 可以強化對真實信號提供更穩定的測量與提取機制。其目的為提升訊號雜訊比以進行信號增強與壓制低相關係數的訊號與波形。以一小時時段為主的PCC 波道資料在經加權疊加後即可形成兩測站間的單一波道的EGF。將以某測站為主並視為虛擬波源 (Virtual Source, VS) 對所有其他測站(此情況下則為虛擬接收測站;Virtual Receiver, VR)處理得到的EGF 按兩站的站間距以聚集方式排列形成EGF 道集(trace gather)或稱為疊後相關係數時間道集(correlogram gather) 。以每個 VS 測站位置為中心所形成的相關道集波形資料可瞭解由VS 到VR 間的背景噪聲波傳現象。不同的VS 測站的疊後相關係數時間道集可顯示出其個別波傳特性、且具有個別明顯且獨特的波形，相位變化，到時與正常时間遷移(Normal Time Moveout, 时間動態遷移)特徵。環境噪聲數據通常歸因於一次或二次微震的產生機制。主要微震是海浪與海底之間作用力的直接耦合效應所造成。二次微震是海浪一旦撞擊海岸，反射回海洋並與另一波傳入的海浪相互作用，然後產生第二次反射波再次撞擊海岸。 ILAN2014 觀測計畫共部署有 163 名接收站，按資料與地質特性我將所有資料分為五組分別進行分析。透過許多數值模型設置進行的許多模擬和分析，根據蘇柏立等人(2019) 所提出的模型。我針對沿海岸線進行了三種不同的環境噪聲激發源情境模式模擬，在該場景下以分析主要噪聲源的可能傳播方向。噪聲波場情境的模擬主要集中於使用譜元法(SEM)探討沿著不規則且高度變化的海岸線幾何形狀和海‐陸界面發生的聲‐彈性震波傳遞耦合現象。使用構建的合成 PCC 波道資料擬合真實數據的EGF 相關道集波形資料的到時與因果或非因果时間動態遷移特徵，分析與模擬結果顯示噪聲源的產生機制主要是以由北向南方向的二次微震聲源波傳特徵為主。另外，我還進行了另一項利用有限差分法 (FDM) 進行陸上環境波場模擬的嘗試。通過許多系統化的受控模擬案例測試，來瞭解噪聲源被激發的可能影響因素。可能影響因素包括激活源數量、源持續時間、頻率源分佈、可能的噪聲源有不同的集中深度以及可能的噪聲源被限制在特定深度或受高速度梯度區的影響等以匹配真實數據特徵。針對沿著蘭陽溪河岸沿線所佈設的測線線型陣列收集到的真實EGF 相關道集波形資料數據，論文中討論了進行不同合成的PCC 波道資料情境模擬結果與其可能的生成過程，研究結果顯示特定環境噪聲源集中在深度3.8‐4.1 km 的高速度梯度變化區內可擬合真實資料的特徵。論文中我還針對群速度和相速度進行頻散分析。由群速度和相速度分佈圖顯示沿著海岸線的接收站的傳播速度比宜蘭平原內部的接收站相對較低。其他地區與海岸線地區相比，表面波傳播速度相對較高，特別是位於宜蘭平原邊緣的站點，由於地形造成的高程變化和地質條件的改變，群速度和相速度的側向變化更為明顯。表面波反演是利用選取的頻散曲線數據，主要是使用速度‐週期數據對來進行反演。環境噪聲層析成像反演主要是基於環境噪聲的頻散特性資訊，其中涵蓋了拾取速度、週期、振幅及其相關的不確定性資訊等。為了反演剪切波速度結構，Fang 等人(2015)提出了基於以小波為基礎的的稀疏資料制約演算法進行淺部剪力波速構造的成像。我使用了三種不同的一維初始速度剖面模型，其中包括使用了自行定義的線性增加Vs 波速度模型、黃有志(Huang, 2003)的平均一維速度與將黃有志(Huang, 2003)和根據蘇柏立等人(2019) 的所提出的大尺度數值模型來建立組合速度。進行反演時，第一種和第二種一維初始速度剖面模型，反演過程僅允許最多十次疊代運算。為了保證能夠得到穩定的解與較深的偵測深度，第三種初始速度剖面在執行逆推時將反演疊代運算次數設置為35 次。第二和第三種初始速度剖面模型都對深度的測深及剪力波速分佈估計提供了相當有用的建議。總體而言，宜蘭平原北部地區速度相對南區為低、盆地位置偏北、沉積層厚、受蘭陽溪控制與貢獻較多。宜蘭平原南部地區速度高、側向速度變化較劇烈、剪力波的波速變化範圍廣、高與低速區相對明顯。關於盆地深度的估計可以達到1.2‐1.3 km、1.7 km 甚至可以檢測到高達2.0 km 或~2.5 km 的深度，這些建議的結論需要緊密的根據Vs 值定義與分佈與速度結構的特徵來進行約束。許多與潛在的地熱異常區與Vs 波速的分佈特徵極為密切相關，透過逆推所得的Vs 波速影像可直接以視覺檢測來尋找較低的Vs 區塊。檢測時可在每個水平速度切片中集中注意嵌入在不同深度的高Vs 速度背景中的極低或低得多的Vs 速度特徵。此剪力波的低速帶(LVZ 區塊)在南部較深處有相當大片區域的分佈。儘管某些測站可能無法提供足夠的探測深度，解決方案則需更多的測站資料與積極小心的態度進行資料收集、處理與逆推解釋。目前可檢測到的地熱異常可以直接從宜蘭平原的Vs 速度分佈圖像中觀察到。剪力波層析成像的手段提供相當正面的結果，並且顯示出相當明顯的特徵。確切的位置及其解釋可能需要以後積極的且儘可能進行更仔細的研究和驗證。;Ilan Plain is a product of the opening of the westmost Okinawa trough. The opening activity also influences the basin shape of Ilan Basin and Ilan Plain that created a triangular-shaped basin. Lanyang River divided the basin into northern and southern parts. ILAN2014 experiment utilized total of 163 temporary stations deployed in Ilan Plain during Aug. 14, 2014 to Jan. 14, 2015, with station spacing varying from 1 to 4 km, to cover the whole study area. The empirical Green’s functions (EGFs) of all available virtual source (V.S) to other receiver pairs were performed via Phase Cross Correlation (PCC) technique before stack. The instantaneous phase information of coherent signals were detected between stations. PCC is sensitive to waveform similarity even for small events but less sensitive to strong amplitude features (amplitude unbiased). Then, the PCC results are further stacked by time-scale Phase Weighted Stacking (ts-PWS) strategy for signal enhancement. Ts-PWS gives more stable measurements for real signals due to its near shift- or time-invariant property on detecting magnitude and phase with less computational cost. Stacked correlogram gathers show obvious and distinct features at each V.S location. Ambient noise data is generally attributed to primary or secondary microseisms. The primary microseism is the direct coupling between ocean waves and seafloor. The secondary microseism is the ocean waves once hits the coast, reflect back to the ocean and interacts with another incoming ocean waves and then generate second reflected wave which strike the coast again. A total of 163 receivers of ILAN2014 observations were divided into five groups. Many simulations and analyses through many model setup including the model proposed by Su et al, 2019 were used. Three different source excitation scenario along the coastline were performed to analyze the possible propagation direction of the predominantly noise sources. Simulations focused mainly on the coupled acoustic-elastic wave propagation phenomena occurred along the irregular and highly varying sea-land coastline geometry and interface using Spectral Element Method (SEM). The north to south wave propagation simulations with constructed synthetic PCC gather resemble the real data gather move-out and features on the causal/acausal arrivals. Another attempt on ambient wavefeld simulations onshore utilized Finite Difference Method (FDM) were also performed. A systematic noise source excitation influence factors were tested through many controlled simulations cases. The influence of number of activated sources, source duration, frequency source distribution, possible concentration at various depths, and possible source of noises being confined in a specific depth to match real data feature. A specific source at depth 3.8-4.1 km resemble the real data phase cross-correlation gather along the Lanyang River array were studied and discussed its possible generation processes. The dispersion analysis on both group and phase velocity were also performed. Both group and phase velocity distribution maps show that receiver stations along the coastline have relatively lower propagation velocities than receivers in the inner part of the Ilan Plain. Other areas show relatively higher surface wave propagation velocity when compared with the coastline area, especially for stations locate at the edge of Ilan Plain with more obvious elevation changes created by the elevated topography and geological site conditions. The last step is to use the picked dispersion curve data information mainly on velocity-period data pair for the inversion. Ambient noise tomographic inversion is based on dispersion properties cover the picked velocity, period, amplitude and its associate uncertainty information were used. In order to invert shear wave velocity structure, the wavelet-based sparsity-constrained method by Fang et al. (2015) is used. Three different initial model including 1D linearly increased velocity model, average 1D velocity from Huang (2003) and combined velocity of Huang (2003) and Su et al. (2019) were used to performed the inversion. The first and second inversion approaches only allow maximum of ten iteration. To ensure stable solution can be obtained, the third approach set the inversion iterations to reach 35 iteration. Both second and third approaches provide useful suggestions on the estimated Vs distribution at various depth. Overall, the estimated velocity structure features regarding basin depth can reach 1.2-1.3 km, 1.7 km or even deeper and up to 2.0 km or ~2.5 km can be detected depends on the constrains defined by the Vs values. Many Vs distribution features which are strongly related to geothermal anomaly can be visually detectable by the lower Vs or even extremely or much lower Vs velocity features imbedded in the high Vs velocity background at various depths can be visually detected. A fairly large distribution of LVZ zone in the southern. Ilan Plain can be inspected directly from shear wave velocity distribution images. Although some station may not provide enough depth solution, however detectable geothermal anomalies are very positive and show fairly distinct features. The exact location and its interpretations may require more careful study and verification later whenever it is possible.
顯示於類別:	[地球物理研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	27	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....