反射探勘數據重建及其淺層地質應用

Abstract

反射信號（reflection data）數據處理一直是地球物理探勘學家相當有興趣的領域。地球物理反射探勘法是以高解析度的反射數據剖面來解釋地質構造或地下埋藏物之幾何形貌。而野外數據常因伴隨的相參性雜波（coherent noise），如：地滾波（ground roll）、諧和波（harmonic wave）、多重反射（multiple），以及儀器本身和外界背景雜波之干擾；或反射信號能量在傳遞過程中，會有本質性 (intrinsic) 及幾何性的衰減（damping factor），導致記錄之數據信雜比（signal-noise ratio, SNR）不佳，進而增加了原始信號本質之分析及解釋上的難度。儘管有很多傳統處理方法可以改善，但往往成效不佳，甚或衍生出更麻煩的人工假象 (artifacts)。 頻譜分析是資料處理的基本方法，本論文以近代新的可適性（adaptive）瞬間時頻分析法，希爾伯特-黃轉換（Hilbert-Huang Transform, HHT）分析技術為架構，以非線性（nonlinear）及非穩態性（non-stationary）的信號處理概念，將數據先用自然對數轉換（natural logarithmic transform, NLT）提高信雜比以及調整數據的動態區間（dynamic range），再配合均和式經驗模組拆解法（ensemble empirical mode decomposition, EEMD），讓原始記錄之反射信號與雜波得以分離，並解析出信號本質上所具有的瞬時特性（instantaneous attributes）之物理意義，進而重構數據，提升信號解析度，得到非線性濾波之功效。 在研究程序上，本論文先對反射探勘法從理論、資料處理到解釋，做簡單而不失周延的整理與探討，再以模擬之數據信號來測試及佐證分析方法之可行性，最後再應用於實際的數據。野外案例主要包含兩部分：第一部分為反射震測資料，以台灣西南外海之海底震測儀（ocean bottom seismograph, OBS）數據為例；第二部分屬透地雷達數據，以校園道路之走離雜波測試及新竹峨眉台三線邊坡之傾斜地層構造等三個野外案例說明。 以數據模擬結果與野外案例顯示，NLT及EEMD共構進行分析，不僅對原始數據的信雜比明顯提升，也助於增強走時曲線在時間域與空間域上能量的連續性與相參性，並提高信號的解析度及時間（深度）顯示。另外，利用可適性瞬間時頻譜分析所得之結果可瞭解，應用NLT及EEMD組合處理的反射數據，除了可有效增進反射數據之可信度外，也讓信號數據處理分析多了一種新的選擇。Exploration geophysicists have long been interested in the processing of reflection data. This dissertation proposed a nonlinear, adaptive procedure to enhance the signal/noise (S/N) ratio of reflection data. The processing methodology is based on the logarithmic transform in conjunction with a newly developed nonlinear data analysis method, the ensemble empirical mode decomposition (EEMD). I use a synthetic model to investigate the capability of signal reconstruction from the decomposed data, and compare the results with those derived from other 2D adaptive filters. Examining the Hilbert-Huang transform (HHT) spectrogram, it indicates that the data attenuation problem is significantly alleviated and the decomposition sensitivity of the EEMD method is greatly improved with the aid of the logarithmic transform. To validate the method, real reflection data include field record of the ocean bottom seismograph (OBS) data and ground penetrating radar (GPR) data are exploited as examples. The OBS data are the record of one of the National Taiwan Ocean University OBSs deployed offshore southwestern Taiwan where is an area very likely deposited with gas hydrate. The GPR records comprise a shot gather acquired on bituminous pavements and a common offset section obtained at a site of dipping layers in Chu-Dong, northern Taiwan. All the real data examples are processed by the similar procedure, and demonstrate the robustness of this method. Therefore, it is can be concluded that instead of Fourier-based approaches, the reflection data can be effectively processed by using an alternative nonlinear adaptive data analysis method. This new method can extract the signal components from noisy data successfully. The achievement of this study suggests a possible nonlinear analysis application in future geophysical data processing.