| 摘要: | 本研究使用資料為TAIGER T6測線，係為台灣2008年的台灣大地動力學國際合作研究計畫(TAIGER)所進行之地殼尺度寬角度折射/反射(WARR)探勘測線之一。北部陣列由西向東跨越北台灣並在其上有四個炸點(N1, N2, N3, N4) ，共配置有456個平均間距為200公尺的受波器。本測線幾何屬於蜿蜒測線，各炸點能量並未一致，各區域躁訊比也不盡相同。其中由於不規律的炸點及測站分布、強烈的側向速度變化及高程變化，使得資料中有明顯的三維波傳現象，因此常用的標準震測資料處理流程並不能完整的施行並滿足研究需求。本研究透過資料修剪、增益處理、解迴旋處理、濾波處理及區域傾斜疊加變換(LSST)，對資料狀況做初步改善。
為了幫助理解逆推的行為及特性，本研究首先使用三種地殼尺度之模型進行模擬: (1) 表面無高程起伏之層狀模型，(2) 表面有高程起伏之層狀模型， (3) 含有高程起伏及側向速度變化之模型。我們以Specfem2D軟體進行模擬運算，並分析了炸點、測站高程效應，與速度變化所造成的影響。結果顯示當逆推過程考慮高程效應時，無論資料做靜態處理與否，均會導致逆推結果之速度較不平滑。我們接著使用全支距及部分支距的實際資料進行波形逆推，從四個炸點的震測資料獲得了七個一維速度模型，並以此產生類二維的速度模型(quasi-2D velocity)。透過Kirchhoff Wave Equation Datuming (KWED)對資料進行靜態修正，處理後的資料對於三種不同的模型有較佳的逆推結果。隨後我們以波形模擬及走時殘差分析對速度模型進行檢驗，其結果與實際資料之誤差非常小。接著我們嘗試對波形逆推所得之模型進行了走時逆推(travel time inversion)，以將類二維模型轉為二維模型，並提升解析度。結果顯示其確實有助於增進近地表區域及位於雪山、中央山脈下高速異常區之解析度。走時殘差分佈亦顯示誤差大幅減少，N1至N3炸點資料之誤差由 ±0.5秒變為 ±0.2秒，N4炸點則由 ±1.5秒變為 ±0.5秒。在與參考研究之模型比較及驗證後，本研究所主張之模型於淺部構造有較高之精確度及解析度。
根據逆推之速度模型及深度偏移成像，我們可以看到西部平原沉積盆地存在厚度接近三公里，具有2.6至4公里/秒相對低速的區域，並由5.0公里/秒至5.5公里/秒之區間勾勒出一向東尖滅之速度構造；此外由偏移成像亦可發現在淺層數公里內也存在有明顯構造邊界，此構造或與低綠片岩相變質基岩有關。此現象可能代表同生裂谷沉積物及基盤於抬升過程中剝露而出。由速度模型5.0公里/秒至5.5公里/秒之區間則可發現一向東尖滅速度構造；由偏移成像亦可發現在淺層數公里內存在有明顯構造邊界，與低綠片岩相變質基岩有關。此現象可能代表同生斷陷沉積物及基盤於抬升過程中剝露而出。位於雪山山脈下的高速異常區與北台灣之電性構造有關。位於深度20公里，側向距離61公里至85公里處，則有Datong Bright Spot (DBS)反射面，可能與超鐵鎂質侵入岩體有關。位於中央山脈下的高速異常區與由超基性岩及大理岩組成的大南澳片岩露頭有關。此外，受限於資料品質及陣列孔徑的不足，莫荷不連續面並沒有在結果中發現。
;TAIGER T6 transects is one of the four crustal-scale Wide Angle Refraction/Reflection (WARR) survey lines under TAiwan Integrated GEodynamics Research (TAIGER) project conducted in 2008 in Taiwan. The north main array consist of four shot points (N1, N2, N3, and N4) across northern Taiwan from west to east. A total of 456 geophones deployed with receiver interval of ~200m on average. However, the recorded data suffers from crooked survey line geometry, uneven shot energy, noise level. Near-surface elevation effects caused by the irregular source and receiver spacing, strong lateral velocity variations and rapid topography changes produce apparent 3D wave propagation so that the standard seismic data processing could not be fully applied. To compensate above-mentioned drawbacks, pre-conditioning of the data including trace editing and enhancements, deconvolution, frequency filtering, and Localized Slant Stack Transformation (LSST) was applied.
To study the inversion behaviors and features, feasibility studies of three crustal scale synthetic models containing: (1) flat layers without topography, (2) flat layers with topographic, (3) laterally varying topographic velocity models. Synthetic data is produced from Specfem2D computations. Effects of source and/or receiver elevation, velocity variation, and the combination of all were studied. Wavefield inversion is robust and stable for all test models. Inverted 1D velocity becomes less smooth when elevation effects are included with and without elevation corrections. Implementation of full-offset and offset-dependent wavefield inversion approach on the real data able to provide reasonable quasi-2D velocity models constructed based on seven 1D velocity profiles from four shots. Together with the inversion results from elevation corrected data through Kirchhoff Wave Equation Datuming (KWED), an adjusted model can be extracted to accommodate the best features among three different models. Quality check of the velocity models through wave simulations and travel time residual distributions show good fit and minimum error compared to the real data. Furthermore, travel time inversion were implemented based on the derived model from wavefield inversion to convert quasi-2D to pure 2D model as well as improve the resolution. Travel time inversion able to improve the resolution related to the near surface of the model as well as the high velocity anomaly below the Hsuehshan Range and Central Range. Also, travel time residual distributions shows that, the travel time inversion significantly improve the error, which range from ±0.5s to ±0.2s for N1-N3 and ±1.5s to ±0.5s for N4. Comparisons and validations between proposed model and several reference model shows improvement at shallow structure shallow part with sufficient accuracy and resolution through limited datasets and consistencies on the large scale.
Based on proposed velocity model and depth migrated image, several structure feature can be outlined. Existing sedimentary basin with layer thickness of ~3 km and relative low velocity from ~2.6 to 4 km/s in the Western Foothill. Thinning eastward feature judging from seismic velocity between 5.0 km/s – 5.5 km/s and highlighted interfaces from migrated image in the upper part within few kilometers presumably can be linked to the low-grade greenschist facies metamorphic basement. Possible indication of their synrift sediments and basement are being uplifted and exhumed. High velocity anomaly linked to the Northern Taiwan Conductor below Hsuehshan Range. Datong Bright Spot (DBS) reflector at ~20 km depth within ~61Km to 85Km of distance, can be linked to ultramafic intrusion. High velocity anomaly below Central Range represent the Tanano Schist outcrop presumably consisted of ultramafic rocks and marble. Due to the limitation of data aperture, Moho discontinuity is not resolved. |
