| 摘要: | 屏東平原為三面環山之沖積平原，平原表面覆蓋著數百公尺之河相沖積物，主要由嶺口礫石、古期沖積扇、現代沖積扇所組成。由於地震波在穿過鬆軟的淺層沉積物後會產生歷時延長及振幅值放大的現象，而屏東平原東界緊鄰長達89公里之活動斷層，因此了解該地之場址特性為相當重要之課題。過去於屏東平原已有許多速度構造調查成果(微地動陣列、接收函數及強震站鑽探)，但由於測點過於稀少且分布不均，故無法充分反映側向之速度變化。目前屏東平原內已進行308點之單站微地動量測，站間距離平均為兩公里，本研究將利用此密集的微地動資料進行速度構造之推估。結合前人所提供之淺層速度構造作為初始模型，我們使用基因赫氏法，藉由赫氏法計算出理論單站頻譜比並配合基因演算法擬合觀測之單站頻譜比，搜尋出最佳之一維速度構造。最後將所得之密集的一維速度構造與地質鑽探紀錄比對，嘗試建立屏東平原地下構造之三維形貌。此反應盆地內鬆軟沖積層放大特性之速度構造於未來將可應用在地動模擬等方面，以期降低地震災害所帶來之損失。;Pingtung plain in the south of Taiwan is covered with hundreds of meters thick of soft sedimentary strata. These unconsolidated alluvial sand and gravel are mainly from the Central Range and Yushan Range. Chiang (1971) indicated that the plain sediments thick gradually from east to west according to the data from seismic surveys, deep boreholes, and surface geology. In this study, the slope of strata in the Pingtung Plain is neglected to simplify the calculation of horizontal-to-vertical spectral ratio (HVSR). Seismic wave will be amplified after propagating through these strata and leads to serious disaster. For Taiwan, a country located on an active plate boundary, it’s essential to clarify the site effect there. To know the site response, correct and detailed shallow velocity structure is necessary. Although there are several studies provided some velocity structures, their scales are all too large to provide enough information. In this study, HVSR of 308 sites were simulated to inverse the shallow velocity structure of Pingtung Plain in detail. In our simulation, the Haskell method had been used to calculate the theoretical HVSR of corresponding S wave velocity structure. To define the searching boundary and initial model, we refer to several previous studies. The S wave velocity profile, measured by Suspension P-S Logging Technique, from Engineering Geological Database for TSMIP. Chen (2016), Chiang (2011) used microtremor array to derived the shallow velocity structure. Lin et al. (2017) used receiver function to inverse the velocity structure of the TSMIP seismometer sites. A model with seven to nine layers covering a half space was set up for the Genetic Algorithm search (GA search). There are two unknown parameters for each layer, i.e., S-wave velocity and thickness. In order to reach the best solution, there are totally 1,250,000 repetitions for one site forward modeling. Finally, the best fitness velocity model will be picked selected and saved. By simulating the 308 HVSRs, we constructed the Pingtung plain near surface shear-wave velocity structure which useful for strong ground motion simulation and prediction in the future. |
