場址效應為地振動預估中重要的參數之一。現今地震工程界普遍使用淺層構造的剪力波波速（Vs30及Z1.0等係數）用以估計場址效應，然而淺層剪力波波速未能完全地反應實際場址之地盤特性，土壤複雜的非線性放大反應，影響地振動預估及模擬的準確性甚鉅。本研究利用氣象局新一代的井下地震監測網的資料，選用2013年三個中大尺度規模(ML6.2-6.5)的地震，計算地表及井下測站的最大加速度(PGA)、最大速度(PGV)之放大倍率，並利用傅式頻譜比得到不同頻率下之放大倍率，分析井下至地表淺層的放大特性，更進一步比較放大倍率和測站Vs30值之間的關係，並使用兩種解迴旋方法進行地層轉換函數地計算。結果顯示PGA、PGV放大倍率與Vs30值的相關性並不高，然而頻譜比的變化情形和測站Vs30值的相關性較高。依頻譜比隨著頻率的變化趨勢可以將測站區分為兩類，(I)類測站之放大倍率隨著頻率的增加而遞增，而(II)類測站在1-3 Hz有較大的放大倍率，推測與鬆軟的沉積層造成特定主頻放大有關。而地層轉換函數已經由兩種不同解迴旋方法計算完成，兩方法皆可很好的重建地表站訊號，未來可望將此轉換函數應用於強地動模擬，作為從井下至地表的淺層場址效應估算模型。;Site effect is one of the most important factors in evaluating earthquake ground motion. For engineering usage, the ground motion prediction equation (GMPE) usually includes the site-effect factor (e.g. Vs30, Z1.0). The site effect, however, is considered as a non-linear response and could be critical for the aspect in broadband waveform modeling or simulation for ground motion prediction. In this study, I utilize three large earthquakes (ML6.2-6.5) data of both surface and borehole seismometers from Central Weather Bureau (CWB) next generation seismic network to analyze the site amplification factors and the transfer functions of sediments over the basement. A spectral ratio method is used to acquire empirical amplification factors (EAF) for PGA, PGV, and frequency spectra from 0.1 to 10 Hz. The relationship between the EAF and Vs30 value is analyed as well. The results show that the corrections between PGA (or PGV) amplification factors and Vs30 are not significant, suggesting that the amplification of PGA or PGV from borehole to surface was not dominated by shallow structures (i.e., top 30 m of strata). On the contrary, the relationships between EAF for frequency range of 0.1 to 10 Hz and Vs30 are strongly correlated and can be classified into two types. For Type I sites, the EAF increased with increasing frequency, and the Vs30 of these sites are generally over 760 (m/s). Type II sites, with Vs30 generally lower than 760 (m/s), show the peak value of EAF around a frequency range of 1 to 3 Hz that may attribute to the soft-soil conditions. The transfer functions obtained by two deconvolution methods well reconstruct the seismograms of surface stations, which could be used to simulate waveform as site-effect models. The empirical amplification and transfer function analysis could be applied to obtain the GMPE for basement rock and also be adopted in ground motion simulation.