|Abstract: ||震源、路徑和場址這三個因子影響地震波。場址特性是對於強地動預測最重要的影響因子。在研究中，我們所利用隨機式點震源模擬法(Boore, 2005)，這種方法已經被廣泛地應用於地動預測和建立模型，其控制觀察地動的參數（Atkinson等人，2009）。|
本研究利用隨機式點震源模擬法(Boore, 1983; Boore, 2003)，模擬台灣地區強度動觀測計畫(Taiwan Strong Motion Instrumentation Program, TSMIP)自1992到2012年間的淺源地震紀錄。選用地震深度小於30km和震矩規模(M_w) 4~6.5之間的地震事件。本研究區域位於台灣北部宜蘭地區，共使用70個TSMIP測站。根據Vs30做分析可分做B,C,D和E場址。本研究隨機式方法使用的震源參數選用前人研究(Sokolov et al., 2006; 2009)中所設定參數。地表放大參數選用半空間。利用雙站法(Borcherdt, 1970)計算觀測與模擬紀錄之間的差異，建立宜蘭地區強地動測站0.2Hz~10Hz的經驗地層轉換函數(Empirical Transfer Function)。地動預測是利用數個目標事件通過選擇隨機點源模擬到半空間來預測。峰值加速度（Peak Ground Acceleration, PGA）的預測是做場址修正後再使用經驗轉換函數做預測。因此，模擬地面運動進行比較時間域和頻率域（Degree of spectrum difference, DSPD）來顯示良好的模擬情況。最後，PGA預測的結果與使用衰減公式預測相比。我們發現模擬結果明顯略勝一籌。
關鍵字:隨機式模擬、場址修正、經驗轉換函數;Seismic waveform is controlled by three factors – source properties, path characteristics, and local site effects. The local site effect is the important factor participate strong ground motion prediction. In this study, we used stochastic point-source method for simulating ground motion (Boore, 2005). This method has been widely used in the development of ground-motion prediction equation and in modeling the parameters that controls observed ground motion (Atkinson et al., 2009).
The shallow earthquake events which recorded by Taiwan Strong Motion Instrumentation Program (TSMIP) from 1992 to 2012 are simulated with the stochastic point-source method (Boore, 1983; Boore, 2003). The earthquakes are selected with the depth from 0 to 30 km and the magnitude (Mw) from 4 to 6.5. The study area is situated in Ilan area which is located in the northeastern Taiwan. There are 70 TSMIP stations which based on the Vs30 consist of site classes B, C, D, and E. Seismic parameters for stochastic method were selected based on previous studies (Sokolov et al., 2006; 2009). The crustal amplification parameter is set to the half space.
The empirical transfer functions from 0.2 Hz to 10 Hz for each station in Ilan area were calculated by H/H method between observed and simulated spectra (Borcherdt, 1970). Ground motion prediction is calculated by selecting several target events for stochastic point-source simulating to the half space. The prediction of peak ground acceleration (PGA) is estimated after doing the site correction with the empirical transfer function. Then, the simulated ground motion was compared in time domain (PGA) and frequency domain (Degree of spectrum difference, DSPD) to show the goodness of the simulation. Finally, the results of PGA prediction was compared with attenuation relationship. Comparison of our simulation results to attenuation relationship, we found that our simulation results showed slightly better.
Keywords : Stochastic point-source method, Site effect, Empirical transfer function