本研究主要針對位於蘭陽平原之中央氣象局強震站，利用隨機式震源模型模擬沖積層站不考慮場址效應下之地動加速度，該模擬值可視為在堅硬岩盤上所收到的資料，藉由理論模擬值與觀測值之間的差異來探討場址效應，並與單站頻譜比法、雙站頻譜比法之結果做比較，討論隨機式震源模型應用在場址效應上的可信度。其中，挑選供模擬用之地震事件有1999年10月13日規模5.0、2000年9月10日規模6.2及2002年5月29日規模6.2三者，並利用數個岩盤站所收到的資料來進行模型參數的校準。比較觀測值除以模擬值之頻譜比與單站頻譜比所得之蘭陽平原主頻等值圖，兩者放大主頻的分佈趨勢大致相同，位於平原邊緣之西側河谷、三星鄉及蘇澳鎮的主頻較高，約3~4Hz以上；主頻較低處分佈於平原中心至海邊，約2Hz以下。結果發現，蘭陽平原在低頻率段的放大倍率有由平原邊緣往平原中央靠海處漸增的趨勢，此分佈與基盤面深度變化趨勢相當類似，可推測低頻率段之轉換函數表現出蘭陽平原區域速度及阻抗差異較大之層面所造成的放大效應；且宜蘭市到蘭陽溪出海口連線在各頻率段中皆有放大倍率高區存在。隨機式震源模型所得轉換函數雖在1Hz以下吻合程度較差，但放大倍率的趨勢仍在。本模型對各筆地震之震源簡化為點震源，省略描述各筆地震複雜破裂面的工作，可進一步納入更多的地震事件進行模擬並計算平均之轉換函數，加強描述蘭陽平原場址效應的客觀性與準確性。 The purpose of this study is using stochastic model to analyze the site effect in Lan-Yang plain. We utilized stochastic model to simulate the ground motion in soil site without taking into account the site effect, the simulated data could be considered as the data recorded by rock site. And we study the site effect by comparing observed with simulated data. The three earthquakes selected in this study are magnitude 5.0 at 13 October 1999, magnitude 6.2 at 10 September 2000 and magnitude 6.2 at 29 May 2002. We calibrated the model parameters by comparing rock sites records with simulated data. The distribution of dominant frequency in Lan-Yang plain calculated by stochastic model and H/V technique are almost the same. The dominant frequencies in west valley, Sanshing and Suau are greater than 3 to 4Hz, but in plain center are less than 2Hz. The result shows that the amplification in low frequency band increases from the plain edge to the seashore of plain center, it is quite similar to the depth variation of basement. It shows that the amplification effect at low frequency was caused by the interface with large difference of velocity and impedance. Additionally, there are high amplification district in each frequency band in the area between Yilan city and the outlet of Lan-Yang stream. Although simulated results are not so good at frequency band less than 1Hz, the results still show the trends of amplification. We take advantage of stochastic model which omits description of complicated source to calculate average transfer function with more earthquake events. More events could improve the externalism and accuracy of the site effect description in Lan-Yang plain.
