為能在強震後即時進行餘震風險的評估，以便提供震後救災工作的重要資訊，傳統上採用RJ模式（Reasenberg and Jones，1989）研究餘震發生的時間及規模分布，其中RJ模式假設餘震的發生時間與其規模為獨立的。實務上，距主震時間越久，發生大規模餘震的機率越小，因此，Chen等人（2015）提出規模與時間相依的修正RJ模式，記作MRJ。本文則進一步利用貝氏方法，根據MRJ模式分析1999年發生在台灣集集附近，芮氏規模7.3的地震之後的餘震序列。本文討論二種貝氏分析，一種是傳統的貝氏分析(Bayesian analysis)，亦即MRJ模式參數的事前分布由之前的相關歷史地震中得知；另一種貝氏分析則在集集餘震序列中隨時調整事前分布(prior distribution)，稱之為調整型貝氏分析。本文將探討上述兩種貝氏析，進行評估或預測規模5以上集集餘震的風險。;To evaluate the risk of aftershock immediately after a strong earthquake for emergency rescue, the Reasenberg-Jones(RJ) model is conventionally employed, which assumes that the occurrence time and magnitude of aftershocks are independent, In practice, the longer it is after a major earthquake, the smaller the chance of a strong aftershock. Hence, Chen et al.(2015) proposed a modified RJ model, called MRJ, that allows the dependence between the time and magnitude. This thesis employs Bayesian analysis to analyze the MRJ model for the aftershock sequence after a M_L7.3, Taiwan, Chichi earthquake in 1999. There are two versions two kinds of Bayesian analysis, one is the traditional Bayesian analysis which is done when the prior distribution of the parameters in MRJ model is given from related historic earthquakes, the other is sequentially adjusted the prior distribution take previous distributions, called the sequentially updated Bayesian method. The two different Bayesian methods are need to evaluate the risk of above chi-chi aftershocks with magnitude at least 5.
