強震及其伴隨的大規模餘震經常對人們的財產與生命帶來巨大的威脅，為了提供震後救災的重要資訊，本文進行強震後餘震的即時風險評估。傳統的RJ模式 (Reasenberg and Jones, 1989) 假設餘震發生的時間與其規模是互相獨立的，因應實務所需，Chen等人 (2015) 提出規模與時間相依的修正RJ模式，記作MRJ模式。因為一個大規模餘震可能引發一個新的餘震序列，本文將流病式餘震序列模式 (ETAS; Ogata,1988) 加以簡化得到SETAS模式，或引用Chen等人 (2015) 結果修正之，得到METAS模式。事實上，本文也將空間訊息納入SETAS及METAS模式發展為可以用來描述餘震空間時間及規模風險的模式。為了說明本文模式的應用，本文分析2011年3月11日發生於日本東部海域，芮氏規模9.0的地震，並且探討如何藉由模式進行強餘震的預測。最後，使用接受者操作特徵曲線及尤登指數評估上述餘震預測的效應。;Strong earthquakes and their large offspring usually cause the huge threat to people′s property and lives. To provide information on the rescue work after the strong earthquakes, the immediate assessment of the risk of aftershocks is necessary. The traditional RJ model (Reasenberg and Jones, 1989) assumes that the time and magnitude of aftershocks are independent. For practical purpose, Chen et al. (2015) suggest the modified RJ model, in which the magnitude and time may be dependent. Since a large aftershock may initiate a new seismic sequence, the ETAS (epidemic type aftershock sequence) model (Ogata, 1988) is simplified to obtain the SETAS model. The SETAS model is also modified by using the results in Chen et al. (2015) to have the METAS model. In fact, the spatial information of the previous aftershocks is incorporated into the SETAS and METAS models. Therefore, in this paper, two models one obtained for describing the spatial-temporal-magnitude hazard of aftershocks. The application of the models proposed in this paper is illustrated based on the 5 years aftershocks after the 9.0 earthquake occurred on March 11, 2011, in Tohoku-Oki, Japan. The model-based prediction of future large aftershocks is also discussed. Finally, the efficiency of the prediction is evaluated by using the receiver operating characteristic (ROC) curve and Youden index.
