台灣地區餘震風險之統計分析(2/2)

Taiwan, located on the boundary of the Eurasia Plate and Philippine Sea Plate, frequently suffers with earthquakes due to continual plate movement. Moreover, sensible aftershocks that occur after a strong main shock might further destroy buildings or damage weaken structures. Therefore, a real time assessment ofthe aftershock hazard after the strong main shock is an important issue, especially on providing valuable information for the ongoing emergent rescue. To do so, the Reasenberg-Jones (RJ) model is developed to describe the magnitude-time hazard of aftershocks where the associate parameters are conventionally estimated based on the maximum likelihood (ML) method. However, the ML-estimated RJ (MLRJ) model may not be an efficient one when evaluating the aftershock hazard in a short period after the main shock due to thesparsity of aftershocks. Therefore, in the first year of the two-year project, aftershock sequences are declustered from the earthquake catalog recorded by the Central Weather Bureau (CWB) of Taiwan. The generic RJ (GRJ) model that contains the information about the magnitude, epicenter, focal depth and focal mechanism of the main shock in the Taiwan area is then developed for evaluating the time-magnitude hazard of the forthcoming aftershock sequences. In the second year, the Bayesian generic RJ (BGRJ) model is obtained by adjusting the GRJ model based on the available aftershocks of the current main shock which provides a near-real time adaptive evaluation of aftershock hazard. In the two-year project, the main shocks and corresponding aftershocksequences under study are divided into the training and validation data sets. The maximum likelihood estimated GRJ model (MLGRJ) is built based on the training data for finding a confidence interval for the number of sensible aftershocks. On the other hand, the credible interval for the number of sensible aftershocks isconstructed based on the BGRJ model that incorporates the GRJ and current available aftershocks. Finally, the performance of MLGRJ and BGRJ models on the evaluation of the aftershock hazard in the Taiwan area is investigated by comparing the estimated number of aftershocks with the real one in the validating data set.