Model comparison for aftershock sequences following the 2005 Kashmir disaster

In an earlier study (Van Lieshout and Stein, in press) we postulated the existence of two major earthquakes in Kashmir instead of a single one, based upon the pattern of aftershocks. In this note we explore this pattern further by fitting several spatial point pattern models. In particular we discuss the Hawkes and the trigger process models for earthquake aftershock sequences following the Kashmir catastrophe in 2005. The minimum contrast method is used for estimation of the parameters. The study shows that the trigger model fits better than the Hawkes model. The most likely number of main shocks is rounded to 2 generating the almost 200 aftershocks, whereas the Hawkes model would estimate a parent process of approximately 18 parents with on average 10 descendants. We conclude that the spatial pattern of aftershocks can best be understood as being generated by the trigger point process model.