Modelling and forecasting three-dimensional-hypocentre seismicity in the Kanto region

This study analyses the seismicity in the Kanto region by fitting the 2-D-epicentre and 3-D-hypocentre ETAS models to the JMA catalogue for events above magnitude M4.0. In the 3-D ETAS model, the focal depth is assumed to follow the beta distribution. Compared with results from the 2-D-epicentre ETAS model, the 3-D ETAS model greatly improves the data fitting. In addition, the stochastic reconstruction method is used when validating the results of the 3-D ETAS model, with results indicating that the shallow events are more productive and their aftershocks decay slightly faster in the time and epicentre dimensions. We also study the changes of seismicity patterns before and after the 2011 Tohoku earthquake. The direct aftershocks of events from the post-Tohoku period are more diffusive in time and epicentre but more concentrated in depth. The seismicity rate increases significantly following the Tohoku earthquake, especially along the interface of the subducting Pacific plate. The curve of cumulative background probabilities for events above M4.0 implies that the background rate decays back to the pre-Tohoku level in about 5 yr after the Tohoku earthquake. However, the occurrence rates of smaller events (from M2.0 to M4.0) indicate that the adjustments of local stress field continue at finer scales. Finally, we verify that the 3-D model can reproduce the focal depths better than the 2-D model and improve the forecasting performance.