[S-TT44] Seismic Big Data Analysis Based on the State-of-the-Art of Bayesian Statistics

[STT44-P03] Forecasting Aftershocks Immediately After the Large Main Shock with Epidemic-type Aftershock Detection Function

*Kosuke Morikawa¹, Hiromichi Nagao^2,3, Naoshi Hirata^4,2 (1.Osaka University, 2.Earthquake Research Institute, the University of Tokyo, 3.Graduate School of Information Science and Technology, the University of Tokyo, 4.National Research Institute for Earth Science and Disaster Resilience)

Keywords:Statistical seismology, ETAS model, Score Matching

Immediate forecasting of the temporal and spatial distribution of aftershock sequences is vital for disaster prevention. However, identifying the number of aftershocks immediately after the main shock is challenging due to the contaminations of arriving seismic waves. To overcome this difficulty, some researchers introduced detection functions of the aftershocks to correct the bias caused by the underdetection of the aftershocks. However, existing models of the detection probability for an aftershock do not utilize the information on aftershock sequences that occurred before the aftershock, which vitiates model flexibility. We aim to propose a model that can fully employ the local information of the aftershock sequences.

On the other hand, the Omori-Utsu law is well-known for the law of aftershock frequency, and the ETAS (epidemic-type aftershock sequences) model extends the Omori-Utsu law to make the past information available. We propose a simple detection function that can use the past information of the aftershocks like ETAS. However, the estimation of the model is cumbersome because the likelihood includes multiple integrals that are unrealistic to compute. Therefore, we avoid using the maximum likelihood estimation and propose an effective method without computing the multiple integrals using the score-matching technique. The proposed method is applied to the 2004 Chuetsu Earthquake and the 2016 Kumamoto Earthquake.