2:30 PM - 2:45 PM
[SCG52-16] Tectonic stress fields estimated from long-term CMT data ranging over different periods
Keywords:stress, inversion theory, subduction zone processes, earthquake dynamics
Tectonic stress in the Earth's crust is the most essential physical quantity that controls earthquake generation. Because of the difficulty in measuring stress in situ at seismogenic depths, the inversion analysis of seismic data has played an important role in estimating tectonic stress fields for more than four decades. The CMT data inversion is a statistical method that estimates the spatial pattern of tectonic stress orientation from centroid moment tensor (CMT) data of seismic events, based on the Bayesian statistical inference and Akaike's Bayesian Information Criterion (ABIC). To obtain reliable inversion analysis results, we need to use as much data as possible, ranging over different periods.
In this study, we improved the CMT data inversion method (Terakawa and Matsu'ura, 2008) to incorporate preceding inversion analysis results into the present analysis as direct prior information about the stress field. In the improved method, in addition to the hyperparameter controlling the roughness of stress fields (indirect prior information) controlling the structure of model parameters, we introduced another hyperparameter that controls the relative weight of direct prior information to the observed data to be analysed. The relative weights of direct and indirect prior information to the observed data are appropriately determined by using ABIC. Given the optimum hyperparameters, we can obtain the best estimates of model parameters by using the maximum likelihood method. In addition, we can evaluate the variance of the uncertainty of direct prior information. The ratio of this quantity to the variance of estimation errors in the preceding inversion analysis gives us a simple indicator of temporal changes in stress orientation between the preceding and present periods.
We applied the improved method to the observed CMT data of seismic events of 24 years (1997-2020) in northeast Japan, where the Tohoku-oki megathrust event (Mw9.0) occurred on 11th March 2011 at the North American–Pacific plate boundary. In this analysis we declustered seismic events into four groups: single events, main shocks, foreshocks, and aftershocks. To construct a data set of independent stress indicators, we selected only single events and main shocks. To investigate temporal changes in tectonic stress orientation, we divided the target period into the three periods: P1 (Jan. 1997-Jan. 2007), P2 (Feb. 2007-Feb. 2011), and P3 (Mar. 2011-Dec. 2020). According to successive inversion analysis results for P1 and P2, we confirmed that the tectonic stress orientation was stable for 14 years before the megathrust event. On the other hand, successive inversion analysis results for P2 and P3 indicated a possibility that the 2011 megathrust event caused significant changes in tectonic stress orientation in four regions. However, three of them are most likely to be apparent. Only the change in stress orientation in the region off the coast of Fukushima and Ibaraki, the southwestern margin of the main rupture area of the megathrust event, seems to be real.
In this study, we improved the CMT data inversion method (Terakawa and Matsu'ura, 2008) to incorporate preceding inversion analysis results into the present analysis as direct prior information about the stress field. In the improved method, in addition to the hyperparameter controlling the roughness of stress fields (indirect prior information) controlling the structure of model parameters, we introduced another hyperparameter that controls the relative weight of direct prior information to the observed data to be analysed. The relative weights of direct and indirect prior information to the observed data are appropriately determined by using ABIC. Given the optimum hyperparameters, we can obtain the best estimates of model parameters by using the maximum likelihood method. In addition, we can evaluate the variance of the uncertainty of direct prior information. The ratio of this quantity to the variance of estimation errors in the preceding inversion analysis gives us a simple indicator of temporal changes in stress orientation between the preceding and present periods.
We applied the improved method to the observed CMT data of seismic events of 24 years (1997-2020) in northeast Japan, where the Tohoku-oki megathrust event (Mw9.0) occurred on 11th March 2011 at the North American–Pacific plate boundary. In this analysis we declustered seismic events into four groups: single events, main shocks, foreshocks, and aftershocks. To construct a data set of independent stress indicators, we selected only single events and main shocks. To investigate temporal changes in tectonic stress orientation, we divided the target period into the three periods: P1 (Jan. 1997-Jan. 2007), P2 (Feb. 2007-Feb. 2011), and P3 (Mar. 2011-Dec. 2020). According to successive inversion analysis results for P1 and P2, we confirmed that the tectonic stress orientation was stable for 14 years before the megathrust event. On the other hand, successive inversion analysis results for P2 and P3 indicated a possibility that the 2011 megathrust event caused significant changes in tectonic stress orientation in four regions. However, three of them are most likely to be apparent. Only the change in stress orientation in the region off the coast of Fukushima and Ibaraki, the southwestern margin of the main rupture area of the megathrust event, seems to be real.