09:30 〜 09:45
[SCG39-21] Comprehensive detection of tremor migration using multiple time windows beneath Kii Peninsula
キーワード：スロー地震、微動、マイグレーション
Deep low-frequency tremors occur downdip of seismogenic zone along the subducting plate interface. Together with short-term slow slip events (SSEs), tremors constitute the Episodic Tremor and Slip (ETS; Rogers and Dragert, 2003). Tremors are known to migrate in the along-strike direction at a velocity of ~10 km/day. After main slip front associated with the ETS passes, relatively small-scale and high-velocity migrations like as Rapid Tremor Reversal (RTR ; Houston et al., 2011) and streak (Ghosh et al., 2010) are often observed. RTR propagates in the direction opposite to that of the ETS main front at a velocity of ~100 km/day, and streak propagates in the along-dip direction at a velocity of ~1000 km/day. Bletery et al. (2017) interpreted these migrations of tremors to represent secondary slip fronts (SSFs) and detected SSFs with wide timescales in Cascadia. In this study, we comprehensively detect SSFs using multiple time windows and estimate migration parameters beneath the Kii Peninsula, southwest Japan.
We use a tremor catalog derived by using the hybrid method based on the envelope correlation considering the spatial distribution of amplitude (Maeda and Obara, 2009). We analyze tremors included in not only major ETS but also minor episodes beneath the Kii Peninsula during 2001 to 2020. In order to detect various SSFs, we use seven different time windows (1 h, 2 h, 4 h, 8 h, 16 h, 32 h,64 h). In this analysis, we assume that all tremor sources in a SSF migrate in one direction in each fixed time window and the migration front extends perpendicularly to the migration direction. First, we estimate the direction of migration as follows: (i) remove spatial outliers from the tremors included in the time window, (ii) calculate the correlation coefficient between time and distance along temporary migration direction, (iii) find a direction which gives the maximum correlation coefficient by a grid search. Next, we estimate the velocity of migration as follows: (i) remove spatiotemporal outliers by the linear regression of space–time plot projected on the direction of migration, (ii) find the slope of the regression line of remaining points. Finally, assuming that SSF area is a rectangle here, we estimate the spatial dimension of SSF as follows: (i) calculate the migration length from the product of migration velocity and duration, (ii) locate the spatial center of SSF in each time window, (iii) calculate standard deviation (σ) of distance between the tremor sources and a straight line in the migration direction through the center, (ⅳ) define the perpendicular width of SSF as 6σ.
As a result, the distribution of direction is isotropic in shorter time windows and is dominated in the along-strike direction as the time window becomes longer. This result is consistent with previous studies in the sense that SSF in the strike direction is dominant in longer time windows (Bletery et al., 2017; Ide, 2012; Obara et al., 2012). The velocity of SSFs becomes slower as the time window increases. For example, average velocity is 270 and 9 km/day for the time window of 1 and 64 h, respectively. Regarding the shape of the SSF, the perpendicular width is longer than the migration length in most of the SSFs. The area of SSFs does not change significantly in shorter time windows. The isotropic distribution of direction in shorter time windows suggest that short timescale migrations occur in a random direction. The along-strike direction in longer time window is interpreted to represent the main slip front. The result that most of the SSF migrates with a spread in the perpendicular direction to the migration direction suggests that migration is two-dimensional phenomena rather than one- dimensional phenomena like as streak.
We use a tremor catalog derived by using the hybrid method based on the envelope correlation considering the spatial distribution of amplitude (Maeda and Obara, 2009). We analyze tremors included in not only major ETS but also minor episodes beneath the Kii Peninsula during 2001 to 2020. In order to detect various SSFs, we use seven different time windows (1 h, 2 h, 4 h, 8 h, 16 h, 32 h,64 h). In this analysis, we assume that all tremor sources in a SSF migrate in one direction in each fixed time window and the migration front extends perpendicularly to the migration direction. First, we estimate the direction of migration as follows: (i) remove spatial outliers from the tremors included in the time window, (ii) calculate the correlation coefficient between time and distance along temporary migration direction, (iii) find a direction which gives the maximum correlation coefficient by a grid search. Next, we estimate the velocity of migration as follows: (i) remove spatiotemporal outliers by the linear regression of space–time plot projected on the direction of migration, (ii) find the slope of the regression line of remaining points. Finally, assuming that SSF area is a rectangle here, we estimate the spatial dimension of SSF as follows: (i) calculate the migration length from the product of migration velocity and duration, (ii) locate the spatial center of SSF in each time window, (iii) calculate standard deviation (σ) of distance between the tremor sources and a straight line in the migration direction through the center, (ⅳ) define the perpendicular width of SSF as 6σ.
As a result, the distribution of direction is isotropic in shorter time windows and is dominated in the along-strike direction as the time window becomes longer. This result is consistent with previous studies in the sense that SSF in the strike direction is dominant in longer time windows (Bletery et al., 2017; Ide, 2012; Obara et al., 2012). The velocity of SSFs becomes slower as the time window increases. For example, average velocity is 270 and 9 km/day for the time window of 1 and 64 h, respectively. Regarding the shape of the SSF, the perpendicular width is longer than the migration length in most of the SSFs. The area of SSFs does not change significantly in shorter time windows. The isotropic distribution of direction in shorter time windows suggest that short timescale migrations occur in a random direction. The along-strike direction in longer time window is interpreted to represent the main slip front. The result that most of the SSF migrates with a spread in the perpendicular direction to the migration direction suggests that migration is two-dimensional phenomena rather than one- dimensional phenomena like as streak.