10:45 AM - 12:15 PM
[SGD02-P04] Attempt to detect approximately 10-minute events from high-rate GNSS data
Keywords:High-rate GNSS, Matched filter analysis, Ultra-slow earthquake
Slow earthquakes with various time-scales have been reported, but single events lasting from about 10 minutes to about a day have rarely been detected (e.g., Nishikawa et al., 2023). Whether this is because few events occur or because many occur but go undetected is unknown. This study is an attempt to detect events of about 10 minutes duration from high-rate GNSS data immediately after the 2011 Tohoku-oki earthquake.
High-rate GNSS data with 30-second sampling (Mitsui and Heki, 2012) are used for about 18 hours immediately after the Tohoku earthquake. Since the noise derived from the coordinate estimation algorithm (Precise Point Positioning) is much higher than that of ordinary daily GNSS data, it is necessary to improve the signal-to-noise ratio of the data. Therefore, the following preprocessing steps are performed: (1) horizontal vector synthesis in the direction of reverse fault slip at the plate boundary, (2) clustering of observation points based on the similarity of time-series data, and (3) common mode removal based on independent component analysis. Template matching based on correlation analysis is then performed to extract events that occur simultaneously at multiple observation points.
The template matching results show several instances of transient displacement at several stations near Ibaraki prefecture, which could be slip events. For the most likely event, an inversion analysis is performed. It assumes sub-faults at the plate boundary. The result is compared with the magnitude-duration scaling relationship of slow earthquakes. The possible event lies between the "ultra-slow earthquake" (Kawasaki et al., 1995) and ordinary earthquakes. This suggests that the possible event may be an induced slow slip event (SSE) due to postseismic deformation of earthquake.
High-rate GNSS data with 30-second sampling (Mitsui and Heki, 2012) are used for about 18 hours immediately after the Tohoku earthquake. Since the noise derived from the coordinate estimation algorithm (Precise Point Positioning) is much higher than that of ordinary daily GNSS data, it is necessary to improve the signal-to-noise ratio of the data. Therefore, the following preprocessing steps are performed: (1) horizontal vector synthesis in the direction of reverse fault slip at the plate boundary, (2) clustering of observation points based on the similarity of time-series data, and (3) common mode removal based on independent component analysis. Template matching based on correlation analysis is then performed to extract events that occur simultaneously at multiple observation points.
The template matching results show several instances of transient displacement at several stations near Ibaraki prefecture, which could be slip events. For the most likely event, an inversion analysis is performed. It assumes sub-faults at the plate boundary. The result is compared with the magnitude-duration scaling relationship of slow earthquakes. The possible event lies between the "ultra-slow earthquake" (Kawasaki et al., 1995) and ordinary earthquakes. This suggests that the possible event may be an induced slow slip event (SSE) due to postseismic deformation of earthquake.