3:30 PM - 3:45 PM
[SCG39-13] Development of the detection method for short-term slow slip events using GNSS data and its application to the Nankai subduction zone
Keywords:Slow slip event, Nankai subduction zone, GNSS
Slow slip events (SSEs), which are kinds of slow earthquakes, are observed in a geodetic time domain and are categorized by their duration into long-term SSEs lasting several months to years and short-term SSEs (S-SSEs) lasting several days to weeks. S-SSEs east of the Bungo Channel have been systematically detected by using tiltmeters [Sekine et al., 2010] and GNSS [Nishimura et al., 2013] and their duration has been determined by tiltmeters based on corresponding low-frequency tremors [e.g., Sekine et al., 2010]. However, it is important to estimate the duration of S-SSEs by using only geodetic measurements because S-SSEs do not always accompany tremors. On the other hand, S-SSEs in the Kyushu region have been detected by only using GNSS data [Nishimura, 2014] and their duration has never been estimated objectively. Therefore, we developed a new S-SSE detection method which can estimate their source parameters including duration and applied it to 23-year-long GNSS data in the whole Nankai subduction zone from the Kyushu to the Tokai regions.
In this study, we use 734 GNSS stations in the period from February 1, 1997, to January 31, 2020. These coordinates were estimated using GIPSY/OASIS version 6.4 software with a strategy of precise point positioning-ambiguity resolution. As preprocessing of GNSS data, we remove post-seismic deformation, co-seismic displacements, long-terms trend including seasonal oscillation, and common-mode error. Firstly, we compute correlation coefficients between a synthetic template representing a time evolution of an S-SSE and 121-day-coordinates of GNSS data for each horizontal component at all stations. Secondly, we average the correlation coefficients weighted by the synthetic displacements predicted from the sub-faults. In this study, we assumed 429 sub-faults on the Philippine Sea plate interface [Iwasaki et al., 2015] and computed synthetic displacements with an elastic half-space dislocation model [Okada, 1992]. We then detected dates and locations of the candidate events by applying thresholds to the weighted average of correlation coefficients. Thirdly, we estimate a rectangular fault model of the events [Matsu’ura and Hasegawa, 1987] by using observed displacements of the candidate event. Fourthly, we stack 121-day-long GNSS coordinates weighted by the displacements predicted from the estimated fault model. Then, we estimate a duration of the candidate event by extracting a maximum correlation coefficient between stacked coordinates and a synthetic template. Finally, we detect S-SSEs by applying the criteria to the fault model and duration estimation result of the candidate events.
In our preliminary result, we detected 280 S-SSEs in the Nankai subduction zone during the analyzed period. Their typical moment magnitude is approximately 6.0, and their duration mainly distributes from 1 to 10 days. Some of S-SSEs detected in offshore Kyushu roughly correspond to repeating earthquakes [Uchida et al., 2020] and very low-frequency earthquakes [Baba et al., 2020]. This is the first geodetic evidence of the synchronization between S-SSEs and other phenomena in offshore Kyushu.
Estimation of slip amount and duration in the developed method enables us to clarify the spatial variation of slip rates of S-SSEs in the Nankai subduction zone. The average slip rate of S-SSEs, which is computed by dividing the cumulative slip by the cumulative duration, shows some regional characteristics (Figure 1). The average slip rate in western Shikoku is about twice larger than that of eastern Shikoku, and the average slip rate in Kyushu is the smallest among the Nankai subduction zone. These differences are statistically significant and possibly relate to the geometry of the subducting Philippine Sea plate.
Acknowledgments
We thank the Geospatial Information Authority of Japan, Japan Coast Guard, Japan Crustal Activity Science Consortium, and International GNSS Service for providing GNSS RINEX data. This research was supported by JSPS KAKENHI Grant Number JP16H06474 in Scientific Research on Innovative Areas “Science of Slow Earthquakes”, the MEXT of Japan under its "The Second Earthquake and Volcano Hazards Observation and Research Program", and ERI under its "Joint Usage/Research Program 2020-A-03".
In this study, we use 734 GNSS stations in the period from February 1, 1997, to January 31, 2020. These coordinates were estimated using GIPSY/OASIS version 6.4 software with a strategy of precise point positioning-ambiguity resolution. As preprocessing of GNSS data, we remove post-seismic deformation, co-seismic displacements, long-terms trend including seasonal oscillation, and common-mode error. Firstly, we compute correlation coefficients between a synthetic template representing a time evolution of an S-SSE and 121-day-coordinates of GNSS data for each horizontal component at all stations. Secondly, we average the correlation coefficients weighted by the synthetic displacements predicted from the sub-faults. In this study, we assumed 429 sub-faults on the Philippine Sea plate interface [Iwasaki et al., 2015] and computed synthetic displacements with an elastic half-space dislocation model [Okada, 1992]. We then detected dates and locations of the candidate events by applying thresholds to the weighted average of correlation coefficients. Thirdly, we estimate a rectangular fault model of the events [Matsu’ura and Hasegawa, 1987] by using observed displacements of the candidate event. Fourthly, we stack 121-day-long GNSS coordinates weighted by the displacements predicted from the estimated fault model. Then, we estimate a duration of the candidate event by extracting a maximum correlation coefficient between stacked coordinates and a synthetic template. Finally, we detect S-SSEs by applying the criteria to the fault model and duration estimation result of the candidate events.
In our preliminary result, we detected 280 S-SSEs in the Nankai subduction zone during the analyzed period. Their typical moment magnitude is approximately 6.0, and their duration mainly distributes from 1 to 10 days. Some of S-SSEs detected in offshore Kyushu roughly correspond to repeating earthquakes [Uchida et al., 2020] and very low-frequency earthquakes [Baba et al., 2020]. This is the first geodetic evidence of the synchronization between S-SSEs and other phenomena in offshore Kyushu.
Estimation of slip amount and duration in the developed method enables us to clarify the spatial variation of slip rates of S-SSEs in the Nankai subduction zone. The average slip rate of S-SSEs, which is computed by dividing the cumulative slip by the cumulative duration, shows some regional characteristics (Figure 1). The average slip rate in western Shikoku is about twice larger than that of eastern Shikoku, and the average slip rate in Kyushu is the smallest among the Nankai subduction zone. These differences are statistically significant and possibly relate to the geometry of the subducting Philippine Sea plate.
Acknowledgments
We thank the Geospatial Information Authority of Japan, Japan Coast Guard, Japan Crustal Activity Science Consortium, and International GNSS Service for providing GNSS RINEX data. This research was supported by JSPS KAKENHI Grant Number JP16H06474 in Scientific Research on Innovative Areas “Science of Slow Earthquakes”, the MEXT of Japan under its "The Second Earthquake and Volcano Hazards Observation and Research Program", and ERI under its "Joint Usage/Research Program 2020-A-03".