9:30 AM - 9:45 AM
[SGD03-03] Detection of short-term SSEs around Okinawa Island using stacking of GNSS time series and displacement vectors
Keywords:Short-term slow slip event, Okinawa main Island, GNSS, Ryukyu subduction zone
1. Introduction
Slow slip events (SSEs) are a type of slow earthquake detected only through geodetic signals without seismic wave emission. Short-term SSEs, lasting from several days to weeks, have been observed around Okinawa Island in the Ryukyu Trench (e.g., Nishimura, 2014), often synchronizing with very low-frequency earthquakes (VLFE) and low-frequency earthquakes (LFE) (Nakamura, 2017). However, GNSS stations are concentrated in southern Okinawa, causing regional variations in detection capability. Although past studies identified SSEs mainly offshore south of the island, it remains unclear if SSEs occur only in the south due to the small displacement (2–3 mm) they cause. In 2019, Kyoto University established three stations in northern Okinawa, improving detection capabilities. This study aims to detect short-term SSEs around Okinawa Island using GNSS time series and displacement vector stacking from over 20 years of data.
2. Data and Analysis Methodology
We used GNSS data from 33 stations operated by the Geospatial Information Authority of Japan (GEONET), Kyoto University, Japan Coast Guard, and Kyushu University, covering April 1997 to September 2024. Daily coordinates were estimated using the PPP-AR method with GipsyX. Stations with over one-third missing data or significant outliers were excluded. After correcting for artificial offsets (e.g., antenna replacements), we transformed coordinates from a geocentric to a local coordinate system (east-west, north-south, vertical). Missing values were interpolated linearly, and outliers exceeding 1.8 times the quartile deviation over 21 days were removed. Common short-period errors, identified from seven GEONET stations, were subtracted from all stations. We then applied a 21-day moving median to highlight SSE-related signals.
Using the processed data, we performed GNSS time series stacking and displacement stacking synchronized with VLFEs. For time series stacking, we partially modified the method of Bletery & Nocquet (2023). Daily inner products of observed and calculated displacements (from the Okada 1992 model) were summed. Calculations assumed a southern fault model (Nishimura, 2014) and a northern fault model shifted 30 km northeast along the trench. We computed a stacking time series s(t) for both models and fitted it with linear functions with and without a step. We evaluated SSE occurrence by calculating the difference in AIC (ΔAIC) between the two models within a 180-day moving window. ΔAIC was set to zero unless a positive step was detected.
For displacement stacking synchronized with VLFEs, we used VLFE data (Asano et al., 2015) from 2004 to 2014. We defined a VLFE burst as a day with at least five VLFEs and averaged the displacement distribution during bursts. We excluded bursts showing anomalous displacements deviating from the southeast direction expected for SSEs.
3. Results and Discussion
The time series stacking analysis showed that ΔAIC variations for both southern and northern fault models were consistent over time, with no distinct negative peak indicating an SSE exclusive to northern Okinawa. This suggests that SSEs do not occur solely in the north. However, displacement stacking synchronized with VLFEs revealed significant displacements not only south of Okinawa but also in the northern and back-arc regions. This implies that SSE faults synchronized with VLFEs may lie further offshore than the southern fault model assumed in this study.
Acknowledgments
We thank the Geospatial Information Authority of Japan, Japan Coast Guard, and Dr. Youichi Asano for providing GNSS and VLFE data (Asano et al., 2015). We also thank Dr. Tomoaki Nishikawa for his invaluable guidance.
Slow slip events (SSEs) are a type of slow earthquake detected only through geodetic signals without seismic wave emission. Short-term SSEs, lasting from several days to weeks, have been observed around Okinawa Island in the Ryukyu Trench (e.g., Nishimura, 2014), often synchronizing with very low-frequency earthquakes (VLFE) and low-frequency earthquakes (LFE) (Nakamura, 2017). However, GNSS stations are concentrated in southern Okinawa, causing regional variations in detection capability. Although past studies identified SSEs mainly offshore south of the island, it remains unclear if SSEs occur only in the south due to the small displacement (2–3 mm) they cause. In 2019, Kyoto University established three stations in northern Okinawa, improving detection capabilities. This study aims to detect short-term SSEs around Okinawa Island using GNSS time series and displacement vector stacking from over 20 years of data.
2. Data and Analysis Methodology
We used GNSS data from 33 stations operated by the Geospatial Information Authority of Japan (GEONET), Kyoto University, Japan Coast Guard, and Kyushu University, covering April 1997 to September 2024. Daily coordinates were estimated using the PPP-AR method with GipsyX. Stations with over one-third missing data or significant outliers were excluded. After correcting for artificial offsets (e.g., antenna replacements), we transformed coordinates from a geocentric to a local coordinate system (east-west, north-south, vertical). Missing values were interpolated linearly, and outliers exceeding 1.8 times the quartile deviation over 21 days were removed. Common short-period errors, identified from seven GEONET stations, were subtracted from all stations. We then applied a 21-day moving median to highlight SSE-related signals.
Using the processed data, we performed GNSS time series stacking and displacement stacking synchronized with VLFEs. For time series stacking, we partially modified the method of Bletery & Nocquet (2023). Daily inner products of observed and calculated displacements (from the Okada 1992 model) were summed. Calculations assumed a southern fault model (Nishimura, 2014) and a northern fault model shifted 30 km northeast along the trench. We computed a stacking time series s(t) for both models and fitted it with linear functions with and without a step. We evaluated SSE occurrence by calculating the difference in AIC (ΔAIC) between the two models within a 180-day moving window. ΔAIC was set to zero unless a positive step was detected.
For displacement stacking synchronized with VLFEs, we used VLFE data (Asano et al., 2015) from 2004 to 2014. We defined a VLFE burst as a day with at least five VLFEs and averaged the displacement distribution during bursts. We excluded bursts showing anomalous displacements deviating from the southeast direction expected for SSEs.
3. Results and Discussion
The time series stacking analysis showed that ΔAIC variations for both southern and northern fault models were consistent over time, with no distinct negative peak indicating an SSE exclusive to northern Okinawa. This suggests that SSEs do not occur solely in the north. However, displacement stacking synchronized with VLFEs revealed significant displacements not only south of Okinawa but also in the northern and back-arc regions. This implies that SSE faults synchronized with VLFEs may lie further offshore than the southern fault model assumed in this study.
Acknowledgments
We thank the Geospatial Information Authority of Japan, Japan Coast Guard, and Dr. Youichi Asano for providing GNSS and VLFE data (Asano et al., 2015). We also thank Dr. Tomoaki Nishikawa for his invaluable guidance.