2:30 PM - 2:45 PM
[STT42-04] Development of a Quasi-Real-Time Analysis System for Seismic Activity in the Shallow Subduction Zone Using Seafloor DAS and OBS Records
Keywords:Tremor, Nankai trough, slow earthquake, DONET, DAS
In the Nankai Trough, it is known that small-sized earthquakes and tremors occur surrounding the region where large megathrust earthquakes are expected to take place. Comprehensive monitoring of these seismic activities is key to understanding the processes of stress accumulation and release processes along the plate boundary near megathrust earthquake rupture zones. Ocean-bottom observations are crucial for such monitoring since shallow subduction zones are located beneath the seafloor. However, compared to land-based networks, ocean-based networks usually have a sparse station distribution. Recently, distributed acoustic sensing (DAS) has garnered attention as a technology that enables dense and cost-effective seismic monitoring. In the Nankai Trough, in addition to the OBS stations of the DONET network, continuous DAS monitoring is being conducted using a submarine cable extending 120 km offshore from Cape Muroto.
This study aims to monitor seismic activity in the shallow part of the Nankai Trough over a wide area in quasi-real-time by jointly using offshore DAS strain-rate records and DONET-OBS velocity records. First, we installed a new analysis server at the land station of the Muroto offshore cable. Then, we established an environment to receive real-time offshore DAS and OBS data from a different server in the same Muroto station and from JAMSTEC Yokohama Institute, respectively. Since the data acquisition timing differs between datasets, adjustments were made to ensure smooth data processing. For the collected data, two types of analysis are performed. The first process focuses on ordinary earthquakes and involves reading arrival times using PhaseNet (Zhu & Beroza, 2019) and determining hypocenter locations using hypomh (Hirata & Matsu’ura, 1987) and HypoNet Nankai (Agata et al., 2024 ArXiv), which is based on a three-dimensional structural model of the Nankai Trough region. The second process targets tremors, a type of slow earthquake, and uses the envelope correlation method (ECM; Mizuno & Ide, 2019) to detect and locate seismic events. Both analyses process data in 120-second windows every 60 seconds. Finally, duplicate seismic events detected in overlapping time windows are removed to obtain the final earthquake catalog. Since August 2024, this system has been in trial operation, confirming that seismic event hypocenters can be estimated with high sensitivity.
In this presentation, we also introduce modifications made in ECM to detect local earthquakes beneath the seafloor. The existing ECM was originally developed for tremors located deeper than 30km beneath land. Therefore, when applied to earthquakes located shallower than 10 km beneath the seafloor, frequent false detections and missed detections occur. It is crucial to clarify at which stage of the ECM these issues arise, in order to improve the detection accuracy. First, we demonstrate that the problem of detecting teleseismic events or T-waves as local earthquakes occurs during the signal association process, where signals with a sufficiently high number of station pairs with high correlations are associated as a single seismic event. Second, we show that the problem of missing local earthquakes arises from poor accuracy in hypocenter determination. To address these issues, we propose solutions that utilize features such as the maximum amplitude of signals and present the results of applying these improvements to past DONET data. In the future, this enhanced ECM will be applied to the aforementioned seismic monitoring system utilizing DAS and OBS.
This study aims to monitor seismic activity in the shallow part of the Nankai Trough over a wide area in quasi-real-time by jointly using offshore DAS strain-rate records and DONET-OBS velocity records. First, we installed a new analysis server at the land station of the Muroto offshore cable. Then, we established an environment to receive real-time offshore DAS and OBS data from a different server in the same Muroto station and from JAMSTEC Yokohama Institute, respectively. Since the data acquisition timing differs between datasets, adjustments were made to ensure smooth data processing. For the collected data, two types of analysis are performed. The first process focuses on ordinary earthquakes and involves reading arrival times using PhaseNet (Zhu & Beroza, 2019) and determining hypocenter locations using hypomh (Hirata & Matsu’ura, 1987) and HypoNet Nankai (Agata et al., 2024 ArXiv), which is based on a three-dimensional structural model of the Nankai Trough region. The second process targets tremors, a type of slow earthquake, and uses the envelope correlation method (ECM; Mizuno & Ide, 2019) to detect and locate seismic events. Both analyses process data in 120-second windows every 60 seconds. Finally, duplicate seismic events detected in overlapping time windows are removed to obtain the final earthquake catalog. Since August 2024, this system has been in trial operation, confirming that seismic event hypocenters can be estimated with high sensitivity.
In this presentation, we also introduce modifications made in ECM to detect local earthquakes beneath the seafloor. The existing ECM was originally developed for tremors located deeper than 30km beneath land. Therefore, when applied to earthquakes located shallower than 10 km beneath the seafloor, frequent false detections and missed detections occur. It is crucial to clarify at which stage of the ECM these issues arise, in order to improve the detection accuracy. First, we demonstrate that the problem of detecting teleseismic events or T-waves as local earthquakes occurs during the signal association process, where signals with a sufficiently high number of station pairs with high correlations are associated as a single seismic event. Second, we show that the problem of missing local earthquakes arises from poor accuracy in hypocenter determination. To address these issues, we propose solutions that utilize features such as the maximum amplitude of signals and present the results of applying these improvements to past DONET data. In the future, this enhanced ECM will be applied to the aforementioned seismic monitoring system utilizing DAS and OBS.