5:15 PM - 6:45 PM
[U15-P09] Fault geometries of the 2024 Noto Peninsula earthquake estimated by hypocenter clustering
Keywords:fault geometry, hypocenter distribution, clustering, Noto Peninsula Earthquake
On 1 January 2024, a devastating large earthquake (Mw 7.5) struck the Noto Peninsula in central Japan. This earthquake is believed to be associated with a series of swarm activities that began in December 2020 in the northeastern area. A substantial number of aftershocks were observed from the western end to the northeastern offshore of the Noto Peninsula. The aftershock distribution, according to the Japan Meteorological Agency Unified Earthquake Catalog, shows a southeast dipping structure beneath the Noto Peninsula, which is consistent with the hypothetical fault model of “F43” by the Ministry of Land, Infrastructure, Transport and Tourism, Japan (2014). To investigate a more detailed geometry of fault planes for this earthquake sequence, we developed a new hypocenter clustering method and applied this method to the relocated aftershocks by Shiina et al. (2024, JpGU).
Our method is based on the hypocenter clustering method proposed by Sato et al. (2022, SSJ; 2023, JpGU). By considering the hypocenter distribution as point cloud data, the two-step clustering of point-cloud normal vectors and relocated hypocenters opens the possibility of detecting detailed fault plane geometries (Sato et al., 2023, JpGU). Point-cloud normal vectors, computed by k-Nearest Neighbor Principal Component Analysis (KNN-PCA), reflect the local planar geometry of surrounding earthquakes. Provided that aftershocks occur around planar faults, the set of events with a similar orientation of normal vector would form a fault plane. Instead of performing the two-step clustering (Sato et al., 2023, JpGU), we developed a new method that uses 6-D feature vectors containing both point-cloud normal vectors and hypocenter positions, which are then classified into distinct clusters using HDBSCAN (Campello et al., 2013). The distance between two feature vectors is defined as the square root of the sum of the modified cosine distance (Sawaki et al., 2023, SSJ) and the squared difference of standardized positions. Then we extract a plane perpendicular to the eigenvector of each cluster with its minimum eigenvalue and calculate the length, width, strike, and dip of the plane.
We obtained southeast-dipping main fault planes beneath the Noto Peninsula. The length of these fault planes is about 20 km, and they are in a chain to a northeast-southwest direction with a slight change in strike. The depth cross-section of aftershocks and obtained fault planes revealed that the main fault planes are situated several kilometers above the lineament of the earthquake swarms and the aftershock activity of the 2023 M6.5 earthquake (Kato, 2024, GRL). For the northeastern seaside, we found it difficult to retrieve plausible planes because of the sparse distribution of relocated aftershocks and their low reliability. In contrast, we extracted an east-dipping plane on the western side of the Peninsula. The depth cross-section of aftershocks also demonstrated a clear change in fault structures around the western side. This east-dipping plane is different from that of the 2007 Mw6.7 earthquake, steeply dipping toward the southeast direction (Sakai et al., 2008, EPS). A drastic rift of < 4 m along the western coast of the Noto Peninsula, caused by the 2024 earthquake, has been confirmed through geodetic and geological observations (GSI, 2024, https://www.gsi.go.jp/uchusokuchi/20240101noto_insar.html; GSJ 4th report of the 2024 Noto Peninsula earthquake, https://www.gsj.jp/hazards/earthquake/noto2024/noto2024-04.html). Reverse fault slip on the east-dipping plane could reproduce the large deformation along the western coast. Slip inversion or dynamic rupture modeling using smaller-scale fault structures would help better reproduction of mainshock fault slip.
[Acknowledgments]
This study was supported by MEXT Project for Seismology toward Research Innovation with Data of Earthquake (STAR-E) Grant Number JPJ010217.
Our method is based on the hypocenter clustering method proposed by Sato et al. (2022, SSJ; 2023, JpGU). By considering the hypocenter distribution as point cloud data, the two-step clustering of point-cloud normal vectors and relocated hypocenters opens the possibility of detecting detailed fault plane geometries (Sato et al., 2023, JpGU). Point-cloud normal vectors, computed by k-Nearest Neighbor Principal Component Analysis (KNN-PCA), reflect the local planar geometry of surrounding earthquakes. Provided that aftershocks occur around planar faults, the set of events with a similar orientation of normal vector would form a fault plane. Instead of performing the two-step clustering (Sato et al., 2023, JpGU), we developed a new method that uses 6-D feature vectors containing both point-cloud normal vectors and hypocenter positions, which are then classified into distinct clusters using HDBSCAN (Campello et al., 2013). The distance between two feature vectors is defined as the square root of the sum of the modified cosine distance (Sawaki et al., 2023, SSJ) and the squared difference of standardized positions. Then we extract a plane perpendicular to the eigenvector of each cluster with its minimum eigenvalue and calculate the length, width, strike, and dip of the plane.
We obtained southeast-dipping main fault planes beneath the Noto Peninsula. The length of these fault planes is about 20 km, and they are in a chain to a northeast-southwest direction with a slight change in strike. The depth cross-section of aftershocks and obtained fault planes revealed that the main fault planes are situated several kilometers above the lineament of the earthquake swarms and the aftershock activity of the 2023 M6.5 earthquake (Kato, 2024, GRL). For the northeastern seaside, we found it difficult to retrieve plausible planes because of the sparse distribution of relocated aftershocks and their low reliability. In contrast, we extracted an east-dipping plane on the western side of the Peninsula. The depth cross-section of aftershocks also demonstrated a clear change in fault structures around the western side. This east-dipping plane is different from that of the 2007 Mw6.7 earthquake, steeply dipping toward the southeast direction (Sakai et al., 2008, EPS). A drastic rift of < 4 m along the western coast of the Noto Peninsula, caused by the 2024 earthquake, has been confirmed through geodetic and geological observations (GSI, 2024, https://www.gsi.go.jp/uchusokuchi/20240101noto_insar.html; GSJ 4th report of the 2024 Noto Peninsula earthquake, https://www.gsj.jp/hazards/earthquake/noto2024/noto2024-04.html). Reverse fault slip on the east-dipping plane could reproduce the large deformation along the western coast. Slip inversion or dynamic rupture modeling using smaller-scale fault structures would help better reproduction of mainshock fault slip.
[Acknowledgments]
This study was supported by MEXT Project for Seismology toward Research Innovation with Data of Earthquake (STAR-E) Grant Number JPJ010217.