09:15 〜 09:30
[SCG50-02] カリフォルニア州及びオクラホマ州におけるマルチスケール断層面推定
キーワード：断層面形状、震源分布、クラスタリング、Ridgecrest、群発地震、誘発地震
The distribution of earthquake hypocenters provides detailed information about fault geometries. Numerous significant earthquakes have occurred in Southern California around the San Andreas Fault system. Most of the earthquakes are distributed vertically across several conjugate systems. In July 2019, the Ridgecrest area in California was impacted by sequences of earthquakes with moment magnitudes (Mw) of 6.4 and 7.1. The aftershock sequences displayed several lineaments, which were not only from the mainshock fault systems but also from various conjugate systems (Shelly, 2020, SRL). Also, a large area of induced seismicity around the Oklahoma–Kansas boundary has been well-documented (e.g., Ellsworth, 2013, Science). The seismicity induced in Oklahoma clearly displays numerous lineaments with a bimodal strike distribution, and many earthquakes with magnitudes greater than 4.0 have occurred. Enhanced monitoring of micro-seismicity has demonstrated that the lineaments of the seismicity represent the histories of faulting for most earthquakes with Mw>=4 (Park et al., 2022, TSR). In short, investigation of a detailed hypocenter distribution is essential for a better understanding of the history and future of earthquake faulting and tectonic architectures.
Hypocenter clustering enables the objective exploration of subsurface crustal faults. The three-dimensional hypocenter distribution often reflects the complexity of fault structures. Hypocenter classification methods using principal component analysis (PCA) or machine learning techniques may be capable of reconstructing the architectures of crustal faults (e.g., Ouillon et al., 2008, JGR; Kamer et al., 2020, NHESS). Furthermore, by considering the hypocenter distribution as point cloud data, the two-step clustering of point-cloud normal vectors by PCA and relocated hypocenters opens the possibility of detecting detailed fault plane geometries (Sato et al., 2023, JpGU). This two-step clustering method was applied to the relocated earthquake catalog for the San Andreas Fault in southern California (Sato et al., 2023, JpGU) and the western Tottori district (Sawaki et al., 2023, AGU). In this study, after making improvements and conducting a parameter study, we applied this hypocenter clustering method to the micro-seismicity catalogs for the 2019 Ridgecrest earthquake sequences (Shelly, 2020, SRL), the 2020 Maacama swarm-like sequences (Shelly et al., 2022, TSR; 2023, GRL) in California, and the induced seismicity in Oklahoma–Kansas (Park et al., 2022, TSR), with the aim of revealing complex fault geometries.
In the 2019 Ridgecrest aftershock sequences, we identified NE–SW trending planes for the Mw 6.4 event on July 4 and NW–SE trending vertical planes for the Mw 7.1 event on July 6. We successfully detected two parallel vertical planes at the southern edge of the Mw 7.1 mainshock. We also identified a fault plane of a Mw 5.4 event on July 5 near the mainshock and multiple conjugate fault planes trending in the NE–SW direction. Estimated planes are mostly consistent with surface fault traces (DuRoss et al., 2020, BSSA) and visually interpreted fault planes (Shelly, 2020, SRL). However, at the northern edge of the mainshock fault, we detected numerous eastward-dipping planes. These were interpreted as complexly intersecting vertical planes due to high-rate shallow aftershocks in the dilatational quadrant (Plesch et al., 2020, BSSA; Shelly, 2020, SRL). In the Maacama sequence, we also identified primary deep planes trending NW–SE and shallow NE-striking planes (Shelly et al., 2023, GRL). We confirmed the clustering method with point-cloud normal can identify smaller faults over a broader area.
[Acknowledgments]
This study was supported by MEXT Project for Seismology toward Research Innovation with Data of Earthquake (STAR-E) Grant Number JPJ010217. We had a fruitful discussion with Rachel Abercrombie.
Hypocenter clustering enables the objective exploration of subsurface crustal faults. The three-dimensional hypocenter distribution often reflects the complexity of fault structures. Hypocenter classification methods using principal component analysis (PCA) or machine learning techniques may be capable of reconstructing the architectures of crustal faults (e.g., Ouillon et al., 2008, JGR; Kamer et al., 2020, NHESS). Furthermore, by considering the hypocenter distribution as point cloud data, the two-step clustering of point-cloud normal vectors by PCA and relocated hypocenters opens the possibility of detecting detailed fault plane geometries (Sato et al., 2023, JpGU). This two-step clustering method was applied to the relocated earthquake catalog for the San Andreas Fault in southern California (Sato et al., 2023, JpGU) and the western Tottori district (Sawaki et al., 2023, AGU). In this study, after making improvements and conducting a parameter study, we applied this hypocenter clustering method to the micro-seismicity catalogs for the 2019 Ridgecrest earthquake sequences (Shelly, 2020, SRL), the 2020 Maacama swarm-like sequences (Shelly et al., 2022, TSR; 2023, GRL) in California, and the induced seismicity in Oklahoma–Kansas (Park et al., 2022, TSR), with the aim of revealing complex fault geometries.
In the 2019 Ridgecrest aftershock sequences, we identified NE–SW trending planes for the Mw 6.4 event on July 4 and NW–SE trending vertical planes for the Mw 7.1 event on July 6. We successfully detected two parallel vertical planes at the southern edge of the Mw 7.1 mainshock. We also identified a fault plane of a Mw 5.4 event on July 5 near the mainshock and multiple conjugate fault planes trending in the NE–SW direction. Estimated planes are mostly consistent with surface fault traces (DuRoss et al., 2020, BSSA) and visually interpreted fault planes (Shelly, 2020, SRL). However, at the northern edge of the mainshock fault, we detected numerous eastward-dipping planes. These were interpreted as complexly intersecting vertical planes due to high-rate shallow aftershocks in the dilatational quadrant (Plesch et al., 2020, BSSA; Shelly, 2020, SRL). In the Maacama sequence, we also identified primary deep planes trending NW–SE and shallow NE-striking planes (Shelly et al., 2023, GRL). We confirmed the clustering method with point-cloud normal can identify smaller faults over a broader area.
[Acknowledgments]
This study was supported by MEXT Project for Seismology toward Research Innovation with Data of Earthquake (STAR-E) Grant Number JPJ010217. We had a fruitful discussion with Rachel Abercrombie.