9:15 AM - 9:30 AM
[SSS10-01] Identification of a non-planar fault from aftershock distribution with implicit modeling
Keywords:2016 Kumamoto earthquake, non-planar fault, surface fault
Fault geometry plays an important role in earthquake source dynamics, and is often identified from earthquake distribution (e.g., Ebllin and Michelini, 1986). The earthquake distribution is usually nonuniform within the source region of interest, and few earthquakes are distributed in some parts of the source region. In particular, aftershock distribution exhibits such heterogeneity, reflecting from slip distribution during a mainshock (Hartzell and Mendoza, 1989). In such a case, we need an analytic tool that enables us to fill the gap. In this study, we developed a method that is capable of filling the gap, and identified the fault geometry as a non-planar fault from aftershock activity following the 2016 Kumamoto, western Japan, earthquake (Mw7.0).
We adopted implicit modeling, and used signed distance and Radial Basis Function (RBF; Carr et al., 2001). In this methodology, the surface of interest is expressed as an iso-value one within a hypothetical scalar field. We analyzed the aftershocks that occurred within 24 hours of the mainshock listed in the earthquake catalog by Shito et al. (2020). The procedure of the analysis is as follows. First, we applied the Gaussian Mixture Model (Melchor and Goulding, 2018) to the aftershocks, and extracted the clusters associated with the fault under consideration. We then computed the normal vectors for each hypocenter from a local neighborhood of the hypocenters. Using the normal vectors, we augmented the data points along both directions of the normal vectors. Finally, we fitted an RBF to the augmented data set, and extracted the iso-value surface as the fault.
The resulting fault surface exhibits variation in both strike and dip directions, extending from northeast to southwest. The fault surface roughly divided into three parts along the strike direction, depending on the dip angle. In the middle part of the fault, the dip angle becomes shallower while near the ends of the fault, the dip angle becomes steeper. A comparison of our non-planar fault with the surface fault in the source region reveals that the identified fault reaches ground surface in close proximity to the Idenokuchi fault, which experienced a surface rupture during the 2016 earthquake. Additionally, the fault trace at the depth of 4 km exhibits strong correlation with the Hinagu fault, which also experienced surface rupture during the 2016 earthquake. These correspondences are consistent with the model of slip partitioning during the 2016 event (Toda et al., 2016), in which the Idenokuchi fault with normal faulting is the shallow extension of the causative fault while the Hinagu fault with strike faulting is vertically blanched from the causative fault. This result illustrates that the identification of a non-planar fault from precisely relocated hypocenters provides some insight into the relationship between the causative fault and the surface fault.
Acknowledgements: This work was supported by the MEXT Project for Seismology toward Research Innovation with Data of Earthquake (STAR-E), Grant Number JPJ010217.
We adopted implicit modeling, and used signed distance and Radial Basis Function (RBF; Carr et al., 2001). In this methodology, the surface of interest is expressed as an iso-value one within a hypothetical scalar field. We analyzed the aftershocks that occurred within 24 hours of the mainshock listed in the earthquake catalog by Shito et al. (2020). The procedure of the analysis is as follows. First, we applied the Gaussian Mixture Model (Melchor and Goulding, 2018) to the aftershocks, and extracted the clusters associated with the fault under consideration. We then computed the normal vectors for each hypocenter from a local neighborhood of the hypocenters. Using the normal vectors, we augmented the data points along both directions of the normal vectors. Finally, we fitted an RBF to the augmented data set, and extracted the iso-value surface as the fault.
The resulting fault surface exhibits variation in both strike and dip directions, extending from northeast to southwest. The fault surface roughly divided into three parts along the strike direction, depending on the dip angle. In the middle part of the fault, the dip angle becomes shallower while near the ends of the fault, the dip angle becomes steeper. A comparison of our non-planar fault with the surface fault in the source region reveals that the identified fault reaches ground surface in close proximity to the Idenokuchi fault, which experienced a surface rupture during the 2016 earthquake. Additionally, the fault trace at the depth of 4 km exhibits strong correlation with the Hinagu fault, which also experienced surface rupture during the 2016 earthquake. These correspondences are consistent with the model of slip partitioning during the 2016 event (Toda et al., 2016), in which the Idenokuchi fault with normal faulting is the shallow extension of the causative fault while the Hinagu fault with strike faulting is vertically blanched from the causative fault. This result illustrates that the identification of a non-planar fault from precisely relocated hypocenters provides some insight into the relationship between the causative fault and the surface fault.
Acknowledgements: This work was supported by the MEXT Project for Seismology toward Research Innovation with Data of Earthquake (STAR-E), Grant Number JPJ010217.