3:30 PM - 3:45 PM
[SSS11-24] Source Rupture Process of the 2024 Hualien, Taiwan, Earthquake (MW 7.4) Estimated from Near-field Strong-motion Waveforms and GNSS Data
Keywords:2024 Hualien earthquake, Collision zone, Source inversion
On April 2, 2024, a reverse-fault earthquake (MW 7.4; depth: 22.5 km) occurred along the eastern coast of Hualien, Taiwan, at the collision zone between the Eurasian and Philippine Sea Plates. This collision zone extends to the surface as the Longitudinal Valley Fault. However, no surface offset along this fault associated with the 2024 Hualien earthquake has been observed. This study aims to enhance understanding of the source characteristics of earthquakes occurring at collision zones. To this end, the source rupture process of this earthquake was estimated through a joint inversion of near-field waveforms (0.05–0.25 Hz) and GNSS data.
Our inversion used the multi-time-window linear inversion method (Hartzell and Heaton, 1983). We assumed a fault plane with strike, dip, length, and width of 25°, 60°, 60 km, and 33 km, respectively, based on the crustal deformation from SAR interferometric images (the strike direction and the extent of the uplift area) and the aftershock distribution. This fault plane was divided into subfaults of 3 km by 3 km, and each subfault was modeled using three time windows, each a smoothed ramp function with a duration of 2.4 s, separated by 1.2 s. The data weighting in the joint inversion was set to a ratio of waveform:GNSS = 1:3, so that the sum of squares of observations for both types of data would be comparable. We computed the Green’s functions for each subfault using the discrete wavenumber method (Bouchon, 1981) and the reflection/transmission coefficient matrix method (Kennett and Kerry, 1979) for near-field waveforms, and the frequency-wavenumber method (Zhu and Rivera, 2002) for GNSS data. The 1D velocity structure model for each station was extracted from the 3D velocity structure model of Huang et al. (2014), and the extracted 1D models for near-field waveform stations were adjusted using the waveform records from an MW 5.0 aftershock.
The inversion result shows that the rupture propagated unilaterally northward and generated two large-slip areas (asperities) at depths greater than 20 km. The main rupture did not extend shallower than 20 km, which is consistent with the absence of a surface offset caused by this earthquake and differs from the case of the 2022 Chihshang earthquake sequence (MW 6.5 and MW 6.9) that occurred along the westward-dipping Central Range Fault within the same collision zone and generated significant surface offsets (Wang et al., 2024). The rupture propagation velocity triggering the first time window was 2.0 km/s, which minimized the data residuals. By examining the spatial distribution of the actual rupture propagation velocity, we suggest that the velocity within the asperity with the largest slip was slightly faster, at 2.5–3.0 km/s (65–75 % of S-wave velocity). The dimension of the rupture area (1980 km2), based on the criterion of Somerville et al. (1999), is slightly larger than the source scaling relationship for inland crustal earthquakes (Irikura and Miyake, 2011), but significantly smaller than that for plate boundary earthquakes (Murotani et al., 2008). The average slip across the rupture area (1.2 m) is smaller compared to the scaling relationship for inland crustal earthquakes, which may be related to the large focal depth and thus high rigidity.
Acknowledgements: This study was based on the 2024 research project “The study on improving the reliability of ground motion evaluation methods using fault rupture models” by the Secretariat of Nuclear Regulation Authority, Japan. The waveform and GNSS data were obtained from the Central Weather Administration and the Nevada Geodetic Laboratory, respectively.
Our inversion used the multi-time-window linear inversion method (Hartzell and Heaton, 1983). We assumed a fault plane with strike, dip, length, and width of 25°, 60°, 60 km, and 33 km, respectively, based on the crustal deformation from SAR interferometric images (the strike direction and the extent of the uplift area) and the aftershock distribution. This fault plane was divided into subfaults of 3 km by 3 km, and each subfault was modeled using three time windows, each a smoothed ramp function with a duration of 2.4 s, separated by 1.2 s. The data weighting in the joint inversion was set to a ratio of waveform:GNSS = 1:3, so that the sum of squares of observations for both types of data would be comparable. We computed the Green’s functions for each subfault using the discrete wavenumber method (Bouchon, 1981) and the reflection/transmission coefficient matrix method (Kennett and Kerry, 1979) for near-field waveforms, and the frequency-wavenumber method (Zhu and Rivera, 2002) for GNSS data. The 1D velocity structure model for each station was extracted from the 3D velocity structure model of Huang et al. (2014), and the extracted 1D models for near-field waveform stations were adjusted using the waveform records from an MW 5.0 aftershock.
The inversion result shows that the rupture propagated unilaterally northward and generated two large-slip areas (asperities) at depths greater than 20 km. The main rupture did not extend shallower than 20 km, which is consistent with the absence of a surface offset caused by this earthquake and differs from the case of the 2022 Chihshang earthquake sequence (MW 6.5 and MW 6.9) that occurred along the westward-dipping Central Range Fault within the same collision zone and generated significant surface offsets (Wang et al., 2024). The rupture propagation velocity triggering the first time window was 2.0 km/s, which minimized the data residuals. By examining the spatial distribution of the actual rupture propagation velocity, we suggest that the velocity within the asperity with the largest slip was slightly faster, at 2.5–3.0 km/s (65–75 % of S-wave velocity). The dimension of the rupture area (1980 km2), based on the criterion of Somerville et al. (1999), is slightly larger than the source scaling relationship for inland crustal earthquakes (Irikura and Miyake, 2011), but significantly smaller than that for plate boundary earthquakes (Murotani et al., 2008). The average slip across the rupture area (1.2 m) is smaller compared to the scaling relationship for inland crustal earthquakes, which may be related to the large focal depth and thus high rigidity.
Acknowledgements: This study was based on the 2024 research project “The study on improving the reliability of ground motion evaluation methods using fault rupture models” by the Secretariat of Nuclear Regulation Authority, Japan. The waveform and GNSS data were obtained from the Central Weather Administration and the Nevada Geodetic Laboratory, respectively.