5:15 PM - 6:45 PM
[U15-P22] Fault Model of the 2024 Noto Earthquake Estimated Using Aftershock, GNSS, and Tsunami Data
Keywords:2024 Noto Peninsula Earthquake, finite fault model, tsunami
On January 1, 2024, at 16:10 (local time), a magnitude (Mw) 7.5 earthquake occurred on the Noto Peninsula, Japan. Japan Meteorological Agency (JMA) seismic intensity scale of 7 was observed at Ishikawa Prefecture, and a tsunami warning was issued for a wide area along the Japan Sea. In this study, we constructed a finite fault model of this earthquake using the aftershock, GNSS, and tsunami datasets.
First, we estimated the fault geometry based on the aftershock distribution detected by JMA. Our analysis used the aftershock locations up to seven days after the mainshock. Since this earthquake can be considered as a multi-fault event (e.g., Fujii and Satake, 2024), we classified the aftershocks into four groups using a k-means clustering. Then, we applied a principal component analysis to each cluster to estimate the strike and dip angles of the fault plane (e.g., Yukutake and Iio, 2017; Shearer et al., 2003). The length and width of each fault were decided by trial and error to agree with the aftershock distribution (Fig 1a).
Based on this fault geometry, we conducted the joint inversion using GNSS and tsunami records to estimate the fault slip model. As GNSS data, we used the F5 solution of GEONET provided by Geospatial Information Authority of Japan (GSI). We considered the coseismic displacement as the difference between the seven-day average before and after the mainshock. We also collected the tsunami records from IOC of UNESCO, Nationwide Ocean Wave information network for Ports and HArbourS (NOWPHAS), and GSI. Since NOWPHAS records were published as images, we digitized them by WebPlotDigitizer (Rohatgi, 2022). As preprocessing, all the tide records were resampled in 60 seconds and the band-pass filter of 600-3600 sec was applied to remove the ocean tide component. Our inversion analysis estimated the rake angle and slip amount of each subfault, which was divided into four from each fault. We assumed that the subfaults belonging to the same fault have the same rake angle, so that, the total number of model parameters was 20 (16 slip amounts and 4 rake angles). We adopted the simulated annealing to solve this problem.
Fig 1b shows the finite fault model obtained by this study. The estimated Mw was 7.51 with the rigidity of 30 GPa, and the variance reductions of the tsunami and GNSS were 76.2% and 94.1%, respectively. The slip was divided into two areas with the hypocenter in between. In other words, the subfaults around the hypocenter did not rupture much. This is consistent with the result by Okuwaki et al. (2023); they found a “quiet rupture” or a low potency-rate density around the hypocenter using teleseismic P-wave. We also investigated the contribution of each fault to the observed tsunami. In the east of the Noto peninsula, the tsunami generated by the eastern fault (F1 and F2) was dominant. In the west, on the other hand, the tsunami by the westernmost fault (F4) hit first, and then, the tsunami propagating from the east of the peninsula arrived.
First, we estimated the fault geometry based on the aftershock distribution detected by JMA. Our analysis used the aftershock locations up to seven days after the mainshock. Since this earthquake can be considered as a multi-fault event (e.g., Fujii and Satake, 2024), we classified the aftershocks into four groups using a k-means clustering. Then, we applied a principal component analysis to each cluster to estimate the strike and dip angles of the fault plane (e.g., Yukutake and Iio, 2017; Shearer et al., 2003). The length and width of each fault were decided by trial and error to agree with the aftershock distribution (Fig 1a).
Based on this fault geometry, we conducted the joint inversion using GNSS and tsunami records to estimate the fault slip model. As GNSS data, we used the F5 solution of GEONET provided by Geospatial Information Authority of Japan (GSI). We considered the coseismic displacement as the difference between the seven-day average before and after the mainshock. We also collected the tsunami records from IOC of UNESCO, Nationwide Ocean Wave information network for Ports and HArbourS (NOWPHAS), and GSI. Since NOWPHAS records were published as images, we digitized them by WebPlotDigitizer (Rohatgi, 2022). As preprocessing, all the tide records were resampled in 60 seconds and the band-pass filter of 600-3600 sec was applied to remove the ocean tide component. Our inversion analysis estimated the rake angle and slip amount of each subfault, which was divided into four from each fault. We assumed that the subfaults belonging to the same fault have the same rake angle, so that, the total number of model parameters was 20 (16 slip amounts and 4 rake angles). We adopted the simulated annealing to solve this problem.
Fig 1b shows the finite fault model obtained by this study. The estimated Mw was 7.51 with the rigidity of 30 GPa, and the variance reductions of the tsunami and GNSS were 76.2% and 94.1%, respectively. The slip was divided into two areas with the hypocenter in between. In other words, the subfaults around the hypocenter did not rupture much. This is consistent with the result by Okuwaki et al. (2023); they found a “quiet rupture” or a low potency-rate density around the hypocenter using teleseismic P-wave. We also investigated the contribution of each fault to the observed tsunami. In the east of the Noto peninsula, the tsunami generated by the eastern fault (F1 and F2) was dominant. In the west, on the other hand, the tsunami by the westernmost fault (F4) hit first, and then, the tsunami propagating from the east of the peninsula arrived.