5:15 PM - 6:45 PM
[U15-P29] Seismic array analysis of strong ground motions from the 2024 Noto Peninsula earthquake recorded by the AN-net
Keywords:The 2024 Noto-Hanto earthquake, AN-net, Seismic array analysis
The 2024 Noto Peninsula earthquake, a significant seismic event that occurred on Jan. 1, 2024, left a trail of severe damage in Noto Peninsula and its surrounding regions. The causative faults of this earthquake extend approximately 150km in a northeast-southwest direction. Because the north-eastern part of the fault is located undersea, seismic stations in Toyama and Niigata Prefectures are important to study the fault motion of this region.
Association for the Development of Earthquake Prediction (ADEP) has operated AN-net in the Nagaoka region, Niigata Prefecture, since 2009. The AN-net consists of 40 stations. At each station, a short-period sensor and an acceleration sensor are installed at the bottom of a borehole with a depth of about 100m, and an acceleration sensor is installed on the surface. The average station separation is about 5km. Strong ground motions from the Noto Peninsula earthquake were recorded by the AN-net. , and seismic waves are coherent in longer periods. So, we conduct a seismic array analysis of strong motion records from the Noto Peninsula earthquake.
Among the stations in the AN-net, we chose five stations in the Southwest and set up a sub-array. This subarray is about 120km from the mainshock hypocenter, and the fault length of about 150km corresponds to 30 degrees in the back azimuth from the array. We used records of the mainshock by surface acceleration sensors for 5 minutes starting from 16:10 on Jan. 1. We decided to use the MUSIC (MUltiple SIgnal Classification) method (Schmidt, 1986)for the array analysis because this method is capable of narrowing the beam and coping with multiple incidences of waves. Calculating the coherence of seismic waves between neighboring stations, we judge that periods longer than 5 s are available in seismic array analysis. Sliding a 20 s long time window with a step of 5 s, we calculate MUSIC spectra at the period of 6.67s (a frequency of 0.15Hz). The best slowness and back azimuth are calculated with steps of 0.05 s/km and 1 degree so as to maximize the MUSIC spectra.
According to the results, the back azimuth corresponds to the mainshock hypocenter and the northeastern part of the fault for the first 25 seconds after the arrival of a direct S wave from the mainshock. Then, the back azimuth points to the southwestern part of the fault for the following 10s. Then, finally, the back azimuth corresponds to the northeastern part of the fault for the following 10s.
In order to check the propagation path effects on the results, we conducted two additional analyses. The first one is to use a moderate-sized aftershock that is large enough to have a good signal-to-noise ratio at the AN-net but is small enough to be recognized as a point source. We used an aftershock of M5.8 that took place at 16:58 on Jan. 1. The second one used synthetic seismograms that were calculated for a point source at the mainshock hypocenter by an open-source 3D finite difference simulation code, OpenSWPC (Maeda et al., 2017). From these two additional analyses, we confirmed that the back azimuth points to the correct location for the direct wave. However, the back azimuth fluctuates for later phases, and this pattern is different from the results for the mainshock. Therefore, we judge that the temporal changes in back azimuths for the mainshock reflect the changes in the source location.
We will be able to set up one more subarray. And we can analyze aftershocks. Through these additional analyses, we will validate our current results further.
Acknowledgements
This study used OpenSWPC (Maeda et al., 2017) and computational resources of Wisteria/BDEC-01 Odyssey (the University of Tokyo). This study was partially funded by Earthquake Res., Inst., the University of Tokyo, Joint Research program 2023-S-A101.
Association for the Development of Earthquake Prediction (ADEP) has operated AN-net in the Nagaoka region, Niigata Prefecture, since 2009. The AN-net consists of 40 stations. At each station, a short-period sensor and an acceleration sensor are installed at the bottom of a borehole with a depth of about 100m, and an acceleration sensor is installed on the surface. The average station separation is about 5km. Strong ground motions from the Noto Peninsula earthquake were recorded by the AN-net. , and seismic waves are coherent in longer periods. So, we conduct a seismic array analysis of strong motion records from the Noto Peninsula earthquake.
Among the stations in the AN-net, we chose five stations in the Southwest and set up a sub-array. This subarray is about 120km from the mainshock hypocenter, and the fault length of about 150km corresponds to 30 degrees in the back azimuth from the array. We used records of the mainshock by surface acceleration sensors for 5 minutes starting from 16:10 on Jan. 1. We decided to use the MUSIC (MUltiple SIgnal Classification) method (Schmidt, 1986)for the array analysis because this method is capable of narrowing the beam and coping with multiple incidences of waves. Calculating the coherence of seismic waves between neighboring stations, we judge that periods longer than 5 s are available in seismic array analysis. Sliding a 20 s long time window with a step of 5 s, we calculate MUSIC spectra at the period of 6.67s (a frequency of 0.15Hz). The best slowness and back azimuth are calculated with steps of 0.05 s/km and 1 degree so as to maximize the MUSIC spectra.
According to the results, the back azimuth corresponds to the mainshock hypocenter and the northeastern part of the fault for the first 25 seconds after the arrival of a direct S wave from the mainshock. Then, the back azimuth points to the southwestern part of the fault for the following 10s. Then, finally, the back azimuth corresponds to the northeastern part of the fault for the following 10s.
In order to check the propagation path effects on the results, we conducted two additional analyses. The first one is to use a moderate-sized aftershock that is large enough to have a good signal-to-noise ratio at the AN-net but is small enough to be recognized as a point source. We used an aftershock of M5.8 that took place at 16:58 on Jan. 1. The second one used synthetic seismograms that were calculated for a point source at the mainshock hypocenter by an open-source 3D finite difference simulation code, OpenSWPC (Maeda et al., 2017). From these two additional analyses, we confirmed that the back azimuth points to the correct location for the direct wave. However, the back azimuth fluctuates for later phases, and this pattern is different from the results for the mainshock. Therefore, we judge that the temporal changes in back azimuths for the mainshock reflect the changes in the source location.
We will be able to set up one more subarray. And we can analyze aftershocks. Through these additional analyses, we will validate our current results further.
Acknowledgements
This study used OpenSWPC (Maeda et al., 2017) and computational resources of Wisteria/BDEC-01 Odyssey (the University of Tokyo). This study was partially funded by Earthquake Res., Inst., the University of Tokyo, Joint Research program 2023-S-A101.