5:15 PM - 6:45 PM
[SCG40-P23] Intersource-Distance Dependence of Seismic Waveform Similarity and Validity of Repeating Earthquake Detection
Keywords:waveform correlation, repeating earthquakes, interplate earthquakes, intraslab earthquakes
Repeating earthquakes are earthquakes that repeatedly rupture the same area on a fault and are mainly found on plate boundary faults. Similarity of seismic waveforms is often used to detect repeating earthquakes. Actually, waveform similarity is frequency-dependent and tends to be high when the sources are close to each other, even if the earthquakes are not repeating earthquakes (e.g., Uchida & Burgmann, 2019; Igarashi, 2020). Still, seismic waveform similarity is useful information in the detection of repeating earthquakes.
We here investigate the relationship between the cross-correlation coefficient of seismic waveforms and the relative locations of earthquake pairs based on actual observation data. The target area is the source region of the 2003 Miyagi-oki Mw7.0 intra-slab earthquake.
First, we relocated the relative centroids of the earthquake by the Double-difference relocation method (Waldhauser & Ellsworth, 2000). The period is from 2003 to 2020, and the earthquake magnitudes are MJMA>1. The procedure is the same as that of Yoshida & Hasegawa (2018, JGR), where large amounts of arrival time difference data derived from waveform correlation were used. Based on the distance from the plate boundary in Hasegawa et al. (1994), we classified the earthquakes as interplate earthquakes (within 1 km; 1562 events) and intraslab earthquakes (deeper than that; 11468 events).
Repeating earthquakes were detected by centroid and source overlapping, using the same criteria as in Yoshida et al. (2022, JGR) for earthquakes with MJMA>2. A total of 597 earthquake pairs were determined to be repeating earthquake pairs. Of these, 595 earthquake pairs were interplate earthquakes (255 events), and only two were intraslab earthquakes (4 events). Despite the overwhelmingly large number of intraslab earthquakes as the population, most of the detected repeating earthquake pairs were interplate earthquakes, suggesting that repeating earthquakes rarely occur within the slab. Our result differs from that of Nakajima & Hasegawa (2023), who suggested that many intraslab repeating earthquakes occurred in this region.
We calculated the mean correlation coefficients of the waveforms between each earthquake pair (M>2) at three stations (TU.KSN, N.TOWH, N.FSWH). We used the time window of 30 seconds starting 1 second before the arrival of the P wave, in two frequency bands, 2-5 Hz and 5-12 Hz. Waveforms with S/N less than 3 were not used, and only the vertical component was used following Igarashi (2020). The relationship between the distances between earthquake pairs and the mean correlation coefficients showed a clear trend of the mean correlation coefficient decreasing with intersource distance (Figure). The distance decay rate is greater in the high-frequency window. In the 5-12 Hz and 2-5 Hz frequency bands, there were 4914 and 6864 pairs with average correlation coefficients exceeding 0.95, respectively. When compared at the same distance, the interplate earthquake pairs tend to have higher correlation coefficients than the intraslab earthquake pairs. This may be related to the difference in the diversity of focal mechanisms.
In the 5-12 Hz frequency band, of the 303 events with correlation coefficients exceeding 0.95, 242 were determined to be repeating earthquakes based on the centroid locations. Only 11 of 303 were intraslab earthquakes. The 61 events of 303 were not classified as repeating earthquakes. Similarly, in the 2-5 Hz frequency band, 494 events have mean correlation coefficients exceeding 0.95. The 195 events of 303 were not classified as repeating earthquakes. This indicates that even though waveform correlations at many stations are quite high, the earthquakes may not have occurred at the same location. This result is not surprising considering the situation where the source size is much smaller than the wavelength, but it is important.
In the above, we used the mean correlation coefficient of the vertical movement of the three stations. In the case where only one station was used, even if we used the three components, high correlation values were obtained for more cases, even for pairs. Additionally, in the above, we used a threshold value of 0.95 for the correlation coefficient, but there is no clear rationale for this. A higher value would yield fewer similar earthquakes, and a lower value would yield far more similar earthquakes. For example, setting the correlation coefficient threshold to 0.8 in the low-frequency condition would double the number of similar earthquakes to 1182. These indicate that careful consideration is needed when detecting repeating earthquakes.
We here investigate the relationship between the cross-correlation coefficient of seismic waveforms and the relative locations of earthquake pairs based on actual observation data. The target area is the source region of the 2003 Miyagi-oki Mw7.0 intra-slab earthquake.
First, we relocated the relative centroids of the earthquake by the Double-difference relocation method (Waldhauser & Ellsworth, 2000). The period is from 2003 to 2020, and the earthquake magnitudes are MJMA>1. The procedure is the same as that of Yoshida & Hasegawa (2018, JGR), where large amounts of arrival time difference data derived from waveform correlation were used. Based on the distance from the plate boundary in Hasegawa et al. (1994), we classified the earthquakes as interplate earthquakes (within 1 km; 1562 events) and intraslab earthquakes (deeper than that; 11468 events).
Repeating earthquakes were detected by centroid and source overlapping, using the same criteria as in Yoshida et al. (2022, JGR) for earthquakes with MJMA>2. A total of 597 earthquake pairs were determined to be repeating earthquake pairs. Of these, 595 earthquake pairs were interplate earthquakes (255 events), and only two were intraslab earthquakes (4 events). Despite the overwhelmingly large number of intraslab earthquakes as the population, most of the detected repeating earthquake pairs were interplate earthquakes, suggesting that repeating earthquakes rarely occur within the slab. Our result differs from that of Nakajima & Hasegawa (2023), who suggested that many intraslab repeating earthquakes occurred in this region.
We calculated the mean correlation coefficients of the waveforms between each earthquake pair (M>2) at three stations (TU.KSN, N.TOWH, N.FSWH). We used the time window of 30 seconds starting 1 second before the arrival of the P wave, in two frequency bands, 2-5 Hz and 5-12 Hz. Waveforms with S/N less than 3 were not used, and only the vertical component was used following Igarashi (2020). The relationship between the distances between earthquake pairs and the mean correlation coefficients showed a clear trend of the mean correlation coefficient decreasing with intersource distance (Figure). The distance decay rate is greater in the high-frequency window. In the 5-12 Hz and 2-5 Hz frequency bands, there were 4914 and 6864 pairs with average correlation coefficients exceeding 0.95, respectively. When compared at the same distance, the interplate earthquake pairs tend to have higher correlation coefficients than the intraslab earthquake pairs. This may be related to the difference in the diversity of focal mechanisms.
In the 5-12 Hz frequency band, of the 303 events with correlation coefficients exceeding 0.95, 242 were determined to be repeating earthquakes based on the centroid locations. Only 11 of 303 were intraslab earthquakes. The 61 events of 303 were not classified as repeating earthquakes. Similarly, in the 2-5 Hz frequency band, 494 events have mean correlation coefficients exceeding 0.95. The 195 events of 303 were not classified as repeating earthquakes. This indicates that even though waveform correlations at many stations are quite high, the earthquakes may not have occurred at the same location. This result is not surprising considering the situation where the source size is much smaller than the wavelength, but it is important.
In the above, we used the mean correlation coefficient of the vertical movement of the three stations. In the case where only one station was used, even if we used the three components, high correlation values were obtained for more cases, even for pairs. Additionally, in the above, we used a threshold value of 0.95 for the correlation coefficient, but there is no clear rationale for this. A higher value would yield fewer similar earthquakes, and a lower value would yield far more similar earthquakes. For example, setting the correlation coefficient threshold to 0.8 in the low-frequency condition would double the number of similar earthquakes to 1182. These indicate that careful consideration is needed when detecting repeating earthquakes.