5:15 PM - 7:15 PM
[SVC32-P18] High-resolution characterization of deep low-frequency earthquake activity beneath Mount Fuji using waveform classification
Keywords:Mount Fuji, deep low-frequency earthquake, waveform classification, hierarchical clustering, waveform cross-correlation, deep magma plumbing system
Deep low-frequency earthquakes (DLFs) beneath volcanoes are possible evidence of deep-seated volcanic activity. Detecting increased activity of DLFs is crucial for early disaster preparedness. Compared to other Japanese volcanoes, Mount Fuji has the highest occurrence rate of DLFs, enabling detailed analysis with high spatiotemporal resolution. A comprehensive detection revealed increased DLF activity following the 2011 Shizuoka Earthquake (Mw 5.9) in the southwestern part of the hypocenter region (Nanjo et al., 2023). However, no increase was observed during the 2008-2010 inflation event linked to deep magma reservoir inflation at the base of the DLF hypocenter region (Mitsui and Kato, 2019). Ikegaya and Yamamoto (2021) classified DLF waveforms at Zao volcano and revealed temporal correlations between the DLF groups and the volcanic activity. To gain a better understanding of deep volcanic fluid behavior, we improved their classification method and applied it to DLFs beneath Mount Fuji from 2006 to 2022.
We used three-component waveform data from seismic stations around Mount Fuji operated by NIED, JMA, and the University of Tokyo. The waveform data were extracted using (i) the JMA unified hypocenter catalog, (ii) the matched-filter detection catalog by Kurihara and Obara (2021), and (iii) the catalog of the Mount Fuji Research Institute. Following the method of Ikegaya and Yamamoto (2021), for catalog (i) including DLFs with relatively large magnitude, we calculated inter-event waveform correlations at multiple stations and performed hierarchical clustering using the nearest-neighbor method, defining distance as 1 minus the cross-correlation coefficient. As an improvement in this study, the largest waveform group was further classified using Ward method. Finally, we calculated waveform correlations between the classified events in catalog (i) and those in catalogs (ii) and (iii), assigning events with a cross-correlation coefficient more than 0.4 to existing groups.
The classification resulted in five waveform groups (G1-G5). G1, G2, and G4 showed waveforms with only low-frequency components (1-2 Hz), whereas G3 and G5 included both low-frequency-dominated waveforms and those containing higher-frequency components (3-8 Hz). Their spatial distributions were distinct: G1 was beneath the summit, G2 in the east, G3 in the east to northeast, G4 in the north to northeast, and G5 in the northeast. The hypocenter depths of all groups ranged from 10 to 20 km. Focal mechanisms estimated from S/P spectral ratios indicated isotropic components in all groups, suggesting fluid involvement, and no clear differences between groups were observed.
G2 activity began in 2008, while G4 and G5 showed increased occurrence rates after 2009 and 2011, respectively. G2 and G4 appeared linked to the inflation event, whereas G5 exhibited increase around the 2011 Shizuoka earthquake. Notably, G4 activity increased when the inflation source migrated from 20 km to 15 km depth, the depth of the DLF hypocenter region (Mitsui and Kato, 2019). In contrast, G1 activity increased in late 2006, and G3 showed a slight increase in 2009 but remained relatively steady, without clear correlations to seismic or volcanic activity at Mount Fuji. Gutenberg-Richter b-values for G3-5 showed G3 had a relatively high b-value (~2), while G4 and G5 had lower values (~1.5). Thus, the increased DLF activity in G4 and G5, located in the northern to northeastern regions of Mount Fuji, may have been driven by increased differential stress. This stress increase may be attributed to deep volcanic fluid movement linked to the inflation event and the 2011 Shizuoka earthquake.
This study indicates that waveform classification captures deep magmatic activity with high spatio-temporal resolution. Further refinement, such as clustering G3 and G5 events based on spectral characteristics, is expected to provide a better understanding of deep volcanic fluid movement.
We used three-component waveform data from seismic stations around Mount Fuji operated by NIED, JMA, and the University of Tokyo. The waveform data were extracted using (i) the JMA unified hypocenter catalog, (ii) the matched-filter detection catalog by Kurihara and Obara (2021), and (iii) the catalog of the Mount Fuji Research Institute. Following the method of Ikegaya and Yamamoto (2021), for catalog (i) including DLFs with relatively large magnitude, we calculated inter-event waveform correlations at multiple stations and performed hierarchical clustering using the nearest-neighbor method, defining distance as 1 minus the cross-correlation coefficient. As an improvement in this study, the largest waveform group was further classified using Ward method. Finally, we calculated waveform correlations between the classified events in catalog (i) and those in catalogs (ii) and (iii), assigning events with a cross-correlation coefficient more than 0.4 to existing groups.
The classification resulted in five waveform groups (G1-G5). G1, G2, and G4 showed waveforms with only low-frequency components (1-2 Hz), whereas G3 and G5 included both low-frequency-dominated waveforms and those containing higher-frequency components (3-8 Hz). Their spatial distributions were distinct: G1 was beneath the summit, G2 in the east, G3 in the east to northeast, G4 in the north to northeast, and G5 in the northeast. The hypocenter depths of all groups ranged from 10 to 20 km. Focal mechanisms estimated from S/P spectral ratios indicated isotropic components in all groups, suggesting fluid involvement, and no clear differences between groups were observed.
G2 activity began in 2008, while G4 and G5 showed increased occurrence rates after 2009 and 2011, respectively. G2 and G4 appeared linked to the inflation event, whereas G5 exhibited increase around the 2011 Shizuoka earthquake. Notably, G4 activity increased when the inflation source migrated from 20 km to 15 km depth, the depth of the DLF hypocenter region (Mitsui and Kato, 2019). In contrast, G1 activity increased in late 2006, and G3 showed a slight increase in 2009 but remained relatively steady, without clear correlations to seismic or volcanic activity at Mount Fuji. Gutenberg-Richter b-values for G3-5 showed G3 had a relatively high b-value (~2), while G4 and G5 had lower values (~1.5). Thus, the increased DLF activity in G4 and G5, located in the northern to northeastern regions of Mount Fuji, may have been driven by increased differential stress. This stress increase may be attributed to deep volcanic fluid movement linked to the inflation event and the 2011 Shizuoka earthquake.
This study indicates that waveform classification captures deep magmatic activity with high spatio-temporal resolution. Further refinement, such as clustering G3 and G5 events based on spectral characteristics, is expected to provide a better understanding of deep volcanic fluid movement.