Source Mechanisms of Volcanic Induced Seismicity - Miyakejima, Japan, 2000

Sarah E. Minson, Douglas S. Dreger, and Roland Bürgmann


Mount Oyama is a volcano on Miyakejima, Japan, part of the Izu Islands volcanic chain located south of Honshu. The most recent period of unrest at Mount Oyama began in June 2000, and this eruptive sequence included the largest recorded earthquake swarm in Japan (Japan Meteorological Agency (JMA), 2000). More than 100,000 earthquakes were recorded in the next two months (Ito and Yoshioka, 2002). The seismic activity showed complex variations in space and time (JMA, 2000), but its most notable feature is that it migrated northwest with an offshore dike intrusion. Many of these earthquakes have large non-double-couple (NDC) components which might be indicative of fluid involvement in the seismic source process. In this study, linear moment tensor inversions in two passbands were used to establish the mechanism of eighteen of these earthquakes; GPS data were used to independently invert of the mechanism of two of these earthquakes and to forward predict the observed seismograms.

Moment Tensor Inversions

The methods of Dreger et al. (2000) and Dreger and Woods (2002) were used to investigate the mechanisms of eighteen events which occurred from June 29 to July 18, 2000 (UT). These eighteen events are comprised of the twelve events which the F-net network determined to have moment magnitudes of 5.5 or larger, plus six smaller events which were chosen because the F-net moment tensors had anomalous characteristics such as large NDC components or solutions with large variances.

Linear moment tensor inversions (Pasyanos et al., 1996; Dreger et al., 2000; Dreger and Woods, 2002), were used to invert complete, three-component, broadband seismograms recorded by the F-net network (Figure 22.1). In this study, all of the events were separately inverted in two frequency ranges: 0.01 to 0.033 Hz, and 0.02 to 0.05 Hz. Two types of moment tensor inversions were used: deviatoric and full moment tensor. In deviatoric inversions, the trace of the moment tensor is assumed to be zero, which implies that there is no volume change.

The depth of each event was established by performing independent moment tensor inversions at a range of depths to find which one produces the best fit to the data, and the ability of the synthetics to fit the data was assessed by the variance reduction (VR) (for example, Dreger and Woods, 2002), where a VR of 100% implies a perfect fit between the data and the synthetic seismograms. The stability of both the full moment tensor inversions and the deviatoric moment tensor inversions was examined by use of the jackknife test, in which data from every subset of the stations used in the original inversion are used to invert for the mechanism. Another type of source mechanism investigated was comprised only of isotropic and double-couple components (ISO+DC).

GPS Inversions

Data from ten GPS stations (Figure 22.1) operated by the Geographical Survey Institute of Japan (GSI) were used to determine displacements caused by EVT3 and EVT15, the two largest earthquakes in this sequence. The station located on the Izu peninsula was used as a reference station, and GPS displacements were calculated using the coordinates of each GPS station from the day before and after each earthquake. The nonlinear inversion methods of Bürgmann et al. (1997) were used to determine slip and rupture geometry. The length and width of the fault were constrained by the empirical scaling relationships between magnitude and rupture length and width which were reported in Wells and Coppersmith (1994), and the fault determined by the GPS inversion was required to be located within 5 km of the earthquake hypocenter reported by JMA.


The majority of the events studied have large CLVD components which are consistent with the opening of a vertical crack (Figure 22.2, Figure 22.3). The results of the jackknife tests show that nearly all of the mechanisms are extremely stable. Extensional CLVD mechanisms such as the ones determined by this study have been repeatedly observed in volcanic areas and have been theorized to be the result of fluid injection (for example, Kanamori et al., 1993; Julian and Sipkin, 1985).

The full moment tensor inversions have an additional model parameter relative to the deviatoric inversions and ISO+DC grid searches. Therefore, it is necessary to determine whether the better fit of the full moment tensor mechanisms is statistically significant. The significance was examined by the use of multiple types of F-tests, but the partial F-test proposed by Helsel and Hirsch (1992) is considered to be most appropriate for this study. In the Helsel and Hirsch (1992) formulation, the F-test analyzes whether the specific parameter which was added to the model yields significant explanatory power in the presence of the other variables in the model. The results indicate that for at least twelve of these earthquakes, the moment tensors in both passbands have isotropic components which are statistically significant with more than 90% confidence, and half of the isotropic components are statistically significant with more than 99% confidence.

Figure 22.1: Study area and GPS inversion results for EVT15. A. Map showing location of F-net broadband seismograph stations and GSI GPS stations used in this study. F-net stations are denoted by triangles, and GSI stations are marked with stars. B. GPS inversion results for EVT15. Black arrows and vertical bar represent observed horizontal and vertical displacements, respectively. Clear arrows and yellow bars are predicted displacements. The fault determined by the GPS inversion is plotted as long with the seismicity reported in the Japan Meteorological Agency (JMA) catalog for the twenty-four hours following EVT15. The model fits the GPS data with a weighted residual sum of squares (WRSS) of 24.0053. C. Same as B. for the area near Miyakejima.
\epsfig{, width=8cm}\end{center}\end{figure}

CLVD components can be caused by shear slip on two fault planes, but isotropic components cannot be formed from double-couples (Julian et al., 1998). Therefore, complex shear faulting can definitely be eliminated as the source of the NDC components of the mechanisms with statistically significant isotropic components. The orientation of the CLVD components of the earthquakes relative to the strike of the inferred dike (Ito and Yoshioka, 2002; Toda et al., 2002) and the observed seismicity indicate that these mechanisms may be consistent with tensile faulting due to opening along the dike.

The GPS data for EVT15 independently predict a mechanism that is very similar to the mechanism determined by the moment tensor inversions. Furthermore, synthetic seismograms generated using the mechanism determined by the GPS inversions fit the observed seismograms with a VR of 89.9%. This is particularly noteworthy because it shows that the GPS data can predict a completely independent data set.

The results of the GPS inversions for EVT3 are less conclusive. The GPS displacements on Miyakejima are consistent with deflation related to the magma chamber that was inferred to exist beneath the island, and they do not appear to be related to EVT3. Therefore, it was necessary to introduce a deflation source beneath the island in order to fit the GPS data, and this other deformation source probably affects the GPS inversion for the slip and geometry of the earthquake fault. The GPS data for EVT3 are clearly better fit by NDC mechanisms, just as the seismic data are much better fit by a NDC mechanism. However, the synthetic seismograms generated from the NDC GPS models do not fit the observed seismograms.

Figure 22.2: Full moment tensor mechanisms for the eighteen events in this study. The mechanisms in this figure were determined by full moment tensor inversions in the 0.02 to 0.05 Hz passband.
\epsfig{, width=5cm}\end{center}\end{figure}


Moment tensor inversions for deviatoric, full moment tensor, and ISO+DC mechanisms in two frequency passbands were performed for eighteen earthquakes related to the eruption on Miyakejima, Japan in 2000. The majority of these events have large CLVD components which are consistent with the opening of a vertical crack (Figure 22.2). The orientation of these mechanisms relative to the dike intrusion and the observed seismicity indicate that these mechanisms might reflect tensile faulting related to inflation along the dike. Inversions of GPS data for EVT15 yield a slip model and fault geometry which does an excellent job of forward predicting the observed seismograms for that earthquake. However, the GPS data for EVT3 do not converge to a mechanism which fits the seismic data as well as the results of the GPS inversions for EVT15 do.


Bürgmann, R., P. Segall, M. Lisowski, and J. Svarc, Postseismic strain following the 1989 Loma Prieta earthquake from GPS and leveling measurements, J. Geophys. Res., 102(3B), 4933-4955, 1997.

Dreger, D., and B. Woods, Regional distance seismic moment tensors of nuclear explosions, Tectonophysics, 356, 139-156, 2002.

Dreger, D. S., H. Tkalcic, and M. Johnston, Dilational processes accompanying earthquakes in the Long Valley Caldera, Science, 288, 122-125, 2000.

Helsel, D. R., and R. M. Hirsch, Statistical Methods in Water Resources, Studies in Environmental Science, 49, pp. 296-299, Elsevier Science Publishers, Amsterdam, 1992.

Ito, T., and S. Yoshioka, A dike intrusion model in and around Miyakejima, Niijima, and Kozushima in 2000, Tectonophysics, 359, 171-187, 2002.

Japan Meteorological Agency, Recent seismic activity in the Miyakejima and Niijima-Kozushima region, Japan - the largest earthquake swarm ever recorded, Earth Planets Space, 52(8), i-vii, 2000.

Julian, B. R., and S. A. Sipkin, Earthquake processes in the Long Valley Caldera area, California, J. Geophys. Res., 90(B13), 11,155-11,169, 1985.

Julian, B. R., A. D. Miller, and G. R. Foulger, Non-double-couple earthquakes, 1: Theory, Rev. Geophys., 36(4), 525-549, 1998.

Kanamori, H., G. Ekstrom, A. Dziewonski, J. S. Barker, and S. A. Sipkin, Seismic radiation by magma injection: An anomalous seismic event near Tori Shima, Japan, J. Geophys. Res., 98(B4), 6511-6522, 1993.

Pasyanos, M. E., D. S. Dreger, and B. Romanowicz, Toward real-time estimation of regional moment tensors, Bull. Seismol. Soc. Am., 86(5), 1255-1269, 1996.

Toda, S., R. Stein, and T. Sagiya, Evidence from the AD 2000 Izu islands earthquake swarm that stressing rate governs seismicity, Nature, 419, 58-61, 2002.

Wells, D. L., and K. J. Coppersmith, New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement, Bull. Seismol. Soc. Am., 84(4), 974-1002, 1994.

Figure 22.3: Waveform comparisons in the 0.02 to 0.05 Hz passband. A. Full moment tensor mechanism for EVT3. This mechanism has a large positive isotropic component which is consistent with inflation due to fluid injection. B. Full moment tensor mechanism for EVT7. The Phase of the Rayleigh waves does not appear to change, which is indicative of isotropic radiation. But the transverse seismogram has large amplitudes, and SH radiation should not be produced by an isotopic component. Therefore, this earthquake may have a complex mechanism consisting of both shear slip and an isotopic component, or it may have an inclined single-force mechanism. The depth of this event is unclear from the moment tensor inversions. But all three components are fit well at shallow depths which are consistent with inferred magma chamber depths. EVT7 occurred on the same day as a caldera began to form on Miyakejima, and it may be related to caldera collapse.
\epsfig{, width=8cm}\end{center}\end{figure}

Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail:
© 2004, The Regents of the University of California