The Absence of Remotely Triggered Seismicity in Japan

Transcription

1 The Absence of Remotely Triggered Seismicity in Japan Rebecca M. Harrington and Emily E. Brodsky Department of Earth and Space Sciences University of California, Los Angeles 595 Charles Young Drive East Los Angeles, CA

2 ABSTRACT We investigate whether remote dynamic triggering by shallow mainshocks of M w 6.7 or greater occurs on a regional scale in Japan to the extent observed in the western United States and Greece following the Denali, Hector Mine, Landers and Izmit mainshocks. Based on depth and magnitude, we consider 10 mainshocks as candidates for triggering in Japan. Earthquakes during meeting either of the following criteria are triggering mainshock candidates: (1) 6.7 M w 6.9 having depths of less than 50 km, or (2) M w 7.0 having depths of less than 100 km. All candidate mainshocks produced shaking amplitudes above 0.1 cm/s in the study region. In addition, we study the seismicity in Japan after the 2004 M w 9.0 Sumatra earthquake because this mainshock produced shaking amplitudes over the entire catalog region comparable to the other studied earthquakes. We use both catalog data and filtered waveforms to search for seismicity increases following each of the mainshocks. We find no significant regional increases within 24 hours of any of the mainshock candidates. Two localized increases in cataloged seismicity occur in Kyushu, which is the only area in the study region with onshore extension. Two factors possibly inhibit regional triggering in Japan: (1) compressional tectonics, and (2) the frequent occurrence of large mainshocks. The ambient seismicity in Japan combined with the lack of observable widespread triggering renders rate-state friction and stress corrosion improbable triggering mechanisms. Fracture unclogging as a triggering mechanism is consistent with the observations. 2

3 INTRODUCTION Remote triggering was observed following the 1992 M w 7.3 Landers, CA, the 1999 M w 7.1 Hector Mine, CA, the 1999 M w 7.4 Izmit, Turkey, and the 2002 M w 7.9 Denali, Alaska earthquakes. In close proximity to the mainshock both dynamic and static triggering mechanisms may induce failure; however, dynamic stress caused by the oscillatory motions of passing seismic waves dominate at farther distances (Hill et al., 1993; Brodsky et al., 2000; Kilb et al., 2000; Gomberg et al., 2001; Husen et al., 2004). Most of the triggered earthquakes occurred in seismically active, geothermal regions. Increased seismicity following large-scale crustal earthquakes suggests that regional seismic triggering occurs outside the aftershock zone. The smallest dynamic shaking which triggered local earthquakes exceeded 0.2 cm/s (Prejean et al., 2004) suggesting that remote triggering occurs in areas with similar geologic features experiencing similar shaking amplitudes. As a seismically active, well-instrumented, geothermal region, Japan constitutes a viable candidate for remote triggering. However, two factors differentiate Japan from the triggered areas in the western US and Greece in the following ways: (1) A high seismicity rate and the Gutenberg-Richter relation dictate that large earthquakes (i.e. M w > 6.5) occur more frequently in Japan. (2) Japan lies on a convergent plate boundary and consequently has more localized compressional stress than much of the western US (Kato et al., 1998, Townend and Zoback, 2004). This has two subsequent physical implications: (a) Japan experiences more locally and regionally compressive stresses, and (b) radiation of mainshock seismic energy comes more frequently from thrust-style mainshocks rather than the strike-slip radiation occurring in the western US and Greece. 3

4 This study examines whether remote dynamic triggering occurs in Japan on a regional scale at distances farther than the conventional aftershock zone, and at levels comparable with the documented examples of the western US and Greece. Comparing the presence or lack of regional remote triggering in Japan with other triggered regions as well as comparing the geophysical factors common among regions, should help constrain viable physical mechanisms for remote dynamic triggering. We consider mainshocks between January 1997 through December 2004 which produced shaking amplitudes in the JMA catalog region comparable to shaking amplitudes produced by distant mainshocks at triggered sites in the western US and Greece (Table 1), and compare the regional seismicity in Japan with the western US and Greece to determine whether remote dynamic triggering occurs in Japan at comparable levels. The seismicity between the two regions is compared in the following ways: (1) searching the JMA catalog for statistically significant increases in the background seismicity rate, and (2) searching filtered waveforms from the Hi-net network for local, high-frequency earthquakes which may have been obscured by the mainshock coda and consequently missed by the JMA catalog. First, we discuss the methods used to compare seismicity between Japan and the western US and Greece using catalog data and waveforms and present our observations. Following the methods and observations section, we discuss what physical mechanisms make Japan different from the western US and Greece and how this relates to our observations, as well as the implications these physical differences have for possible triggering mechanisms. 4

5 METHODS AND OBSERVATIONS All other documented examples of triggering have occurred following shallow earthquakes near magnitude 7 or greater, so we examine the seismicity occurring close in time to shallow earthquakes with M w of 6.7 or greater according to the Harvard CMT catalog. We examine earthquakes of magnitude from 6.7 to 6.9 with depths less than 50 km, and magnitude 7 or greater with depths less than 100 km. Mainshocks are obtained by searching the JMA and the Harvard CMT catalogs from 1997 to 2005 for such events between latitudes of 20 N to 50 N and 120 E and 150 E with the previously mentioned depth and magnitude specifications. Although the Sumatra mainshock occurs at a teleseismic distance, it produces shaking amplitudes of roughly 0.2 cm/s or greater over the entire study region. Therefore, Sumatra is included as a mainshock triggering candidate because the shaking amplitudes compare with those regional mainshocks considered in the study. Table 1 lists the earthquakes considered as possible mainshock candidates for triggering. Catalog Study Using the JMA catalog, we look for a regional change in cataloged seismicity immediately following the mainshock triggering candidates. Table 1 lists 10 mainshocks around which seismicity at depths less than 33 km is examined. By observing the occurrence (or lack) of triggering and considering the influence of geophysical factors present, we hope to constrain a remote triggering mechanism; therefore, comparing similar types of seismicity is important. Including seismicity only above 33 km avoids comparing the shallow, crustal seismicity in Greece and the western US with deep, slab events in Japan. We compare the cataloged seismicity over the entire region with the 5

6 cataloged seismicity over the regions of previous observed triggering, and quantitatively evaluate the seismicity changes in Japan by calculating the statistical significance of the events occurring after the mainshock. Restricting our study to the intermediate and farfield as discussed in the introduction requires not counting any earthquakes within two fault-lengths of the mainshock epicenter. For a penny-shaped crack with seismic moment M 0, M 0 = 16/7 σa 3, (1) where a is the radius and σ is the stress drop. We assume that 2a = L, the fault length, and use a value of 3 MPa for the stress drop σ (Sholz, 1990). A standard Gutenberg- Richter relation, assuming a b value of 1, indicates that the JMA catalog completeness includes M j 2. Based on the catalog completeness test, included events have M j 2. Probability distribution functions characterize seismicity and quantify whether the seismicity level for a given time interval is statistically anomalous (e.g. Matthews and Reasenberg, 1988). Based on this reasoning, we evaluate the statistical significance of the catalog data by calculating the probability distribution function of the daily seismicity, and determining the cumulative probability of n earthquakes or less occurring in a given day. The probability that more than n earthquakes will occur in a given day is 1 minus the cumulative probability. Should the probability of n or greater earthquakes occurring be less than 5% according to our data, this measure indicates that there is a 95% chance that the observed number of earthquakes n was anomalous. We define the term statistically significant as a 95% chance that the number of events was anomalous. A few events require special consideration. The May 1998 earthquake coincided with a volcanic swarm on the eastern Izu peninsula that commenced before the 6

7 mainshock, and gradually increased in intensity. Without this volcanic swarm and two others like it in August 2000 at Miyake-jima, and March 2000 at Mt. Usu, the maximum number of events observed on any day of a mainshock in Table 1 is 64 (observed on May 3, 1998). The May 3, 1998 value corresponds to an 84% level of significance. For reference, the mean level of seismicity is 44 events (Figure 1). There are no statistically significant increases in seismicity beginning on the same day as any of the mainshocks in Table 1. However, a statistically significant seismicity peak above background occurs on the second and third day following the 3/28/2000 Volcano Islands mainshock (mainshock (g) of Table 1). The observed increase results from a swarm occurring in the Ibusuki Volcano field in southern Kyushu. The probability calculated from the catalog data indicates that such swarms occur in this volcano field 0.8% of the catalog days, i.e. a small, but non-negligible percentage. A local catalog supplied from the Fukuoka observatory shows that the swarm begins at 30 March 15:00:00 local time, nearly 43 hours after the mainshock. Because the microearthquakes did not commence with the incident mainshock shaking, we cannot definitively link either of these swarms to dynamic triggering by mainshock surface waves. The Ibusuki swarm is the only example of a potentially triggered increase in seismicity at a high enough level to be seen in the regional catalog. The chance of finding an increase somewhere in Japan in some interval increases as the number of intervals examined increases. Searching 10 times for levels of seismicity which occur with a probability of 5% will produce an anomalous level 40% of the time (Blom, 1989). Although the catalog only records significant seismicity on a regional scale delayed more than one day after the mainshock, the seismicity level increases 7

8 significantly on the day of the mainshock when restricting the spatial region to include only Kyushu. The catalog lists 35 earthquakes in Japan above magnitude 2 on the day of the Sumatra mainshock, a seismicity level corresponding to 38% significance (Figure 1b). Figure 1b reveals an insignificant peak in cataloged seismicity evident on the day of the mainshock, however, closer examination indicates that nearly 1/3 of the seismicity (11 earthquakes) occurs in Kyushu alone (Figures 1b). The same seismicity plot for Kyushu reveals a 95% significant peak in seismicity on the day of the mainshock, according to a probability distribution of the daily seismicity in Kyushu (Figure 3). Next, we compare cataloged seismicity in Japan with the western US and Greece. All of the mainshocks in the western US and Greece produced anomalous seismicity increases above the 95% level on the day of the mainshock corresponding to 72 events or more in the western US, and 16 events in Greece except for Hector Mine, CA which produced an increase at the 86% level (corresponding to 32 events). However, seismicity did increase significantly above the 95% level on the second day after the mainshock (72 events). Therefore, the regional cataloged seismicity levels between the western US and Greece with Japan indicate that none of the mainshocks in Japan produce a regional, cataloged increase in seismicity at a comparable level (Figures 1 (a) and (b)). The 9/25/2003 Tokachi Oki mainshock, while larger and closer to the catalog region than Denali, produces no noticeable increase in seismicity above the completeness threshold of the catalog (Figure 1b). The fact that shaking amplitudes following the 9/25/2003 mainshock well exceed those which trigger seismicity in the western US, and because triggered earthquakes are often small in magnitude warrants checking the seismicity levels below the completeness threshold as well. A histogram of the 8

9 seismicity above M j > 1 for the 2003 Tokachi Oki (offshore of Hokkaido) earthquake indicates an increase in seismicity the day following the mainshock (figure 2). There is an increase of significance from 5% to 16 % after including events with M j 1. Relaxing the standard of only including events above the completeness threshold shows a slight increase in regional seismicity, however, not at any significant level as observed in US and Greece. Waveform Study In previous triggering studies, other authors found additional uncataloged, triggered events hidden in the mainshock coda by using a high-pass filter to remove the surface shaking of the mainshock. For example, following Denali, filtered waveforms reveal events in places such as Mount Rainier, WA, the Geysers, CA, and Long Valley, CA locations in the western US located thousands of kilometers from the mainshock (Gomberg et al., 2004; Husker and Brodsky, 2004; Prejean et al., 2004). We use Hi-net and F-net waveform data to search for such events in the waveforms of stations in Japan experiencing similar shaking amplitudes ( roughly > 0.2 cm/s) at distances greater than two fault lengths for the hour before and after the mainshock, using a bandpass filter from 5 to 20 Hz. Only five of the mainshock candidates produce shaking amplitudes exceeding 0.2 cm/s over any significant portion of the study area in Japan: M w 9.0, 12/26/2004; M w 8.3, 9/25/2003; M w 7.0, 5/26/2003; M w 6.8, 3/24/2001; M w 6.8, 10/6/2000. We examine the seismograms of stations at distances greater than two fault lengths for evidence of local events following these mainshocks in order to minimize any ambiguity as to whether events at this distance could be considered part of a regular nearfield aftershock sequence. The Antelope Seismic System software performed picking, 9

10 associating and filtering (highpass of 5 Hz), of all three components of 687 stations in the Hi-net network (with sample rates of 100 Hz) following the 12/26/2004 Sumatra and the 9/25/2003 Tokachi-Oki earthquakes. Following Tokachi-Oki there were no additional events outside of the catalog. For Sumatra, there were two individual events in eastern and western Honshu not located by the catalog, but an estimation of magnitude, using P-S arrival time and amplitude, revealed that the magnitudes were below the catalog completeness threshold. Additionally, a burst of microseismicity occurs in the filtered waveforms of Hi-net station YKW which commences where the Rayleigh has its peak amplitude shaking. Additionally we searched by hand 27 F-net stations (with sample rates of 80 Hz) for the remaining mainshocks, and found no indication of local seismicity increases. DISCUSSION Two notable geophysical factors differentiate Japan from the western US and Greece, and likely affect the physical mechanisms involved in dynamic triggering. (1) Japan experiences a higher level of seismicity, which might affect mechanisms such as fracture unclogging. (2) Only the western half of Kyushu occupies a tectonically extensional regime, unlike much of the region in the western US and Greece. Shaking amplitudes which exceed 0.2 cm/s by an order of magnitude occurred over most of Japan following the two largest candidate mainshocks, Tokachi Oki and Sumatra. We note that areas in eastern Honshu experiencing the largest shaking amplitudes from the Sumatra mainshock are populated with geothermal fields; however, we observe a lack of dynamic triggering in these areas (figure 4). Given that these regions concurrently experience tectonic compression might suggest that tectonic extension encourages dynamic 10

11 triggering. The next section discusses the geophysical factors which differentiate Japan from the western US and Greece, and their implications for possible mechanisms of remote dynamic triggering. High Ambient Seismicity One physical difference between Japan and the western US and Greece, is the number of mainshocks over M w > 7 that occur. From January 1977 to January 2005 Harvard CMT Catalog lists 30 events (of M w > 7, at depths less than 100 km) in Japan, and lists only six in spatial region of the same size in the western US. Brodsky and Prejean (2004) suggest that remote dynamic triggering requires large amplitude, lowfrequency shaking resulting from mainshocks of M w 7 or greater. Japan more frequently experiences mainshocks over M w 7 than the western US and Greece, and consequently much of the region experiences large-amplitude, low-frequency shaking more often. The high seismicity rate has implications for many mechanisms. Previous observations of remote dynamic triggering suggest that triggering occurs often in geothermal regions. Brodsky et al. (2003) suggest that the network of clogged fractures within saturated hydrological units coupled with the constant precipitation of material into fractures from geothermal fluids make these regions susceptible to triggering effects via fracture unclogging from the dynamic stresses from the waves of distant earthquakes. In a region with more frequent large earthquakes, constant shaking activity should constantly disturb the fragile blockages in fractures and faults, preventing the formation of long-term obstacles to fluid flow. Without a means to accumulate blockages, there is no chance to build or maintain a differential pressure between hydrological units, eliminating the opportunity to trigger seismicity in occasional bursts 11

12 when blockages are released. Given the occurrence rate of large mainshocks in Japan versus the western US, one can say that Japan experiences a higher level of seismicity. If one assumes that fracture unclogging is the mechanism for dynamic triggering, one might reason that seismicity rate makes Japan less susceptible to remote dynamic triggering than the western US and Greece. Rate and state friction has also been suggested as a triggering mechanism, however, a high ambient seismicity is inconsistent with this model. According to Dieterich (1994), seismicity rate after a perturbance, R, is proportional to steady state seismicity, r. If stressing rate is assumed constant, then seismicity rate as a function of stress step is directly proportional to the background rate, and should be high given a high background rate. For instance, the instantaneous rate-change at the time of a stress step τ is predicted to be: τ R = r exp( ). (2) Ασ A is a material property, and σ is the normal stress, which is assumed constant (Dieterich, 1994). Given equation (2), we would expect to see large rate changes in seismicity given a high background rate, contrary to what we observe. Finally, if one makes the usual assumption that the seismicity is proportional to the stressing rate, then one can apply the equations of stress corrosion. Brodsky and Prejean (2004) show that the equations governing stress corrosion or sub-critical crack growth are identical to the governing equation for evolution of faults with rate-state stress in the commonly applied limits. Therefore, given the common assumptions about seismicity and stressing rate are true, we conclude that in the context of both rate-state friction, and consequently sub-critical crack growth models, the lack of remote dynamic 12

13 triggering in regions of high ambient seismicity is described more effectively by fracture unclogging. Tectonic Regime Unlike the western US and Greece, Japan lies on a convergent plate boundary. GPS data indicates that all of the Japanese islands with the exception of Kyushu occupy a dominantly compressional tectonic regime (see Kato et al. (1998), Figures 3 and 5). The same data indicates that the strain in the western half of Kyushu is dominantly extensional. Compression vs. Extension Much of the triggered seismicity in the western US and Greece occurs in locally extensional areas. This raises the question of whether remote dynamic triggering requires a locally extensional regime. Assume a model consistent with Anderson s theory of faulting, and consider a principle compressive stress oriented vertically in a region of normal faulting and horizontally in a region of thrust faulting. The Mohr circle diagram for this scenario would indicate that a thrust fault requires a higher differential stress to rupture. Keeping this scenario in mind, one might expect remote dynamic triggering to occur more readily in regions of localized extension, where faults are weaker. For example, we might expect to see variable changes in seismicity between Kyushu and the rest of Japan, because the western half of Kyushu comprises the only extensional region in our study area. If Kyushu triggers more readily than the rest of the study area, then it is possible that Kyushu experiences increases in seismicity after the mainshocks but the seismicity of the rest of Japan dwarfs the signal. 13

14 The catalog indicates that seismicity in Kyushu increases at a significance level of 95.1% (corresponding to 11 events) on the day of the Sumatra mainshock (Figure 3) which had the largest shaking in the region (Figure 4). The limited regional extent of the extensional stress regime makes it difficult to measure the seismicity in the same meaningful way given the limited data points, which consequently limits the statistical argument in Kyushu for remote dynamic triggering. The statistical significance of the seismicity increase is marginal compared with the increases observed in the western US. Some localized triggered areas in the western US experienced 60+ earthquakes following Denali. The most analogous tectonic area to Japan in the western US is the Cascades. Prejean et al., (2004) report triggering at Mount Rainier following the 2003 Denali earthquake. Moran et al. point out that many of the earthquakes occurring at Mount Rainer from 1989 to 1997, which have a sufficient number of polarities to calculate focal mechanisms, have normal to oblique-normal focal mechanisms (2000). Therefore, although Mount Rainier is located in a regionally compressional regime, the triggering occurred in a locally extensional regime. The fact that no known triggering occurs in any tectonically compressional locations in Japan, coupled with the fact that observed triggering in the western US occurs in locally extensional locations, suggests that remote dynamic triggering may require an extensional environment. Mainshock Focal Mechanism All of the mainshocks which triggered seismicity in the western US and Greece had mainly strike slip motion (Table 2). Both the Sumatra and Tokachi Oki mainshocks, which produced the largest amplitudes shaking over the entire study area, are thrust events which produces no significant increase in seismicity with our completeness 14

15 threshold and fault length standards imposed (or otherwise). In contrast, the Denali earthquake is a strike-slip event, which triggered seismicity thousands of kilometers away. Strike-slip earthquakes may radiate more energy than thrust-style events (Choy and Boatright, 1995). If the dominantly thrust-style mainshocks in Japan radiate less energy than strike-slip earthquakes of similar magnitude in the western US, then it is possible that thrust-style faulting may be a physical factor responsible for the lack of observed triggering in Japan. However, the focal mechanism is not important if the waveforms indicate shaking occurs at comparable levels to the western US. Figure 4 indicates that the raw shaking amplitudes from Sumatra over the entire study area are in fact comparable with those observed to cause triggering in the western US at locations like the Geysers, Long Valley, and Coso Geothermal field (Brodsky and Prejean, in press; Prejean et al., 2004). It is possible that shaking only in some particular frequency range is important for triggering and therefore the comparison of raw amplitudes in Fig. 4 is too restrictive. Table 3 lists the mainshock shaking amplitudes over a range of frequency bandwidths typically observed at various stations in Japan at distances outside of five fault lengths from the mainshock epicenter. The table indicates that over all frequency ranges, shaking in Japan is larger than or comparable to amplitudes observed in the western US. Therefore, nothing suggests that the thrust-style mainshocks in Japan should trigger seismicity any less effectively than strike-slip mainshocks in the western US, given that shaking amplitudes are comparable over much of Japan to triggered sites in the western US. 15

16 CONCLUSIONS Our study investigates whether regional remote dynamic triggering occurs in Japan by examining cataloged seismicity and waveforms from the period of January, 1997 through January, There are 10 mainshocks between 1997 and 2005 at shallow depths which produce shaking amplitudes on the order of, or greater than the mainshocks for which others previously observed examples of triggering. Ukawa et al. (2002) report triggering of micro-earthquakes in the coda of regional mainshocks at Iwojima volcano. Additionally, Miyazawa and Mori (2004) observe triggering of deep lowfrequency events (DLF s) at regional distances, and observe shallow, crustal triggering primarily in regions within our two-fault length approximation. We recognize that remote, dynamic triggering may existent in Japan, but catalog and waveform data indicate that it does not occur at comparable levels as observed in the western United States and Greece following mainshocks with M w 7. Two factors, which may work either separately or in combination, differentiate Japan from other triggered locations and might influence physical triggering mechanisms: (1) the high relative seismicity, i.e. the more frequent M w > 7 earthquakes occurring in the region, and (2) tectonic extension. The lack of extensive triggering in geothermal regions is consistent with fracture unclogging as a triggering mechanism because the faults and fractures may be consistently unclogged in the presence of frequent large-amplitude, low-frequency shaking from large mainshocks. In this scenario the high seismicity rate prevents the occurrence of remote dynamic triggering on a regional scale in Japan to the extent observed in the western US and Greece following mainshocks which produce comparable dynamic stresses at comparable distances. The marginally significant triggering in 16

17 Kyushu alternatively indicates the importance of tectonic extension encouraging triggering. We conclude that either, or possibly both, fracture unclogging and wideranging compressional tectonics inhibit remote dynamic triggering, and that triggering does not occur in Japan to the extent observed in the western United States and Greece. ACKNOWLEDGEMENTS We used seismic data from the Fukuoka Observatory, the Japan Meteorological Agency, and the National Research Institute for Earth science and Disaster Prevention, the Hi-net and F-net seismic networks. The construction and the maintenance of the Hinet and F-net are supported by all the staff in the National Disaster Information Center and the Solid Earth Research Group, NIED, as well as the help from a number or related companies. Thanks to Osamu Nakagome from the Fukuoka Observatory for supplying additional catalog data. We thank H. Kanamori and S. Prejean for instructive comments on an early draft. AUTHOR AFFILIATION Rebecca M. Harrington Graduate Student UCLA Department of Earth and Space Sciences 595 Charles Young Drive East Los Angeles, CA Phone: (310) Fax: (310) rebecca@moho.ess.ucla.edu Emily Brodsky Assistant Professor of Geophyisics UCLA Department of Earth and Space Sciences 595 Charles Young Drive East Los Angeles, CA Phone: (310) Fax: (310) brodsky@ess.ucla.edu 17

18 REFERENCES Atkinson, B. K., 1984, Subcritical Crack Growth in Geological Materials, J. Geophys. Res., v. 89, no. B6, p Blom, G., 1989, Probability and Statistics: Theory and Applications, New York: Springer-Verlag. Brodsky, E. E. and S. G. Prejean, New Constraints on Mechanisms of Remotely Triggered Seismicity at Long Valley Caldera, J. of Geophys. Res., in press. Brodsky, E. E., E. Roeloffs, D. Woodcock, I. Gall and M. Manga, 2003, A Mechanism For Sustained Ground Water Pressure Changes Induced by Distant Earthquakes: J. Geophys. Res. Vol. 108, No. B8, 2390, doi: /2002JB Brodsky, E. E., Karakostas, V., and Kanamori, H., 2000, A New Observation of Dynamically Triggered Regional Seismicity: Earthquakes in Greece Following the August 1999 Izmit, Turkey Earthquake. Geophys. Res. Let., v. 27, no. 17, p Brodsky, E. E., Sturtevant, B., Kanamori, H., 1998, Earthquakes, Volcanoes, and Rectified Diffusion. J. Geophys. Res., v. 103, no. B10, p. 23, Choy, G. L., Boatwright, J. L., 1995 Global Patterns of Radiated Seimic Energy and Apparent Stress. J. Geophys. Res., v. 100, no. B9, p Dieterich, J., 1994, A Constitutive Law for Rate of Earthquake Production and its Application to Earthquake Clustering. J. Geophys. Res. v. 99, no. B2, p Global Volcanism Program, Smithsonian National Museum of National History, 18

19 Gomberg, J., P. Bodin, K. Larsen and H. Dragert, 2004, Earthquake Nucleation by Transient Deformations Caused by the M = 7.9 Denali, Alaska, Earthquake, Nature, v. 427, p Gomberg, J., Reasenberg, P. A., Bodin, P., and Harris, R. A., 2001, Earthquake Triggering by Seismic Waves Following the Landers and Hector Mine Earthquakes. Nature, v. 411, p Hill, D. P., et al., 1993, Seismicity Remotely Triggered by the Magnitude 7.3 Landers, California, Earthquake. Science, v. 260, p Husen S., Taylor R., Smith R.B, and H. Healser, 2004, Changes in Geyser Eruption Behavior and Remotely Triggered Seismicity inyellowstone Natinal Park Produced by the 2002 M 7.9 Denali Fault Earthquake, Alaska. Geology, v. 32, no. 6, p Husker, A.L and E. E. Brodsky, 2004, Triggered Seismicity in Idaho and Montana: A Window into the Geologic Context for Seismic Triggering, BSSA, v. 94, S310- S316. Agency of Natural Resources and Energy and Geological Survey of Japan, Ministry of International Trade and Industry, 1976, Index to the Geothermal Field of Japan. Kato, T., El-Fiky, G. S., and Oware, E. N., 1998, Crustal Strains in the Japanese Islands as Deduced from Dense GPS Array. Geophys. Res. Let., v. 25, n. 18, p Kilb, D., Gomberg, J., Bodin, P., 2000, Triggering of Earthquake Aftershocks by Dynamic Stresses. Nature, v. 408, p Lay, T. and Wallace, T. C., 1995, Modern Global Seismology. San Diego: Academic 19

20 Press. Matthews, M. V., Reasenberg, P. A., 1998, Statistical Methods for Investigation Quiescence and Other Temporal Seismicity Patterns. Pure and Applied Geophysics, v. 126, n.24. Miyazaki, S, and Heki, K., 2001, Crustal Velocity Field of Southwest Japan: Subduction and Arc-arc Collision. J. of Geophys, Res., v. 106, no. B3, p Miyazawa, M., and Mori, J. J., 2003, Active Response of the Structure Beneath Japan due to Static and Dynamic Stress Changes From Large Earthquakes in Annuals of Disaster Prev. Res. Inst., Kyoto Univ., n. 47 C. Moran, S. C., Zimbelman, D. R., Malone, S. D., 2000, A Model for the Magmatic Hydrothermal System at Mount Rainier, Washington, from Seismic and Geochemical Observations. Bull. Volc., v. 61, p Prejean et al., 2004, Remotely Triggered Seismicity on the United States West Coast Following the M w 7.9 Denali Fault Earthquake. BSSA, v. 94, p. S348-S359. Scholz, C. H., 1990, The Mechanics of Earthquakes and Faulting, 2 nd ed., Cambridge, UK: Cambridge Univ. Press Stein, R. S., 1999, The Role of Stress Transfer in Earthquake Occurrence. Nature, v.402, p Townend, J., Zoback, M. D., 2004, Regional Tectonic Stress near the San Andreas Fault in Central and Southern California. Geophy. Res. Let., v. 31, L15S11. Ukawa, M., Fujita, E., Kumagai, T., 2002 Remote Triggering of Microearthquakes at the Iwo-jima Volcano. J. Geography, v. 111, p

21 Mainshock Latitude Longitude Depth (km) Plunge T/B axes (Harvard M w (Harvard CMT) CMT) (a) / /26/2004 (b) / /28/2004 (c) / /25/2003 (d) / /26/2003 (e) / /24/2001 (f) / /6/2000 (g) / /4/2000 (h) / /28/2000 (i) / /20/1999 (j) 5/3/ / Table 1. Mainshocks considered as candidates to possibly have caused remote dynamic triggering in Japan. To focus on mainshocks with large surface shaking, we consider those with 6.7 Mw < 7 with depths < 50 km, and Mw > 7 with depths < 100 km. All of these mainshocks produced shaking amplitudes in the catalog area exceeding 0.2 cm/s the lower amplitude threshold of shaking observed with triggered seismicity in the Western United States (Prejean et al, 2004). 21

22 Mainshock Latitude Longitude Depth (km) Plunge T/B axes 11/3/ / /16/ / /17/ / /28/ / M w Table 2. Mainshocks which dynamically triggered remote seismicity in the Western United States and Greece. All of the mainshocks have steeply plunging B axes, suggesting strike-slip earthquakes. In contrast, the Japan mainshocks are mostly thrust events with shallowly plunging B axes. 22

23 > 5 Hz Hz < 0.1 Hz (a) 12/26/2004 (c) 9/25/2003 (e) 3/24/2001 (f) 10/06/2000 NEW (Washington) HLID (Haley, ID) cm/s 0.7 cm/s cm/s 0.6 cm/s 0.7 cm/s 3.5 cm/s 0.2 cm/s 0.1 cm/s 3.7 cm/s 0.2 cm/s - 2 cm/s* 1.5 cm/s*.01 cm/s* 0.6 cm/s* 0.5 cm/s* Table 3. Maximum horizontal shaking amplitudes in different bandpasses in Japan and at two triggered sites in the Western US. Stations in Japan are located outside of five fault lengths as defined in the text. Shaking amplitudes for stations in the US result from the Denali mainshock. Seismic amplitudes from mainshocks in Japan are comparable in all frequency bands to amplitudes in the western US which triggered seismicity, however, no triggering is observed in Japan. * Record off-scale for E components, so the amplitude indicated is a lower limit. 23

24 Figure 1. (a) Histogram of the cataloged seismicity in the western United States and Greece. We separated the events in 24-hour time bins, with the mainshock set at time zero. We count earthquakes beyond beyond two fault lengths (defined by equation 1) of epicenter, and above M j 2. (b) shows the same plot for the same sized region in Japan. Green dashed lines represent seismicity including unrelated volcanic swarms (see text). 24

25 Figure 2. Histogram of seismicity in Japan 10 days before and after the 9/25/2003, Mw 8.3 Tokachi Oki mainshock. The dashed histogram indicates events that are M j > 1, and the solid histogram with asterisks indicates event of M j > 2. All events lay outside of 320 km (two fault lengths) from the epicenter. 25

26 Figure 3. Cataloged seismicity within the circle on Figure 4 of M j > 2 in Kyushu, the region in Japan with notable onshore extensional strain (Kato et al., 1998). A statistically significant seismicity increase of 95.1% occurs on the day of the 12/26/2004 Sumtra mainshock (correspondsing to 11 events). A swarm of seismicity occurs at the Ibusuki Volcano field the second day after the March 2000 mainshock see text for discussion. 26

27 Figure 4. Contour plot of shaking amplitudes for the 12/26/2004 Sumatra mainshock, taken from the 687-station Hi-net borehole network. Contour amplitudes are in cm/s. Transverse and vertical components exhibit similar shaking amplitudes. 27