2003 AGU 2003 Fall Meeting
San Francisco, California December 8-12 , 2003
Source Process of the 1999 Chi-Chi, Taiwan Earthquake: Comparisons to Shallow Faulting in Subduction Zones
The 1999 Chi-Chi, Taiwan earthquake (Mw 7.6) occurred along a zone of rapid convergence in central Taiwan. Since there were large amounts of slip at shallow depths on the Chelungpu fault (8 to 10 meters of thrust faulting at depths of a few kilometers to the surface), the earthquake provides a first-hand opportunity to study many of the physical properties associated with large slip on a fault. There has been a large variety of geological and geophysical observations of this event, including, a dense network of strong-motion instruments that recorded the earthquake, GPS observations of the pre-, co- and post-seismic deformation rates, borehole sampling into the fault, and surface geology studies. The combination of these studies produces one of the most complete pictures of the source process of a large thrust earthquake. Especially, the detailed imaging of the area of large slip in the northern part of the fault provides information about the fault dynamics that may control the large amounts of slip on a fault. The faulting process appears very different between the large slip areas in the north and the more moderate slip regions in the south. The large slip regions in the north appeared to be associated with 'smooth' rupture, that generated relatively low levels of high-frequency radiation and caused relatively low levels of damage. This is in contrast to the more typical high-frequency radiation produced by the fault in the south, which caused considerably more damage, even though the amounts of fault slip were less. The rather unusual characteristic of the large slip in the north may be associated with low levels of dynamic friction on the fault. The thrust faulting of the Chi-Chi earthquake may be similar to the shallow faulting in subduction zones, such as the Nankai Trough. The geometry of the Chelungpu fault, which is connected to more shallowly dipping fault structures at depth, is similar to the splay faults that are mapped in the Nankai subduction zone. For the Nankai trough, it is unclear if slip on the splay faults occurs during the great subduction earthquakes or in separate earthquakes. Part of the answer to this question may come from the results of current projects, such as the borehole sampling of the Nankai splay faults and comparisons with similar borehole samplings of the Chelungpu fault.
Initiation Process of Moderate Earthquakes in the Northern Miyagi, Japan, Earthquake Sequence.
Sato, K. and J. Mori
Scaling relationships between the durations of the beginning portion of P-waves and final sizes of earthquakes have been reported by several studies (e.g. Iio ). These studies mainly targeted microearthquakes and moderate to large earthquakes have been analyzed in only few studies. We tried to determine whether such scaling relationships hold for larger events, because it especially important for understanding the initiation processes of large earthquakes. We analyzed several moderate (magnitude up to 5) earthquakes and two large (magnitude 6 to 7) earthquakes and concluded that there may be a scaling relationship for small to moderate events but the scaling probably does not hold for the larger events. An earthquake sequence including several moderate earthquakes (magnitude up to 6.2) began on July 26, 2003 (JST) in the Northern Miyagi region, Japan. To study characteristics of initial phases of the moderate earthquakes, we analyze 4 $<$ M $<$ 6.2 events in the sequence. In this study, we employ the method which we used in the past study, a source model of Sato and Kanamori . The initial rupture of an earthquake is described with two parameters in this model: trigger factor and initial crack size. The trigger factor represents the contrast of surface energy on the edge of a crack and off of the crack. A large value for the trigger factor means that very high surface energy exists only on the edge and smaller surface energy off of the edge and rupture propagates spontaneously. A smaller value means that similar values of surface energy exists on and off of the edge and rupture propagates gradually depending on the initial crack size. We can use clearly recorded data from several near source borehole stations which are operated by Hi-net, NIED. We analyze 11 events with magnitude between 4.0 and 6.2. Hypocenter locations determined by the Japan Meteorological Agency and focal mechanisms determined by F-net, NIED are used. For these events, we obtain the results that all events can be explained by a small trigger factor and their initial crack sizes are between 15 and 50m. We can see a scaling relationship that the largest event (M6.2) has the largest initial crack size. However, we can also see that similarly sized events began from various initial cracks.
TI: Determination of Fault Planes and Rupture Velocities of Small Earthquakes in a South African Gold Mine: Constraints on Radiation Efficiency
Yamada, T., J. Mori, H. Ogasawara, Y. Iio, H. Kawakata, S. Ide
Analyses of source processes of small earthquakes are important for investigating whether or not there are dynamic differences between small and large earthquakes. However, it is difficult to resolve details of the source of small earthquakes because close station spacing near the hypocenter and data with high sampling rates are necessary. Such observations of mining induced earthquakes are being carried out in a South African gold mine. Nine tri-axial borehole accelerometers were installed within 200 m along a 2,650-m-deep haulage tunnel in the Mponeng gold mine. Many seismic events (-2.7 $<$ M $<$ 3.3) were recorded with a sampling frequency of 15 kHz from February to October, 1996. In this study we focused on the rupture velocity, which is important for investigating characteristics of initiations, arresting mechanisms, and radiation efficiency of earthquakes. We carried out kinematic wave-form inversions for three larger events (M1.4, 1.1, and 0.8) that occurred within 200 m of the stations. First, we determined the velocity structure using arrival time data. Velocities of P and S waves were estimated to be 6.00 km/s and 3.83 km/s, respectively. Next, we determined focal mechanisms from amplitudes of P, SH, and SV waves. Finally we carried out kinematic wave-form inversions for both nodal planes of focal mechanisms assuming various rupture velocities to distinguish the fault plane and the best-fitting rupture velocity. We could determine the fault plane for all events because the model fit to the data for the fault plane was significantly better than for the auxiliary plane. Also, the deduced fault planes were consistent with the results of sub-event locations for these events by Yamada et al. (2002). On the other hand, we could not determine the rupture velocities with complete confidence. One problem is that in general, residuals are likely to be smaller by assuming higher rupture velocities. However, we can conclude that rupture velocities were not less than 50% of the S-wave velocity, on the basis that the slower rupture velocities could not explain wave-forms very well. Therefore, we conclude that rupture velocities of small earthquakes in the South African gold mine are almost the same as those of larger natural earthquakes. The radiation efficiency can be written as a function of the rupture velocity and becomes greater with increase of the rupture velocity. This study indicates that radiation efficiencies of small earthquakes in the South African gold mine are almost equal to those of larger natural earthquakes.
Rupture Areas of Large Earthquakes In Papua New Guinea in 1971 and 2000
Park, S. and J. Mori
Two large earthquakes with magnitudes greater than 7.0 (Mw) occurred near New Ireland and New Britain, Papua New Guinea, on November 16-17, 2000. Both earthquakes were thrust events with shallow dip angles in the New Britain subduction zone. Also, another two large earthquakes (both Ms 7.9) occurred in the same area on July 14 and July 26 in 1971, almost 30 years before the 2000 events. Aftershock zones of the 1971 events covered large areas, including the fault zones of the 2000 events. In this study we estimate the rupture areas of the 1971 and 2000 events using teleseismic waveform inversion and investigate their inter-relationships. To understand the rupture processes of the 1971 events, waveform inversions using Pdiff waves was carried out. Paper recordings of the Pdiff waveform data at seismic stations for distance between 100 and 140 degree were scanned from microfilm and digitalized. Synthetic seismograms were calculated with theoretical Green's Functions and the best fit between observed and synthetic seismograms was obtained by a least-square method. Fault mechanisms and rupture processes were obtained by the waveform inversion. Results of the waveform inversion showed that both earthquakes were thrust events having shallow dip angles. The July 14 event has two areas with large slip and the July 26 event has a single area of large slip. Teleseismic P wave inversions were previously carried out to understand the rupture processes of the 2000 events. P-wave data from stations at distance between 30 and 90 degree were retrieved from the IRIS data center and used for waveform inversion. As the result of the waveform inversion, both events seem to have fairly simple rupture patterns with a single area of large slip. Based on the result of the waveform inversions, we will discuss the relationships between the rupture areas of the 1971 and 2000 events, and their rupture mechanisms.
2003 Seismological Society of Japan Fall Meeting
Kyoto, October 6-8, 2003
Swarm Activity of DLF Earthquakes Beneath Western Tottori Region
Ohmi, S. and J. Mori
Scaling of Radiated Energy for Intermediate Depth Earthquakes
Mori, J. and A. Kim
Estimate of Fault Planes and
Rupture Veloclties of Small Earthquakes in a South
African Gold Mine:
Constraints on Radiation Efficiency
Yamada, Y., J. Mori, H. Ogasawara, Y. Iio, H. Kawakata, S. Ide, The Research Group for Semi-Controlled Earthquake-Generation Experiments in Deep African Gold Mines
Rupture Areas of Large Earthquakes in Papua New Guinea in 1971 and 2000
Park, S. and J. Mori
Initiation Process of Earthquakes in Northern Miyagi
Sato, K. and J.Mori
23rd General Assembly of the International Union of Geodesy and Geophysics
Sapporo, Japan, June 30 - July , 2003
Frictional Heat and Radiated Energy Budget for the 1999 Chichi, Taiwan Earthquake
Mori, J. and H. Tanaka
We examined the energy balance of the 1999 Chichi, Taiwan earthquake (Mw 7.6) using several estimates of radiated and thermal energy. Estimates of radiated energy from regional seismograms give a value of about 1.0x1016 joules. The static stress drop from the total moment and the fault area is about 3 MPa. Temperature measurements from 2 shallow boreholes in the northern and southern sections of the fault show temperature profiles that increase across the narrow fault zone. If we assume this temperature increase was caused by frictional heating during faulting of the earthquake, thermal modeling gives the results that the fault generated 2.5 x 106 joules per square meter in the north and 4.5 x 106 joules per square meter in the south. If these frictional values are extrapolated to depth, using higher normal pressure, we estimate that the earthquake produced a total of about 2 x 1017 joules of frictional heat. Adding the radiated and thermal energy gives a total energy of the earthquake (neglecting the fracture energy) of about 2.1 x 1017 joules. This implies an average seismic efficiency is about 5%.
The average energy values for the earthquake can be quite different from the energy balance on smaller portions of the fault. For example, most of the radiated energy is generated by a large asperity on the northern part of the fault, which has an area that is about 20% of the whole fault surface. For this region of large slip, it has been suggested that the dynamic friction may be very low. If we use a value of 0.2 for the coefficient of friction, which is consistent with the borehole temperature data, the thermal energy for region of the asperity will be about 3x1016 joules and the seismic efficiency for the asperity is about 15%, which is much higher than the average value for the whole earthquake.
Image of the Fault Plane and Crustal Structure for the 2001 Gujarag, India Earthquake from Aftershock Observations: A Deep Crustal Event with High Stress Drop
Negishi, H., J. Mori, T. Sato, S. Kumar, R. Singh, P. Bodin, B.K. Rastogi
A large earthquake (Mw 7.7) occurred in the state of Gujarat in northwestern India on January 26, 2001, causing severe damage over a widespread area.About 20,000 people were killed and the total damage amounted to about US$ 5.2 billion. Surprisingly, this crustal earthquake did not have obvious surface displacements for the main fault, although the magnitude is large. It is important to know the orientation and size of the fault for evaluations of seismic hazards and understanding the tectonics of this region.
We deployed 7 seismographs around the damaged area from February 28 through March 6, 2001, and recorded waveforms from over 1,300 events. Hypocenters were determined using the Joint Hypocenter Determination program, which simultaneously calculates hypocenters, station corrections and a one-dimensional velocity structure (e.g., Kissling et al., 1994). Our aftershock locations show a trend that dips toward the south at about 50 degrees which is interpreted as the fault plane of the mainshock. The depth range of the aftershocks is from 10 to 35 km, which somewhat deeper than other crustal earthquakes, and indicates that the faulting did not reach the surface. The area of the fault is about 40 km x 40 km, which is small for a Mw 7.7 earthquake and results in a high static stress drop of 13 to 25 MPa. There are no mapped faults or obvious topographic features along the surface projection of this fault. These findings show that very large damaging earthquakes can occur without producing surface faulting.
A tomographic inversion for the 3-D velocity structure was also carried out. Because the resolution of the inversion depends on the number of data and ray coverage, we used the merged data set collected by the National Geophysical Research Institute of India, Center for Earthquake Research and Information, Memphis University and our data in this analysis. We tried to determine optimal damping parameters quantitatively by using the Cross-Validation Technique. Our 3-D velocity model shows that the aftershock distribution corresponds to the high velocity anomalies. Low Vp/Vs anomalies are generally found at depths of 10 to 35 km, i.e. the depth range of the aftershock distribution. However, relatively high Vp/Vs and low Vs characterize the deeper region below the hypocenter of the mainshock, at depths of 30 to 40 km. This anomaly may be due to a weakly fractured and fluid filled rock matrix, which might have contributed to triggering this earthquake. This earthquake occurred on a relatively deep and steeply dipping reverse fault with a large stress drop. Theoretically, it is difficult to slip such steep faults. Our tomographic investigation may provide information, which helps explain why this thrust event could occur at such depths in the crust.
Deep Low-frequency Earthquake Associated with Active Faults in Central-Southwest Japan
Ohmi, S. and J. Mori
Deep low-frequency (DLF) earthquakes have been discussed in association with fluid (magma) activity around volcanoes. In this paper, however, we will show several examples of DLF activity beneath active faults in central - southwest Japan which do not have volcanoes in their near vicinity.
On October 6, 2000, a
Mw6.7 crustal earthquake occurred in western Tottori prefecture, southwest
Japan. Beneath the focal region of the earthquake, DLF earthquakes were
observed at depths of around 30 km. Five DLF earthquakes were detected during
the 3 years before the mainshock and one occurred 9
hours before the mainshock. The focal mechanism of
the DLF earthquake that occurred 9 hours before the mainshock
was analyzed by using amplitude ratios of the S-waves to the P-waves and
polarization patterns of the S-waves (Ohmi and Obara, 2002).The result indicated that a single-force
source mechanism is more preferable than a double-couple source mechanism,
which suggests the transport of fluid, such as water or magma. More than 100
DLF earthquakes, most with magnitudes up to about M1.5, were observed during
the 2 years after the mainshock. There are a few
larger DLF earthquakes (M>2.0), that have source time durations considerably
longer than those expected from scaling relations between source dimension and
seismic moment for normal earthquakes. This indicates that the source
mechanisms of these DLF events are quite different from that of ordinary
In central Kyoto prefecture, southwest Japan, DLF earthquakes have been observed since 1978. They are located below the Mitoke Fault system. Continuous shallow swarm activity is also observed in the region. There are no volcanoes in the area. In central Japan, DLF earthquakes were mainly concentrated in the Hida Mountain range. One prominent area of activity is near the northeastern part of the Atotsugawa fault, where the fault enters the region of volcanoes in the Hida Mountain range. Mt. Tateyama, which is an active volcano, is located 10 km from the focal region of the DLFs. In all the regions, the maximum magnitudes of the DLF events are slightly larger thanM2.0, and the focal depths ranges from 25 km to 35 km.
Seismic tomography analysis (e.g. Zhao et al., 2000) indicates the existence of a low-velocity body at depths from the lower-crust to the upper-mantle in the focal region of these DLFs. The distribution of DLF activities, described above, corresponds well to the low-velocity region rather than to the distribution of volcanoes. The occurrence of DLF events is probably direct evidence of fluid activity below the seismogenic zone, where fluids might be injected from the lower-crust beneath active faults.
Triggering of Large Earthquakes in the 2000 New Ireland, Papua New Guinea Sequence
Park, S. and J. Mori
A major earthquake (Mw8.0), followed by two more large events (magnitudes greater than Mw7.0), occurred near New Ireland, Papua New Guinea, on November 16-17, 2000. Since the events happened within two days and their hypocenters are located within 250 km, this is one of the best opportunities to study triggering of large earthquakes. The occurrence of the first event has a large rupture and significantly alters the stress field in the local area. In this study we calculate the static Coulomb stress changes resulting from these earthquakes and investigate how these stress changes may trigger the following large earthquakes.
Teleseismic P wave inversions were carried out for the three events to understand the source processes and to obtain the slip distributions. P-wave data retrieved from the IRIS data center were used for the inversion. Using the data and synthetic seismograms, we carried out an inversion process to search for the best fitting fault mechanism, slip distribution, and total seismic moment for each event. Using the slip distribution of the first event, Coulomb stress changes for the location and orientation of the fault plane of the second event were calculated. We also calculated the Coulomb stress changes for the fault plane of the third event using the slip distributions of the first two events.
The rupture process of the first event shown by the slip distribution seems very complex according to the teleseismic P wave inversion. The fault does not appear to be a pure strike-slip event, but has some component of dip-slip. There is an area of large shallow slip that is consistent with the observed surface displacements of about 5 meters on southern New Ireland. Details of the slip distribution of the first event and the depth of the second event make it difficult to say clearly if the second event initiated in an area of stress increase or decrease caused by the first event. The third event clearly occurred in an area which showed a decrease of Coulomb stress change from the first and second events. So static stress changes do not provide a good explanation for the triggering of this third event.
2003 Japan Earth and Planetary Science Joint Meeting
Makuhari, Chiba, May 26-29, 2003
Absolute value of stress prior to the 2000 Tottori earthquake
Mori, J. and K. Yoshikawa
The well-determined focal mechanisms of aftershocks of the 2000 Western Tottori earthquake show a bimodal distribution in the direction of P-axes. We interpret these two directions as representing the local stress field before and after the mainshock. At the time of the earthquake, the slip on the fault causes a local change in the stress field, which is reflected in the rotation of the focal mechanism. Using this change in direction along with the calculated stress change due to slip during the mainshock, we can estimate the absolute value of the local stress field before and after the mainshock. For the direction of the pre-mainshock stress field, we assume east-west compression, which is consistent with the focal mechanisms of the mainshock and moderate earthquakes that occurred during the previous several years. For the direction of the post-earthquake stress field, we use the focal mechanisms of aftershocks that show a rotation of 15 to 20 degrees that is different from the pre-earthquake direction. Our results show that the level of stress prior to the mainshock was relatively low, on the order of 5 to 20 Mpa at the locations of the aftershocks. The fact that we see any rotation of the stress field, is an indication that the level of stress is rather low. For some of the aftershocks, the level of stress after the earthquake, is higher than the level before the earthquake. This is consistent with the idea that some aftershocks occur in regions where stress is increased due to the fault slip during the mainshock.
Complexity in Earthquake Initiation Process
Sato, K. and J. Mori
In the recent years, earthquake initiation process and its scaling relationship has been often studied. One of the studies performed by Iio [1992,1995] indicated that there is a scaling relationship in duration of a slow initial phase that appeared at a beginning of the P-wave. Sato and Kanamori  showed a model to explain the slow initial phase in which model initial crack size of an event scaled with its final earthquake size. Using this model, some studies (e.g. Sato and Mori , Hiramatsu et al. ) estimated initial crack sizes and discussed their scaling relationships. Sato and Mori  analyzed moderate to large earthquakes and concluded that scaling relationships appeared in microearthquakes continued to M4 events, but it did not appear in larger events (M6-7). They removed events with 'complex' waveforms (or chose 'simple' waveforms), although all of the larger events have 'complex' waveforms. This may cause the break of the scaling. In general, while larger earthquakes have complex rupture processes, small events can have both simple and complex rupture processes, and we may conclude that larger earthquakes than a certain size often have complex rupture pattern. In this study, we analyze frequency distribution of multiple shock events which have complex rupture pattern and its scaledependency. We use Hi-net waveform data for events in the JMA catalog between June 2002 and January 2003. In this data set, we can use 146 events with magnitudes larger than 3. In addition, we calculate sizes of the first subevents and initiation related parameters for multiple shock events and discuss scaling relationship of earthquake initiation process using the first subevent sizes.
Rupture Velocities of Small Earthquakes in a South African Gold Mine: Constraints on Fracture Energy
Yamada, T., J. Mori, H. Kawakata, H. Ogasawara, S. Ide, M. Yoshimura, International Research Group for Semi-controlled Earthquake Generation Experiment at South African Gold Mine
Analyses of rupture velocities of earthquakes are important to investigate characteristics of fracture energies, initiations, and arresting mechanisms. But it is especially difficult to resolve rupture velocities of small earthquakes because close station spacing near the hypocenter and high sampling rates are necessary. Such observations are being carried out in a South African gold mine for mining induced earthquakes. Nine tri-axial borehole accelerometers were installed within 200 m along a 2,650- m-deep haulage tunnel in the Mponeng gold mine. More than 25,000 seismic events with magnitudes between -2.7 and 3.3 were recorded with a sampling frequency of 15 kHz from February to October, 1996. We carefully picked 3 events with magnitudes 1.4, 1.1, and 0.8 having good azimuthal coverage and analyzed the waveforms to try to determine rupture velocities.
The studied events have rather complicated waveforms and individual subevents could be identified. Arrival times of the subevents were picked relative to the initial arrival. These differential arrival times were used to locate the subevents relative to the initial hypocenter. Approximate rupture velocities could be obtained by dividing the distance to the subevent by the delay time. We obtained results that showed rupture velocities ranging from 2.10 to 3.01 km/s. These values are from 55 to 78 % of the shear-wave velocity and consistent with those of larger natural earthquakes. This result suggests that the ratios of fracture energies to radiated energies of small earthquakes in a South African gold mine are not particularly large and almost the same as those of larger natural earthquakes.
The Mj5.3 Earthquake at Central Tottori Prefecture (Sep. 16, 2002) and the Seismicity of the San-in District
Nakao, S., H. Katao, J. Mori, T. Shibutani, K. Watanabe, K. Ito
Major earthquakes (more than M6) along the coastline of the San-in district have individually NW-SE trend of aftershock distribution. Only the 1943 Tottori earthquake (M7.2) shows aftershock distribution of E-W direction. Precise aftershock distribution is important to know the geometry of the mainshock rupture. During 1980-1989, seismic activities along the San-in district migrated from east to west. During 1990-2000, active seismicity was observed around western Tottori prefecture. Then, the Western Tottori Earthquake (Mj7.3) occurred in 2000. After the Western Tottori Earthquake, seismicity in the eastern Shimane prefecture, east side of Mt Daisen and the northern Hyogo prefecture are activated. On September 16 in 2002, a Mj5.3 earthquake occurred at central Tottori prefecture. The epicenter is located at east foot of Mt. Daisen. This area is western end of the aftershock zone of the 1943 Tottori earthquake. Numerous number of aftershock occurred following the mainshock. Aftershock distribution shows E-W linear trend which is consistent with the focal mechanism of the mainshock. Also from the analysis of the directivity of the waveform, it is confirmed that the mainshock rupture is E-W direction. But 1 or 2 days after the mainshock, the aftershock area extended to South, and formed another E-W trend line. Finally entire aftershock distribution shows Z-shape. Total aftershocks are located in the area of 4km x 3km. Depth range of the hypocenter is 3-11km. Focal mechanisms of the major aftershocks are strike-slip, and the P-axis directions are nearly NE-SW.
imaging of subsurface structure of Rabaul volcano
Tanimura, T. and J. Mori
We carried out a tomographic inversion to determine the three-dimensional P-wave velocity structure of Rabaul Caldera, Papua New Guinea. This is an active volcano that has had two eruptions during the 20th century, including the 1994 eruption that destroyed Rabaul Town. The three-dimensional velocity structure helps to identify the magma system that underlies the
caldera. We used 3756 P-wave arrival times from 455 earthquakes recorded at the Rabaul Volcanological Observatory (RVO) permanent network of 12 stations. Added to this, 1854 P-wave arrival times from marine seismic shots fired in the sea areas around Rabaul as part of the Rabaul Earthquake Location and Caldera Stracture (RELACS) program were used. The gird intervals were 1.5km in the horizontal direction and 1km in the depth. Grids were selected to cover Rabaul caldera (total 10x10x6 nodes). Checkerboard tests showed a good resolution within the caldera to 4km depth. A low-Vp zone stretching from north to south at center of caldera was found at 1-3km depths. The location of this low-Vp zone is consistent with ground uplift deformation area for the period 1973-1985. This suggests that the low-Vp zone is a magma chamber. This low-Vp zone is located at the center of the caldera (not beneath the active vents that produced recent eruptions). This implies there are narrow pathways from the large magma reservoir in the center of the caldera to the active vents on the outer ring fault.
Activity of Deep Low-Frequency Earthquakes in the Western Tottori Region
Ohmi, S. and J. Mori
On October 6, 2000, a Mw6.7 crustal earthquake occurred in western Tottori prefecture, southwest Japan. Beneath the focal region of the earthquake, DLF earthquakes were observed at depths of around 30 km. Five DLF earthquakes were detected during the 3 years before the mainshock and one occurred 9 hours before the mainshock. The focal mechanism of the DLF earthquake that occurred 9 hours before the mainshock was analyzed by using amplitude ratios of the S-waves to the P-waves and polarization patterns of the S-waves (Ohmi & Obara, 2002). The result indicated that a single-force source mechanism is more preferable than a double-couple source mechanism, which suggests the transport of fluid, such as water or magma. More than 100 DLF earthquakes, most with magnitudes up to about M1.5, were observed during the 2 years after the mainshock. There are a few larger DLF earthquakes, whose magnitudes are greater than 2.0, that have source time durations considerably longer than those expected from scaling relations between source dimension and seismic moment for normal earthquakes. This indicates that the source mechanisms of these DLF events are quite different from that of ordinary tectonic earthquakes.