Nonvolcanic Deep Tremor Associated with Subduction in Southwest Japan

See allHide authors and affiliations

Science  31 May 2002:
Vol. 296, Issue 5573, pp. 1679-1681
DOI: 10.1126/science.1070378


Deep long-period tremors were recognized and located in a nonvolcanic region in southwest Japan. Epicenters of the tremors were distributed along the strike of the subducting Philippine Sea plate over a length of 600 kilometers. The depth of the tremors averaged about 30 kilometers, near the Mohorovic discontinuity. Each tremor lasted for at most a few weeks. The location of the tremors within the subduction zone indicates that the tremors may have been caused by fluid generated by dehydration processes from the slab.

Long-period events and tremors with typical periods in the range of 0.2 to 2 s are often observed at active volcanoes and reflect the internal dynamics of the volcanic system (1). A possible tremor-generating mechanism is flow-induced oscillation in channels transporting magmatic fluid (2). We have identified and studied anomalous long-period tremors from a nonvolcanic area in southwest Japan by using the National Research Institute for Earth Science and Disaster Prevention's (NIED) high-sensitivity seismograph network (Hi-net), which is composed of about 600 stations installed throughout Japan to detect microearthquakes (3). The densely distributed high-sensitivity seismic stations provide a high-level detection capability for microearthquakes and offer us an opportunity to find and investigate very small amplitude tremors. Because the amplitudes of these tremors are very small, it is difficult to identify them with a single station or a sparse network.

We observed small-amplitude tremors that lasted from a few minutes to a few days, as shown in fig. S1A. The tremors were observed simultaneously at several Hi-net stations, which indicates that they are not related to artificial noise. The predominant frequency of the tremors ranged from 1 to 10 Hz and was lower than that of ordinary earthquakes of similar size (10 to 20 Hz). We transferred the raw seismogram to root-mean-square (rms) amplitude for the filtered output (fig. S1B), and tremors were clearly seen for time windows of 35 to 50 min. The envelope shapes of the tremors were very similar at different stations. The envelopes had gradual rise times and differed from those of a normal earthquake, which has a spike-like envelope shape. The similar envelope amplitude pattern seemed to have been propagated with a velocity of 4 km/s, which we roughly estimated from paste-up traces plotted with the increasing epicentral distance. This means that the source of the tremors was located at a deep portion and the envelopes were propagated not by P-wave, but by S-wave velocity. Because it was very difficult to identify the initialP- and S-wave onset for the hypocentral determination, we applied a cross-correlation technique to get the distribution of the relative arrival time of the envelope (4). The spatial distribution of the arrival time was used to determine the hypocenter of the tremor using a depth-dependent layered S wave–velocity model, which was used for the local hypocentral determination. In this method, the hypocenter of the tremor was obtained once every 1 min, if the coherent envelope of the tremor continued. However, each hypocenter was sometimes scattered because of the lack of P wave–arrival time data and contamination of the envelope shape by occurrence of tremors in different areas at the same time. Therefore, the center of the distribution of tremors determined for 1 hour was plotted (Fig. 1) as the epicenter of the tremor in the time interval if the hypocenters were distributed within an area whose radius was 10 km. The tremors were distributed along the strike of the subducting Philippine Sea plate over a length of 600 km from the Tokai area to the Bungo channel, between Shikoku and Kyushu Island. The epicentral distribution of the tremors corresponded to the seismicity with the depth range from 35 to 40 km in the Shikoku area, and with the depth range from 40 to 45 km in the Kii peninsula. No tremor has been detected around the Kii channel between the Shikoku and the Kii peninsula, nor in the east part of the Shikoku. The well-estimated source depths of the tremors are concentrated at a depth of about 30 km, near the Mohorovic discontinuity.

Figure 1

Epicentral distribution of the deep long-period tremors in the year of 2001. The circle indicates the center of epicenters of tremors calculated every hour. The crosses represent the Hi-net stations. The depth contour line indicates the maximum frequent depth-distribution of earthquakes inside the subducting Philippine Sea plate, and the gray line represents the leading edge of the subducting Philippine Sea plate (11).

We looked for the tremor activity in every 1-hour envelope seismogram and determined the frequency in units of 1-hour periods. The active period of most tremors generally continued for several days, and sometimes for a few weeks (Fig. 2). After an active period, the region was quiet for a few months. The tremors sometimes seemed to be triggered by a nearby relatively large earthquake. For instance, active tremor sequences started in the Tokai area from 10 April and 2 June, 2001, after M (local magnitude determined by Japan Meteorological Agency) 5–class earthquakes occurred at a location 40 to 50 km away from the tremor region, on 3 April and 1 June, respectively. The Geiyo earthquake (M 6.7) occurred on 24 March, and then a tremor became active in the Shikoku area. On the other hand, tremor activity sometimes finished right after a nearby earthquake. For example, in the Tokai area, tremor activity lasted for about 2 weeks in September and then stopped when the Western Aichi earthquake (M 4.1) occurred near the tremor.

Figure 2

Time sequence of the tremor activity in the Tokai, the Kii peninsula, and the Shikoku area. The frequency in units of 1-hour periods in which the tremor was recognized is plotted. The arrows indicate major earthquakes greater than M 4, which occurred near the tremor active zone. The arrows attached with the name of the earthquake indicate relatively large earthquakes, which might have affected the tremor activity.

The tremors did not always remain in one region in an episode but sometimes migrated (fig. S2). In the western Shikoku area, tremor activity started at the beginning of January at approximately 133.0°E and then moved gradually toward the west with a velocity of about 13 km/day. In the same area, the tremor activity in August moved from west to east with the same velocity as observed in January.

Deep long-period tremors often include impulsive body waves with a predominant frequency of about 1 to 2 Hz. Such an impulsive phase, which is thought to be an S wave, has been established by hypocentral determination to be a deep low-frequency microearthquake (5). Such a tremor might be a continuous sequence of low-frequency earthquakes. Such low-frequency earthquakes (long-period events), with a depth of around 30 km, occur not only around active volcanoes (6, 7) but also near active fault systems (8), according to Hi-net. These low-frequency earthquakes are sometimes followed by tremor-like long wave trains, where each wave train has a duration of about several minutes to several hours. However, the tremors observed in southwest Japan had a longer duration and a larger source area than did the low-frequency earthquakes.

Considering the long duration and mobility of the tremor activity, the generation of tremors may be related to the movement of fluid in the subduction zone. The subducting slab may liberate aqueous fluid by dehydration. Therefore, large amounts of fluid may be generated in the slab and move to the slab surface or the depth of the Mohorovic discontinuity. At high temperature and pressure, aqueous fluid mixed with silicate melts exists as a supercritical fluid (9). The presence of supercritical fluid may reduce the friction and change the fracture criterion of the rock by increasing the pore pressure and/or create new cracks through hydraulic fracturing. Therefore, tremor activity with a long duration time might be caused by a chain reaction of small fractures caused by the supercritical fluid. If the condition of the tremor generation is unstable, the additional supply of fluid to an almost saturated system, or stimulation by nearby earthquake shaking, might be able to trigger the observed tremor.


View Abstract

Navigate This Article