Double seismic zone and seismicity in the mantle wedge beneath the Ogasawara Islands identified by an ocean bottom seismometer observation

Abstract

Around the Ogasawara Islands, only a few seismic stations in the area can be used to determine the hypocenters of regional earthquakes; thus, hypocenter location precision tends to be low. To more precisely determine hypocenter locations, we deployed a temporary seismic observation network of pop-up ocean bottom seismometers around the Ogasawara Islands from July to October 2015. We identified a double seismic zone in the 70–200 km depth range associated with the subducting Pacific slab. The slab-normal distance between the two planes of the double seismic zone is about 30–35 km, similar to such distances observed along the Japan and Mariana trenches. Furthermore, we found unusual seismicity in the mantle wedge at 20–50 km depth beneath the Ogasawara trough that might be related to structure formed at the onset of the oceanic slab subduction. The hypocenters determined from the ocean bottom seismometer observation were horizontally separated by a few tens of kilometers from hypocenters published by the Seismological Bulletin of Japan. USGS locations (Preliminary Determination of Epicenters) seem to be offset westward about 30 km compared with the locations determined in this study.

Introduction

The Japan Meteorological Agency (JMA) compiles seismic data from many sources and publishes hypocenter data in the Seismological Bulletin of Japan (SBJ) with the help of the Ministry of Education, Culture, Sports, Science and Technology of Japan. Although inland areas of Japan are covered by a dense seismic network, the number of seismic stations is limited in small island regions far from the main islands of Japan. In the Ogasawara Islands, in particular, there are only three seismic stations, located almost in a line on Chichijima, Hahajima, and Iwoto islands; the next closest station is on Aogashima, an island 600 km north of Chichijima. Hypocenters in the Ogasawara region are determined by using data from this sparse local network together with data from some stations on Honshu and other Japanese islands. Epicenters in the Ogasawara region published by SBJ are often located several tens of kilometers east of corresponding epicenters in the Preliminary Determination of Epicenters bulletin (PDE), published by the US Geological Survey (USGS) (Fig. 1).

Fig. 1

Comparison of hypocenter locations observed in 2013: gray circles, SBJ; blue circles, PDE. The locations of OBSs (orange triangles) and the land stations on Chichijima and Hahajima (inverted yellow triangles) used in this study are also shown. The bathymetry (gray lines) is shown with a contour interval of 500 m. The inset map shows the area of this study and the direction of movement (arrow) of the Pacific plate (PAC) relative to the Philippine Sea plate (PHS)

In the Ogasawara region, a number of reflection and refraction seismic surveys have been conducted with ocean bottom seismometers (OBSs) (e.g., Takahashi et al. 2009). One survey, for example, deployed OBSs along a 580-km-long baseline and estimated the velocity structure down to 30 km below the seafloor (Kodaira et al. 2010). However, no OBS observations targeting natural earthquakes are conducted in this area.

The subduction of the Pacific plate (PAC) beneath the Philippine Sea plate along the Izu-Ogasawara trench has recently led to notable seismic activity, including a normal-fault earthquake (M_w 7.3) on 22 December 2010 (JMA 2010) and a deep-focus earthquake (M_w 7.9) on 30 May 2015 (JMA 2015). Along the Izu-Ogasawara trench, the occurrence of shallow earthquake greater than M 8 has not been known. Scholz and Small (1997) suggested that great earthquakes might occur where the Ogasawara Plateau is subducting beneath the Hahajima island group because of increase in interplate seismic coupling caused by seamount subduction. Ishibashi and Harada (2013) suggested a possibility that the source area of the 1605 Keicho earthquake may have been along the Izu-Ogasawara trench. Knowledge of the precise locations of seismic activity in this area could thus provide important information about future great earthquakes.

In this study, we determined the precise distribution of earthquake hypocenters with a pop-up OBS network that was deployed around the Ogasawara Islands for about three months in 2015. We aimed to grasp a wide range of seismic activity in the Ogasawara region with a horizontal distance of about 150 km and a depth down to about 200 km by using an observation network with the observation station intervals of 50 to 100 km.

Observation with pop-up OBSs

The OBS seismic network was deployed from 15 July to 10 October 2015 by R/V Keifu-Maru, a JMA research vessel. Twelve OBSs (stations A–L) were deployed in a 200 km × 100 km area, with five of them (stations B–F) deployed in the vicinity of Nishinoshima volcano (Table 1, Fig. 1). Data from OBSs, along with data recorded by the permanent land stations on Chichijima (station X) and Hahajima (station Y), were used for hypocenter determination. The horizontal coordinates of the OBSs were estimated by acoustic ranging, which usually has a location error of 20–30 m (Yamazaki 2011). Each OBS consisted of a three-component seismometer with a natural period of 4.5 Hz (installed on a gimbal with oil-damping) and a hydrophone. Data at 16-bit resolution were continuously recorded at a sampling rate of 100 Hz. The difference between the clock time of each OBS and the Global Positioning System timing signal was measured before installation and after recovery of each OBS; then the OBS clock times were corrected by assuming that the time-shift rate was constant.

Table 1 Parameters of OBS stations used in this study

Hypocenter determination

Events were detected by applying the short-term-average/long-term-average (STA/LTA) method (Urabe and Hirata 1984) to the continuous record of seismic data from nine seismic stations (A, G–L, X, and Y; Fig. 1). Data recorded at stations B–F, close to Nishinoshima volcano, were excluded from the STA/LTA analysis because of the high level of volcanic activity during the OBS deployment period, but they were used to hypocenter determination. We set the decay constants for STA and LTA calculation at 2 and 60 s, respectively. When the STA/LTA exceeded a value of 3.0, detection status was turned on. When STA/LTA fell below 2.0, the status was turned off. A continuous ON status for more than 5 s was judged as a triggered event. These processes were performed at each observation station. The triggered event at only one station may be a noise or a small earthquake for which the hypocenter cannot be determined. Therefore, we selected the triggered events which got on at two or more stations simultaneously within 20 s and detected 1601 events as candidates for earthquake events. The WIN system was used to pick seismic wave onset times and maximum amplitudes (Urabe and Tsukada 1992). Examples of seismic records are shown in Fig. 2. Hypocenters of earthquakes for which one or more P-wave arrivals and five or more total P- and S-wave arrivals were detected were estimated with the HYPOMH program (Hirata and Matsu’ura 1987). We assumed a one-dimensional P-wave velocity structure based on the results of a seismic survey by Takahashi et al. (2009) (Fig. 3). The S-wave velocity was assumed to be 1/1.73 of the P-wave velocity, and velocity was assumed to be constant above the seafloor. Because the thickness of the sediment layer at the seafloor and base rock velocities beneath stations usually varies spatially, we applied an iterative station correction method several times to correct for differences in site conditions (Iwasaki et al. 1991; Sakai et al. 2005; Nakata et al. 2016). First, the correction was done by adding the time difference of S phase by sedimentary layer from the arrival-time difference between the PS and P phases (Table 1). After second time, the corrections were applied to both P and S phases. We adopted the seventh result of the iterative station correction method (see Additional file 1 for details). The standard deviations and station-correction values at this time are shown in Table 1. We adopted the 269 hypocenters within the observation network. We also tried the HypoDD method (Waldhauser and Ellsworth 2000) for comparison. Though clustered hypocenters tended to become compact, we could not get a enough number of reliable hypocenter locations for sparsely distributed ones with the HypoDD method.

Fig. 2

Examples of seismograms observed by OBSs A, H, and J

Fig. 3

One-dimensional P-wave velocity model used for hypocenter determination derived from Takahashi et al. (2009). For comparison, the velocity model used by JMA2001 (Ueno et al. 2002) is also shown

Seismic activity in the Ogasawara region

The hypocenter distribution results obtained in this study for the Ogasawara region, including the magnitude-time diagram and the magnitude-frequency distribution (Fig. 4, Table 1), can be summarized as follows:

Fig. 4

Hypocenters determined in this study. a Map view and N–S and E–W vertical cross sections showing the hypocenter distribution (open circles) determined from the data recorded by OBSs (orange triangles) and the two island stations (inverted yellow triangles) from 15 July to 10 October 2015. The diameters of the circles indicating hypocenters vary according to the magnitude of the earthquake. Hypocenters in the upper seismic zone are shown by red circles, and those in the lower seismic zone by blue circles. Hypocenters away from the plate boundary are shown by green circles. SBJ hypocenters determined with high precision from January 2002 to December 2017 around the deployment area of the OBSs are also shown by solid gray circles. The dashed purple line in the map view shows the survey line of Takahashi et al. (2009). The dotted black line in the E–W cross section indicates the upper surface of the Pacific slab along 27.3°N, as determined by Hayes et al. (2012). α and β indicate two clusters of hypocenters at 60–80 km depth. b Magnitude-time diagram. The green line shows the cumulative number of earthquakes with M 2.2 or greater. c Magnitude–frequency distribution. Open circles denote the number of earthquakes in each magnitude bin, and solid circles show the cumulative number of earthquakes. The best-fit line according to the Gutenberg–Richter law is also shown (Gutenberg and Richter 1944)

• (1) Seismic activity was stationary during the observation period, and no swarm activity was recognized (Fig. 4b).

• (2) The lower limit of detectable magnitude was about 2.0. The best-fit line for the magnitude-frequency distribution (Gutenberg and Richter 1944) had an estimated b-value of 0.72 (Fig. 4c). Here, the best-fit line was deduced from the maximum likelihood methods in the range of M 2.2 to 5.3.

• (3) Seismicity related to the subducting PAC was recognized when the hypocenters were projected onto an east–west cross section (Fig. 4a). Double seismic zone (DSZ) activity observed along the subducting slab (red circles, upper plane; blue circles, lower plane) is similar to that observed beneath the Tohoku area. The plate configuration inferred by Hayes et al. (2012) at 27.3°N, shown by a dotted line on the cross section, corresponds well to the activity observed along the upper plane.

• (4) Some earthquakes were also distributed at about 20–50 km depth in the mantle wedge (green circles in Fig. 4a).

Double seismic zone (DSZ)

Seismicity was observed along two planes in the depth range from 70 to 200 km (Fig. 4a). The slab-normal distance between the two planes was 30–35 km, similar to corresponding distances observed along the Japan and Mariana trenches (Umino and Hasegawa 1975; Hosono and Yoshida 2001; Yoshida and Hosono 2002; Nakajima et al. 2009; Shiobara et al. 2010). In general, DSZs are considered to relate to pressure and thermal conditions within a subducting plate, with the upper and lower planes attributable to dehydration of the oceanic crust and of mantle serpentinite, respectively (e.g., Yamasaki and Seno 2003; Seno 2009). The relationship between the slab-normal distance obtained in this study and plate age (145.3 Ma, Syracuse et al. 2010) in the Ogasawara region is consistent with the observed global trend that slab-normal distances increase with plate age (Brudzinski et al. 2007). The DSZ is recognizable down to at least 200 km (Fig. 4a), but its extension to depths below 200 km cannot be confirmed because those depths are not covered by the observation network. Wei et al. (2017) showed that, globally, the maximum depth of DSZs is positively correlated with the thermal parameter calculated by multiplying the vertical subduction speed by plate age (Kirby et al. 1996). The relationship between the maximum depth obtained in this study and the thermal parameter in the Ogasawara region (40.8 km, Syracuse et al. 2010) is consistent with results compiled by Wei et al. (2017).

Clusters of hypocenters are recognized at 60–80 km depth in the upper plane (α and β on the map view in Fig. 4a). Kita et al. (2006) found a seismic belt at depths 70–100 km in the upper plane beneath northeast Japan, which is related to dehydration reaction in the oceanic crust. The clusters in Fig. 4 are distributed in the same depth range of the seismic belt found beneath northeast Japan. The clusters may correspond to the southward extension of the seismic belt beneath northeast Japan. The depth range of the seismic belt varies according to thermal condition (Kita et al. 2010). The relatively shallow depths of the seismic belt in the Ogasawara region may indicate the absence of such low-temperature materials seen beneath Hokkaido area.

Only 11 earthquakes were observed in the lower plane of the DSZ. Similarly, in the DSZ in the Tohoku area, less seismicity is observed in the lower plane compared with the upper plane (Hosono and Yoshida 2001). According to Umino and Hasegawa (1975), a down-dip extension-type focal mechanism is dominant in the lower plane. We tried to estimate a composite focal mechanism using data from all 11 earthquakes in the lower plane, but we obtained no clear result.

Seismicity in the mantle wedge

Some seismic activity was observed in the 20–50 km depth range inside the mantle wedge (green circles in Fig. 4a). These earthquakes, which are distributed in a wide area without forming any clusters, are located beneath the Ogasawara Trough, a forearc basin (e.g., Takahashi et al. 2009) between the current volcanic arc and the Ogasawara ridge. The seismic records of one of these earthquakes are shown in Fig. 5, and they indicate no special features such as low frequency or long duration. Active seismicity in the mantle wedge has also been observed in areas with complex plate configurations such as southern Hokkaido or the Kanto area of Japan. In the Hokkaido area, the Kuril and northeast Japan arcs are colliding (e.g., Kimura 1996), and beneath the Kanto area, the Pacific and Philippine Sea plates interact as they both subduct beneath the Kanto area (e.g., Ishida 1992; Yoshida and Hosono 2002). However, seismic activity is usually low in the mantle wedge in areas without complex plate interactions, with some exceptions. For example, Uchida et al. (2010) found seismicity at depths of 25–50 km above repeating earthquakes at the plate boundary. Davey and Ristau (2011) suggested that fluid produced by dehydration of serpentine in the mantle wedge is related to the seismicity observed in the mantle wedge beneath northeast New Zealand. Laigle et al. (2013) reported seismicity in the mantle wedge along the Lesser Antilles subduction zone. The seismicity in the mantle wedge beneath Lesser Antilles is distributed over trench-normal distance more than 50 km, which is wider than that seen in Fig. 4a. Depth ranges and sparse distribution seem similar between Ogasawara region and Lesser Antilles. Supraslab earthquakes offshore Sanriku (Uchida et al. 2010) are distributed mainly near the toe of the mantle wedge, and form clusters. Earthquakes in the mantle wedge beneath northeast New Zealand (Davey and Ristau, 2011) form clusters near the boundary of subducting slab. The seismicities offshore Sanriku and beneath northeast New Zealand look different from the seismicities around Ogasawara Islands and Lesser Antilles.

Fig. 5

Vertical seismic records of an earthquake in the mantle wedge, with the P- and S-wave arrival time picks used for hypocenter determination indicated

The tectonic features of the forearc basin in the Lesser Antilles and Ogasawara Island regions are similar. Thus, seismicity in the mantle wedge may be a feature of forearc basin formation, and serpentine dehydration might account for the seismic activity in the mantle wedge in the Ogasawara region. Ogasawara Islands was formed with volcanic activities in 41–48 Ma related to the beginning of the Pacific plate subduction in this area (Ishizuka et al. 2006). The backarc basin of the Lesser Antilles, Grenada basin, and the forearc basin, Tobada basin, were formed as a single structure, and the Lesser Antilles were formed by volcanic activities with dividing the single basin into the two (Aitken et al. 2011). The volcanic activity started at about 40 Ma (Briden et al. 1979) due to the rollback of the Atlantic oceanic slab subduction (Aitken et al. 2011). Both Ogasawara Trough and Tobada basin are related to the beginning of oceanic slab subduction. The seismic activity in mantle wedge may be related to dehydration of minerals which were formed at the time of subduction onset.

Comparison with SBJ and PDE locations

We compared the hypocenter locations in this study with SBJ and USGS PDE locations (Fig. 6 and Table 2). We considered earthquakes with an origin time difference less than 3.0 s to be the same earthquake. SBJ hypocenters are categorized into “high-precision hypocenters” and “low-precision hypocenters” according to hypocenter determination errors (JMA 2019). In the Ogasawara region, the locations of hypocenters in the latter category are uncertain because of the small number of used observation stations which are unevenly distributed. The magnitudes of the earthquakes with low-precision hypocenters are generally smaller than the others. Only two high-precision SBJ hypocenters matched with our OBS hypocenters (connected by lines in Fig. 6a), and the SBJ and OBS hypocenter locations were a few tens of km apart. Fifteen low-precision hypocenters (flagged with “S” in Table 2) also matched with OBS hypocenters. When all 17 earthquakes are considered, the distances between SBJ and OBS hypocenter locations are distributed from a few km to over 100 km.

Fig. 6

Comparison of hypocenters. a Map view and N–S and E–W vertical cross sections comparing SBJ hypocenters (gray circles) with corresponding hypocenters determined in this study (open red circles) and PDE by the USGS (open blue circles). b Relationship between magnitudes determined in this study and SBJ magnitudes. Open and filled circles denote low- and high-precision SBJ hypocenters, respectively

Table 2 Comparison between SBJ hypocenters and those of this study

We also compared the locations of two OBS hypocenters with the PDE locations estimated for the same earthquakes (Table 3) and found that the PDE hypocenters were about 30 km west of the OBS hypocenters. These two earthquakes were same as those matched with SBJ. The high-precision SBJ hypocenter locations were closer to the OBS than the PDE, although there were only two data. When we compared the hypocenter locations of the SBJ and the PDE occurring over a 1-year period (Fig. 1), we found that the difference between them is remarkable in the east–west direction. The SBJ earthquake locations were estimated by using onset times obtained at Chichijima, Hahajima, Iwoto, and stations on distant main islands of Japan. The PDE earthquake locations were estimated by them at Chichijima, some stations on main islands of Japan, and distant stations around the world. The distribution of travel-time residuals for an earthquake occurring at 20:53:13 20 July 2015 in the SBJ data (Fig. 7) shows an area with small residuals with an east–west orientation, and the resolution of hypocenters is considered to be poor in the east–west direction. This is considered to be a cause of the hypocenter difference in the east–west direction.

Table 3 Comparison between PDE hypocenters and those of this study

Fig. 7

An example of the travel-time residual distribution of an earthquake in SBJ. The gray circle indicates the hypocenter location, and the inverted yellow triangles show the locations of three stations situated on islands

We also examined the relationship between OBS magnitudes and SBJ magnitudes (Fig. 6b), including those of the low-precision SBJ hypocenters. We consider the precision of the hypocentral distances used in magnitude determination to be reliable because the hypocentral distance is constrained by the observed P–S time difference even if data are available from only a few stations. The greater magnitudes of this study, which are higher by 0.5 relative to the SBJ magnitudes (Fig. 6b), can be attributed to amplification by the soft sediment layers beneath the OBS stations (Aoki and Yoshida 2005).

Conclusions

A seismic observation by ocean bottom seismometers in 2015 revealed some notable features of seismicity in the Ogasawara region.

• (1) Seismicity relating to the subduction of the Pacific plate (PAC) was observed, and a double seismic zone (DSZ) was observed in the 70–200 km depth range. The distribution of seismicity is mostly consistent with the plate model proposed by Hayes et al. (2012).

• (2) Seismic activity observed within the mantle wedge, above the seismicity of the PAC, may be related to onset of the oceanic slab subduction.

• (3) Seismological Bulletin of Japan locations seem to be a few tens of kilometers apart from the locations determined in this study. USGS locations (Preliminary Determination of Epicenters) seem to be offset westward about 30 km compared with the locations determined in this study.