Background

Beneath the Kanto district of Japan, which contains the Tokyo metropolitan area, the Philippine Sea plate is subducting under the continental plate along the Sagami trough, and the Pacific plate is subducting below the Philippine Sea and continental plates along the Japan trench (Figure 1a). The dual subduction of these oceanic plates has resulted in a highly complex tectonic scenario beneath Kanto. Our understanding of seismotectonics in this region has vastly improved in the last decade following comprehensive studies on the geometries of the subducting Pacific and Philippine Sea plates (e.g., Sato et al. 2005; Hirose et al. 2008; Nakajima et al. 2009; Uchida et al. 2010), the distribution of small repeating earthquakes (e.g., Kimura et al. 2006; Uchida et al. 2009), seismic velocity structures (e.g., Matsubara et al. 2005; Wu et al. 2007), and stress regimes derived from focal mechanism of earthquakes (e.g., Hori 2006; Nakajima et al. 2011). One of the most important findings was a low-velocity and high-Vp/Vs mantle of the Philippine Sea slab beneath eastern Kanto, which is interpreted as being the serpentinized mantle (Figure 1b) (Nakajima et al. 2009). Nakajima and Hasegawa (2010) emphasized the important role of the western boundary of the serpentinized mantle on seismogenesis beneath Kanto and suggested that the western unmetamorphosed mantle moves 0.5 to 1 cm/year relative to the eastern serpentinized mantle.

Figure 1
figure 1

Study area and S-wave velocity image in the Philippine Sea slab. (a) The tectonic setting of central Japan. Red triangles denote active volcanoes. Iso-depth contours of the Pacific and Philippine Sea slabs (Nakajima and Hasegawa 2007; Hirose et al. 2008; Nakajima et al. 2009) are shown by blue and pink curves, respectively. Green shading represents an area where the upper surface of the Pacific slab is in contact with the lower surface of the Philippine Sea slab, termed the slab contact zone (Nakajima et al. 2009; Uchida et al. 2009). Arrows represent the directions of motion of the Philippine Sea and Pacific plates. The extent of the studied area is marked by a dashed rectangle. (b) Map showing S-wave velocity distribution along a curved surface located 10 km above the Pacific slab (after Nakajima et al. 2009). Red stars denote two earthquakes that occurred in the Philippine Sea slab in 1921 (M7.0) and 1987 (M6.7). A dashed pink line represents the western boundary of the serpentinized mantle of the Philippine Sea slab.

Full size image

Seismic attenuation provides additional insights into seismogenesis and plays a crucial role in the estimation of strong surface motions, but there are few studies on three-dimensional (3-D) seismic attenuation beneath Kanto. A pioneering work by Sekiguchi (1991) imaged the subducting Philippine Sea slab as a marked low-attenuation area, and Nakamura et al. (2006) revealed the existence of a pronounced E-W trending high-attenuation area at depths of 20 to 60 km. However, we still do not have enough information on seismic attenuation compared to seismic velocity. Therefore, precisely estimating seismic attenuation is of considerable importance to understand the complicated structure and mechanical interactions between the two subducting slabs. In this study, the 3-D P-wave attenuation (Q p -1) structure is estimated by using a large number of high-quality waveform data recorded at a nationwide seismograph network in Japan. The results are then discussed with reference to seismotectonics beneath Kanto.

Methods

Corner frequency, attenuation term, and site response

The three-step approach developed by Nakajima et al. (2013) was applied to estimate 3-D Q p -1 structure, which can minimize potential tradeoffs among unknown parameters to be solved, in addition to attenuation terms (t*) along a ray path. Firstly, corner frequencies (f c ) of earthquakes were estimated from spectral ratios of S-coda waves, and then a joint inversion was performed in order to determine attenuation terms and frequency-dependent site-amplification factors. Finally, the set of t* was inverted for the 3-D Q p -1 structure. The same criteria for waveform analysis and parameter selections given in Nakajima et al. (2013) were adopted for this study, and the outline of the analysis is described in the following paragraphs.

Velocity waveforms were collected for 336 earthquakes with focal depths (H) of less than 30 km and 2,621 earthquakes with H ≥ 30 km, all of which occurred from January 2003 to October 2013. Magnitudes of these earthquakes ranged from 2.5 to 5.0. Spectral amplitudes of the transverse component were calculated for S-coda waves, with a time window of 10 s taken at twice the theoretical S-wave travel time for the 1-D seismic velocity model (Hasegawa et al. 1978). Spectral ratios were calculated at common stations for available earthquake pairs that had a magnitude difference of ≥0.5, and the spectral ratios were subsequently stacked. Then, an ω2 source model (Brune 1970) was fitted to the stacked spectral ratio in the 1- to 32-Hz frequency range, and values of f c for the earthquake pair were estimated. As a result, f c for 1,915 earthquakes were determined.

P-wave and noise spectral amplitudes were subsequently calculated from the vertical component with a window length of 2.56 s after and before the onset of P-wave arrival, respectively. We assumed a frequency-dependent attenuation term, t* (f), as t * f = t 0 * f ‒ α , where t0* is the attenuation term at 1 Hz (e.g., Stachnik et al. 2004). A frequency-dependent term, α, was estimated by the method of Stachnik et al. (2004), yielding the optimal range of 0.2 to 0.3. The most relevant α for the mantle, 0.27 (e.g., Jackson et al. 2002), lies in this range, and therefore, α of 0.27 was adopted as the frequency-dependent factor. Observation equations were constructed as a set of equations for many earthquakes at one station and a joint inversion was carried out for t0*, frequency-dependent site-amplification factors, and an offset that controls the level of spectra.

Figure 2a shows path-average Q p -1 plotted on each station, which were calculated from t0* for an earthquake (M3.9) at a depth of 79 km. Stations in the western part of the study area show low attenuation, whereas stations in the eastern part show moderate to high attenuation. Ray paths from the earthquake to four stations, N.CHSH, N.YKIH, TOKYO, and N.FJOH, are shown in Figure 2b. Observed spectral amplitudes (black curve) at the four stations are explained well by theoretical spectral amplitudes (red curve) calculated with the estimated t0*, site-amplification factors, and the ω2 source model (Figure 2c). Observed and theoretical spectral amplitudes that are corrected for site-amplification factors show a rather smooth pattern over the frequency range of interest and are explained by the estimated t0* together with the ω2 source model (Figure 2d).

Figure 2
figure 2

Examples of results obtained for an earthquake (M3.9) with a depth of 79 km. (a) Map showing path-average Q p -1 plotted at station locations (colored squares). Colors represent Q p -1 values and color scales are shown in (b). The epicenter of the M3.9 earthquake is shown by a red star. (b) Vertical cross section of the Q p -1 model obtained in this study along latitude of 35.6 °N. Stations and earthquakes located in the gray band in (a) are shown by colored squares and gray circles, respectively. Black lines denote ray paths from the M3.9 earthquake (red star) to four stations (N.CHSH, N.YKIH, TOKYO, and N.FJOH) that are calculated with a 3-D P-wave velocity model (Nakajima et al. 2009). Dashed lines denote the upper surfaces of the Philippine Sea and Pacific slabs (Hirose et al. 2008; Nakajima et al. 2009). (c) Spectral amplitudes for P waves (black line) and noises (dashed line) observed at the four stations. The names of stations are shown above each panel. A red line represents the theoretical spectral amplitude calculated with t0*, site-amplification factors, and the ω2 source model. (d) Spectral amplitudes for P waves (black line) and noises (dashed line) corrected for site-amplification factors. A red line represents the theoretical spectral amplitude calculated with t0* and the ω2 source model. Inverted triangles denote the P-wave corner frequency of the M3.9 earthquake. Values of t0* and path-average Q p -1 are shown above each panel. (e) Site amplification factors estimated for the four stations.

Full size image

Tomographic inversion

The 3-D Q p -1 structure was estimated by the inversion of t0* (e.g., Rietbrock 2001). The ray-tracing technique of Zhao et al. (1992) was used to calculate ray paths and travel times for the 3-D P-wave velocity model of Nakajima et al. (2009). The Conrad and Moho discontinuities in the continental plate (Katsumata 2010), and the upper boundary of the subducting Pacific plate (Nakajima et al. 2009) were introduced as velocity discontinuities in the model space. Two 3-D grid nets were set in the model space, and a value of Q p -1 for each grid node was estimated. One net covered the crust and mantle in the continental and Philippine Sea plates, with grid nodes spaced at a horizontal interval of 0.1° and vertical intervals of 10 to 20 km down to a depth of 200 km. The other net covered the Pacific slab, where grid nodes with a horizontal interval of 0.2° were set at distances of 5, 20, and 40 km from the slab surface and orientated subparallel to it. Initial values of 0.0033 were used for Q p -1 for the crust and mantle and 0.001 for the Pacific slab.

The distributions of the 1,915 earthquakes and 165 seismograph stations used in this study are shown in Figure 3. A total of 80,644 t0* were used in the inversion, and Q p -1 values were estimated at grid nodes with more than 100 rays. The optimum damping parameter was chosen to be 50, based on the tradeoff curve between model variances and data residuals. The final results were obtained after four iterations, and the RMS of t0* residuals was reduced from 0.034 s in the initial model to 0.020 s upon optimization.

Figure 3
figure 3

Distribution of earthquakes (colored circle) and seismograph stations (white squares) used in the tomographic inversion. Colors represent the focal depths according to the scale given in the lower figure. Red triangles denote active volcanoes.

Full size image

Checkerboard resolution tests (CRTs) were used to assess the resolution of the obtained images. Q p -1 values of 0.007 and 0.001 were assigned alternately to grid nodes with every 0.2° interval in the horizontal direction and synthetic t0* was calculated with the same source-receiver geometry as the observations. Random noises with a standard deviation of 0.01 s were added to the synthetic data, and the synthetic t0* were inverted to obtain checkerboard images. The results show that checkerboard patterns are recovered well down to a depth of 80 km (Figure 4), whereas those in the subducting Pacific slab are slightly smeared, probably due to the limited and less-crisscrossed ray coverage. The results of CRTs demonstrate that the data set can resolve 3-D Q p -1 structure above the Pacific slab with a spatial resolution of approximately 20 km.

Figure 4
figure 4

Results of the checkerboard resolution test. The depth of each layer is shown at the upper-left corner of each panel. The black line shown at depths of 50 to 80 km denotes the iso-depth contour for the Pacific slab at each corresponding depth. Two panels in the bottom right show the results for layers in the subducting Pacific plate (where distance is measured from the slab surface).

Full size image

Results

High-attenuation (Q p -1 > 0.005) areas are distributed over a wide region of Kanto at a depth of 20 km, and a significant E-W trending high-attenuation area is observed along latitudes of 35.5 to 36.0 °N at depths of 30 and 40 km. This latter high-attenuation area coincides spatially with an area characterized by low velocity and high Vp/Vs (e.g., Kamiya and Kobayashi 2000; Matsubara et al. 2005). When compared with iso-depth contours for the Philippine Sea slab (broken curve in Figure 5), it is evident that the E-W trending high-attenuation area occurs within the continental plate at a depth of 30 km, whereas it corresponds to the subducting crust at a depth of 40 km. At a depth of 50 km, a N-S oriented high-attenuation area is imaged in the Philippine Sea slab immediately above the Pacific slab. This area was also partly observed by Nakamura et al. (2006). Altogether, the Philippine Sea slab is imaged as a low-attenuation zone (Q p -1 < 0.002) at depths of 60 and 70 km.

Figure 5
figure 5

Results of tomographic inversion. Map showing Q p -1 at depths of (a) 20, (b) 30, (c) 40, (d) 50, (e) 60, and (f) 70 km. The color scale is shown at the bottom. Red triangles represent active volcanoes. White stars and gray dots represent small repeating earthquakes (Uchida et al. 2009) and other earthquakes, respectively, that occurred within 5 km of each depth slice. Earthquakes (M ≥ 2) that occurred from 2001 to 2012 are represented by gray dots. Dashed and black lines show the iso-depth contours of the Philippine Sea and Pacific slabs, respectively, at each corresponding depth. Only the regions with hit counts of >100 are shown.

Full size image

Figure 6 shows that high-attenuation areas observed in this study coincide spatially with low-velocity areas. The continental lower crust situated immediately above the Philippine Sea slab shows high attenuation, and high attenuation continues to the surface in some areas (lines A and E). The continental mantle shows distinct zones of high attenuation beneath volcanic areas along the western limit of lines A-C. Interestingly, a localized high attenuation is observed at the corner of the mantle wedge where no volcanic activity occurs at the surface (lines D and E). Furthermore, high attenuation is imaged in the Philippine Sea slab crust and mantle, the latter being observed along the eastern part of lines A-C. Consequently, the mantle of the Philippine Sea slab is characterized by high attenuation and low velocity in eastern Kanto (line D) and by low attenuation and high velocity in western Kanto (line E). Such an E-W variation in attenuation is confirmed in Figure 2b, which shows that rays that propagate to eastern stations tend to have path-average Q p -1 higher than those that propagate to western stations.

Figure 6
figure 6

Comparison between (left) Q p -1and (right) P -wave velocity. Vertical cross-sections along the lines (a) A, (b) B, (c) C, (d) D, and (e) E in the inset map are shown. Solid lines indicate the upper surfaces of the Philippine Sea and Pacific slabs, and dashed lines denote the Moho for the continental plate and Philippine Sea and Pacific slabs. Black bars and red triangles on the top of each panel denote the land area and volcanoes, respectively. White stars and gray dots represent small repeating earthquakes and other earthquakes, respectively, that occurred within 5 km from each line. Earthquakes (M ≥ 2) that occurred from 2001 to 2012 are represented by gray dots. The P-wave velocity model is reproduced from Nakajima and Hasegawa (2010).

Full size image

Discussion

High-attenuation areas in the lower crust of the continental plate

The high-attenuation area that is laterally extensive at a depth of 20 km can be divided into a region located around active volcanoes in the western part of the study area and a region located in non-volcanic areas. As high attenuation observed in the volcanic areas in western Kanto is comparable to that observed in volcanic areas, for example, in northeastern Japan (Nakajima et al. 2013) and Mexico (Chen and Clayton 2009), it is probably caused by high temperatures and magmatic fluids. In contrast, high attenuation in the lower crust of the non-volcanic areas cannot be due to a high-temperature effect, as this area has the lowest heat flow values (e.g., Tanaka et al. 2004) and the deepest cutoff depth (>25 km) of crustal earthquakes in the Japanese Islands (Omuralieva et al. 2012).

A key observation to interpret these non-volcanic, high-attenuation areas of the lower crust is that almost no seismicity occurs in regions characterized by high attenuation and low velocity (Figures 5 and 6). A plausible cause of the aseismic and high-attenuation lower crust is the localization of overpressured fluids with a fluid pressure to lithostatic pressure ratio of approximately 1. Such high pore pressures may no longer sustain significant regional shear stresses, which promotes aseismic deformation and enhances high attenuation. The enrichment of slab-derived silica in the lower crust (e.g., Breeding and Ague 2002) may also promote aseismic deformation. Abundant fluids, supplied from the subducting Philippine Sea slab over geological time scales due to slab metamorphism, are likely to play important roles in the promotion of a large-scale aseismic deformation in the lower crust.

The high-attenuation area observed at the corner of the mantle wedge corresponds to an area of low velocity and moderate-to-high Vp/Vs ratio, where serpentinization may occur (e.g., Kamiya and Kobayashi 2000; Matsubara et al. 2005). Assuming that high attenuation represents the serpentinized mantle, serpentinization is confined to a small region immediately below the continental Moho. Because seismic ruptures along the plate boundary can propagate to greater depths beyond the localized serpentinized mantle (e.g., Collings et al. 2012), comprehensive evaluations involving the slip deficit along the plate boundary and frictional properties of serpentine are required to conclude whether the serpentinized mantle wedge could act as a barrier to megathrust seismic ruptures.

High-attenuation mantle of the subducting Philippine Sea slab

The most notable observation in this study is the existence of the high-attenuation area in the Philippine Sea slab mantle beneath eastern Kanto (Figures 6a,b,c and 7a). Although high temperatures are considered as the dominant cause of high attenuation in the mantle (e.g., Faul and Jackson 2005), a thermal effect is considered unlikely for the origin of high attenuation within the subducting mantle. The high-attenuation mantle corresponds spatially to the proposed serpentinized slab mantle with low velocity (Figure 1b) and high Vp/Vs (Nakajima and Hasegawa 2010), and hence marked high attenuation is interpreted to be attributable to the serpentinized mantle with high water content, low viscosity (e.g., Hilairet et al. 2007), and small grain sizes (e.g., Mizukami and Wallis 2005). The wedge-shaped mantle of the Philippine Sea plate has been serpentinized prior to subduction (Kamimura et al. 2002), and serpentinization is considered to have been preserved after subduction because temperatures in the eastern portion of the Philippine Sea slab remain in the serpentine stability field (e.g., Iwamori 2000). Additional serpentinization has probably also occurred after subduction, given that fluids have been supplied by dehydration reactions occurring in the underlying Pacific slab. An important consequence of this interpretation is the enhancement of attenuation by serpentinization.

Figure 7
figure 7

Seismic attenuation and velocity structure inside the Philippine Sea slab. (a) Q p -1 and (b) P-wave velocity along a curved surface located 10 km above the upper surface of the Pacific slab. White stars represent small repeating earthquakes that occurred on the Pacific slab. Other symbols are the same as in Figure 1.

Full size image

Small repeating earthquakes along the upper surface of the Pacific slab (white stars in Figure 7) (Uchida et al. 2009) occur in areas that are overlain by low-attenuation and high-velocity mantle of the Philippine Sea slab and are absent below the serpentinized slab mantle (Figure 7). The results of deformation experiments performed on serpentine show that its viscosity is significantly lower than that of the major non-metamorphic mantle-forming minerals; thus brittle failures are unlikely to occur (e.g., Moore et al. 1997; Hilairet et al. 2007). Observations in this study are consistent with the interpretation made by Hilairet et al. (2007) in that interplate earthquakes do not occur when the plate boundary is in contact with serpentinized mantle.

Two M ~ 7 earthquakes in the Philippine Sea slab (red stars in Figure 7) are interpreted to have occurred as a result of right-lateral deformation along the western boundary of the serpentinized mantle (Nakajima and Hasegawa 2010). The western boundary of the observed high-attenuation mantle coincides with that of the serpentinized mantle (Figure 7a), suggesting that these two earthquakes occurred at the sharp boundary between the high-attenuation mantle to the east and the low-attenuation mantle to the west. Although it is inconclusive as to whether the western boundary of the serpentinized mantle corresponds to the thermal limit of the stability field of serpentine or whether it is controlled by the distribution of fluids, frictional properties and pore fluid pressures may change abruptly across the boundary. These observations highlight the importance of heterogeneity in the subducting mantle for the facilitation of intermediate-depth earthquakes.

Conclusions

This study characterizes seismic attenuation structure and discusses spatial relations between seismic attenuation and seismic activity beneath Kanto, Japan. The attenuation model yields additional insights into seismogenesis in this tectonically complex region. The main conclusions are summarized as follows.

  1. 1.

    The lower continental crust above the Philippine Sea slab shows high attenuation and is almost aseismic. Ductile deformation is likely to be dominant, associated with overpressured fluids supplied from the underlying Philippine Sea slab.

  2. 2.

    The corner of the mantle wedge in the continental plate may be serpentinized, but the extent of serpentinization is limited to a small region below the Moho.

  3. 3.

    The serpentinized mantle of the Philippine Sea slab shows high attenuation as well as low velocity, which strongly suggests that serpentinization enhances attenuation in the seismic frequency band.

The attenuation model obtained in this study will provide a practical constraint on the evaluation of realistic strong motions at the surface, which is crucial for the assessment and management of seismic hazards in the Tokyo metropolitan area.