In the summer of 2000, an ocean bottom seismometer (OBS) array was deployed along the western end of the Nankai Trough, off Cape Muroto. The array crosses an area where a buried subducting seamount was identified under the Nankai accretionary prism sediments during a multichannel seismic (MCS) reflection survey. In this paper we report the first results of shear-wave splitting (SWS) analysis of the three-component OBS data. One interesting result is a peculiar anomaly in the behaviour of shear-wave polarizations observed near the subducting seamount. We attempt to explain the observed spatial pattern in shear-wave polarizations by complex structure of stress field in accretionary prism sediments around seamount. We conclude, that the spatial pattern of SWS consistent with expected stress perturbations around seamounts, but there are several other reasons that can result in a similar type of seismic anisotropy. Finally, we have a somewhat broader look at the spatial pattern of shear-wave polarizations observed with another OBS array near the active front of Nankai Trough.
The Nankai Trough subduction zone, off southwest Japan, is an active plate-convergent margin where the Philippine plate is subducting beneath the Eurasian plate. The subduction at the Nankai Trough is the cause of great interplate earthquakes with a recurrence interval of 100–200 yr, and with the history that can be traced back for over a thousand years (Ando 1975). The Nankai Trough is considered to be one of the best-suited convergent plate margins for studying interplate subduction earthquakes and the formation of accretionary prisms.
Another interesting feature of the Nankai Trough is the Kinan seamount chain situated in the centre of Shikoku Basin, with some of the seamounts subducting immediately beneath the accretionary wedge, being largely responsible for the deformation of the wedge (Park et al. 1999; Kodaira et al. 2000). Several cases of subducting seamounts have been reported at various active convergent margins around the world and there were a number of attempts to model the effect of seamount subduction on deformation of accretionary prism sediments and seafloor morphology (e.g. Lallemand et al. 1994; Domingues et al. 2000). Most of these efforts are focused on the deformation of the overriding plate and the forearc wedge. In some of these studies the subducting seamounts are viewed as seismogenic asperities, associated with seismic rapture and slip in the subduction zone (Cloos 1992). The seamount subduction can enhance seismic coupling related to the occurrence of large interplate earthquakes (Scholz & Small 1997). Recently, a high-resolution multichannel seismic (MCS) reflection survey was carried out for the first time across the western Nankai Trough to study the deformation pattern of an accretionary wedge by seismic methods (Park et al. 1999).
In this paper we employ methods and ideas of shear-wave splitting (SWS) analysis and seismic anisotropy to look at the stress state of accretionary wedge sediments near a subducting seamount. When a shear wave propagates through an anisotropic elastic solid, it splits into two nearly perpendicular polarizations, which travel with different velocities. The two most important parameters of SWS are the fast shear-wave polarizations and the timedelay between the two shear waves (Crampin 1981). The directions of fast polarization are relatively stable and approximately the same for seismic events with different source-to-receiver azimuth at any given location and, in most cases, are parallel to the maximum regional stress direction. These regionally consistent fast shear-wave polarizations have been observed worldwide for a variety of tectonic, geological and topographical environments. There are several physical mechanisms believed to be responsible for the origin of seismic anisotropy, which are different for upper and lower crustal and mantle rocks (Ando et al. 1983; Peacock et al. 1988; Kaneshima 1990; Savage 1999). It is generally agreed, that seismic anisotropy of most of the sediments is directly related to the level of fracturing and to the preferred orientation of cracks or fractures (Kaneshima et al. 1988). Whenever this preferred orientation is aligned with local principal stress directions, SWS can be used for interpretation of the stress state of crustal rocks by seismic methods.
There are several SWS inland studies on Shikoku Island, but an accurate offshore analysis of SWS became possible only with the advance of the three-component ocean bottom seismometer (OBS) technique. Below, we present the first results of SWS analysis of the three-component OBS data collected with two different OBS arrays in the western and eastern segments of Nankai Trough. One of these OBS lines is deployed over the area where a subducting seamount was identified during a MCS reflection survey (Park et al. 1999). An interesting result of our SWS analysis is an observed anomaly in the behaviour of shear-wave polarizations near the subducting seamount. We attempt to explain the observed spatial pattern in shear-wave polarizations as caused by a complex spatial structure of stress field in accretionary wedge sediments around a subducting seamount. To do this, we heavily rely on the results of 2-D numerical simulation (Baba et al. 2001). At this stage, we do not have enough reliable OBS records to prove unambiguously our suggestion about the connection between the stress perturbations around subducting seamount and SWS, but the observed spatial pattern of polarizations of leading shear waves seems to be consistent with this possibility.
In discussion we demonstrate some new results recorded by another OBS array near the active front of Nankai Trough and consider a problem of unique stress regime of Nankai Trough accretionary wedge from a more general geophysical position.
Several SWS studies have been carried out on Shikoku Island during the last decades, but reliable offshore records demonstrating SWS became possible only with the advance of the three-component OBS technique. For that reason, the present study focuses in the western (off Cape Muroto) and the eastern (off Kii Peninsula) segments of Nankai Trough (Fig. 1), where two OBS arrays had been recently deployed: the first one (six seismometers) during 2000 July to September off Cape Muroto and the second one (12 seismometers) during 2001 November to 2002 January off the Kii Peninsula. All digital recording OBSs were 4.5-Hz short-period seismometers, except for one precision measurement device (PMD) (Obana et al. 2001).
Fig. 2 shows the position of the OBS and the events for the two data sets. However, due to technical problems (noise, very shallow events, difficulties in determining seismometer orientations), analysis was restricted to four OBS off Muroto (S02, S04, S06 and S09) and six OBS off Kii (KM01, KM04, KM15, KM16, KM22 and KM24).
Onshore seismic stations have the advantage that the orientations of the instrument axes can be determined directly. However, for our free-fall and pop-up type OBS, the orientations of the two horizontal components were unknown and needed to be corrected for true north and east. This was done in the following way: for each event with epicentral distance/depth ratio >1 (to assure sufficiently large horizontal P-wave component) the P motion on the horizontal plane was inspected. When the polarization direction of the P onset could be identified, the event was located with relation to pseudo-backazimuth Nsta. Subtraction of the above angle from the event backazimuth gives the position of Nstawith respect to North. Finally, the original seismograms were rotated to the backazimuth angle Nsta (Table 1 and Figs 3a and b).
The hypocentral determination was performed by a grid-search method, in which a hypocentre was searched for in a 3-D volume, and the residuals of relative P- and S-arrivals between OBS's and S−P times at each OBS were calculated. The hypocentres were considered to be determined with an accuracy of about 1 km in the horizontal and 3 km in depth (Obana et al. 2001).
The steep velocity gradient under Nankai accretionary wedge caused by the poorly consolidated water-saturated sediments of the ocean floor results in large vertical P components for the majority of events, whereas the horizontal P onsets for many seismograms were only marginally above noise level. For that reason we examined as many local and far-field events with different azimuths as possible.
The difficulty in the determination of seismometer orientation for the OBS is counterbalanced by the advantage of not dealing with the problem of shear-wave window (SWW). For inland seismic stations shear-wave arrivals must be within the critical cone or shear-wave window, for shear-wave seismograms not to be disturbed byS-to-P conversions at the free surface (Booth & Crampin 1985). This severely restricts the number of events that can be reliably analysed for inland seismic stations. However, with OBS, this restriction does not apply for short-period arrivals from local and regional earthquakes. As shear waves arrive almost vertically to the seafloor, we do not expect serious distortions even for events well outside the conventional SWW. And indeed, examining events for S09 and S04 outside the geometrical SWW, the polarization directions did not show any significant variation from the events well inside the critical cone.
SWS has been measured by using a conventional, visual method (Liu et al. 1997; Volti & Crampin 2003). Only clear and impulsive onsets were used, which makes the identification of the first shear-wave arrival unambiguous. As is usually the case, only a small number of the impulsive offsets with unambiguous fast directions allowed the second (slow) arrival to be reliably identified and measured. The steps followed for a typically analysed event are shown in Fig. 4.
In Fig. 5, the line OBS1 to OBS98 is a wide-angle seismic profile (Kodaira et al. 2000, 2002) along which a 50-km-wide subducted seamount extending down to 25 km depth, has been discovered in the off Muroto accretionary wedge (the large ellipse). A MCS reflection survey (Park et al. 1999) also identified a smaller seamount nearer to the subduction front (the small ellipse), 15 km wide and 2 km high, extending down to 7–8 km depth. Fast shear-wave polarizations were calculated for S02, S04, S06 and S09. The onshore station, KTG (IFREE report 2002) located to the N of the profile (SE of Shikoku) is also shown for comparison.
Fig. 5 reveals several other features: Towards the two ends of the seismic profile, KTG and S09 show fast polarizations, which are consistent with the regional near E–W compression regime believed to be predominant in the area through the westward subduction of the Pacific plate at the Japan trench (Kaneshima et al. 1988; Ishikawa 1995). The other three stations (S06, S04 and S02) show more complicated patterns: S04, midway between the two seamounts, has the majority of events striking approximately N45°E; whereas the events recorded by S02 and S06 have a variety of orientations. The average direction for S06 is approximately parallel to the profile (NNW—SSE), whereas for S02, this profile-parallel tendency becomes obvious only for events towards the North. To the south of S02, directions are more scattered and vary between trench-parallel and E–W directions. The complex polarization pattern in this area and its possible relation to the stress perturbation caused by subducting seamount will be discussed in Section 5.
Time delays and fast polarizations versus depth are plotted in Figs 6(a) and (b), respectively. Time delays are generally less than 0.2 s, which is typical for crustal earthquakes onshore and increase generally with depth. S04 shows the smallest values of time delays (mostly <0.1 s) and for S02 time delays were observed with values up to 0.2 s. S09 is scattered all over the range, while for S06 values are quite small (<0.1 s). We can conclude that fast shear-wave polarizations in the study area do not show any clear dependence on depth.
This absence of a polarization—depth relationship indicates strongly that the observed SWS is due to the seismic anisotropy of accretionary wedge and, consequently, the observed anomalies in fast shear-wave polarization can be caused by the deformation of accretionary wedge sediments by subducting seamount.
Stress Pattern Around Seamount
The results of previous section show that there is an observable level of anomalous perturbation in the polarization directions of first S-wave arrivals in the vicinity of the seamount. This is particularly evident for S06 station, where the first arrivals are aligned almost perpendicular to the maximum regional stress field. In this section we attempt to explain a possible reason for these anomalies invoking a perturbed stress pattern around a subducting seamount. Fig. 7 show an expected stress field around seamount in the vertical plane along the direction of plate motion as given by 2-D numerical simulation of subducting seamount (Baba et al. 2001). Originally, the simulation was performed to explain a possible scenario for the generation of a thrust fault imaged by a multichannel seismic reflection survey at the seaward flank (right flank, as shown in Fig. 7) of the Muroto seamount, which is subducting under the Nankai Trough accretionary prism (Park et al. 1999). The distribution of maximum incremental shear-stress value is shown in Fig. 7(a), meanwhile the minimum and maximum principal stress components are indicated with arrows in Figs 7(b) and (c), respectively. It should be noted that the vertical gravitational (overburden) component of stress field has been subtracted from the presented results. Thus, the stress pattern as seen in Fig. 7 is a perturbed stress field caused by the motion of the seamount through accretionary prism sediments. For our purpose we can roughly identify five regions with qualitatively different stress-field characteristics (Fig. 7d):
Regions 1 and 5 on both flanks of the seamount, where the stress perturbation is relatively small and local stress field is close to the regional stress field;
Region 2, where there is a strong subhorizontal maximum (compressional) stress component;
A relatively narrow region 3, where the subhorizontal stress component is small, but there is a strong extensional subvertical (perturbed) stress component (a possible location of thrust fault); and, with less degree of certainty, some area at the rear (seaward) flank of seamount where there is an additional increase in subhorizontal stress component.
We shall emphasize that the stress pattern in Fig. 7 is a result of 2-D simulation and, therefore, can be taken as a reasonable approximation for a stress field around 3-D seamount only in the nearest vicinity of the vertical symmetry plane (x–z plane). Out of the symmetry plane, additional stress perturbations in y-direction can significantly distort the stress pattern for all principal stress directions. Nevertheless, for seismic rays close enough to the vertical symmetry plane (along the direction of plate motion) we can follow our (relatively coarse) classification of regions with distinctively different stress characteristics.
If we assume that there is a strong regional horizontal stress in y-direction far enough from the seamount, which is still smaller in absolute value then the perturbed subhorizontal stress component in x-direction near the seamount, we can expect the following polarization pattern of shear-wave first arrivals:
Parallel to y-directions for regions 1 and 5 (as ‘seen’ by the fast shear waves at stations KTG and S09 in Fig. 5);
Parallel to the plate motion direction (arguably, there is some inconclusive trend in this direction for a number of S02 events in Fig. 5);
Parallel to y-direction or highly variable in the region 3 (station S04 in Fig. 5);
Parallel to the plate motion direction in region 4 (as seen by station S06).
It seems that our hypothetical scenario is in a reasonable agreement with the observed polarization pattern. Unfortunately, a very limited number of events recorded by ‘critical’ for our arguments stations S02 and S06 renders it difficult to relate seamount stress perturbation and S-wave polarization patterns with a higher degree of certainty.
It is difficult to estimate the effect of off-plane 3-D stress perturbations in absence of appropriate 3-D numerical simulation of seamount subduction. The seismic events recorded and processed at the stations S02, S04, S06 and S09 are reasonably close to the symmetry plane along the plate motion direction. This alone, of course, can not guarantee that the off-plane stress perturbations are negligible. A wider picture of shear-waves arrivals could have resolved this problem. There is, however, another complication which is clearly seen in Fig. 7. The principle stress axes are pronouncedly tilted in x–z vertical plane. This can strongly affect the polarizations of first arrivals, particularly, for non-vertical seismic waves with significant initial offsets. We address this question using a very simple model to compare the effects caused by stress-magnitude and stress-orientation variations.
In this section we use the ANISEIS seismic modelling package (Taylor 2000) to calculate synthetic seismograms for a simple anisotropic model shown in Fig. 8(a). There are three layers of sediments sandwiched between the top half-space of sea water and the bottom half-space representing a subducting ocean plate. The upper layer has parameters of poorly consolidated ocean-bottom sediments, meanwhile the lower layer of accretionary prism sediments with a thickness 2.5 km is anisotropic, with a type of elastic anisotropy consistent with a triaxial differential stress field under consideration (see below). We use a horizontal point-force source with orientation 45° from radial so that seismic energy is distributed more or less evenly between the radial and transverse components of S wavefield. We assume an average source depth of 20 km, with two possible offsets 5 and 11.55 km corresponding to the straight-line incidence angles of 15° and 30°, respectively, and take a 30° step in azimuth around the station (Fig. 8b). According to our analysis in previous section, the presence of a seamount may have two separate effects on the triaxial stress field.
Change in the relative values of horizontal principal stresses
The effect of triaxial stress on effective elastic anisotropy of fluid-saturated sediments can be modelled using a simple code for anisotropic poro-elastic model (APE) (Zatsepin & Crampin 1997) together with ANISEIS. In essence, the model computes an effective stress-induced elastic anisotropy of randomly fractured fluid-saturated rock, when the evolution of the distribution function of cracks and/or fractures over orientations is governed by applied triaxial stress. The APE model is scale invariant (the dominant size of cracks or fractures is not specified, provided their size is much less then a characteristic seismic wavelength) and, in the simplest form, has only one model parameter, dimensionless crack or fracture density.
Figs 9(a) and (b) show the synthetic seismograms corresponding to the relative values of principal stress components as expected in regions 1, 5 (σxx : σyy : σzz= 1 : 2 : 6) and regions 2, 4 (σxx : σyy : σzz= 3 : 2 : 6) of Fig. 7, respectively. The seismograms are computed for 5 km offset along 0° azimuth. The polarizations of fast S wave follow strictly the direction of specified maximum horizontal stress field. When the principal horizontal components of stress field are equal (σxx=σyy) the polarization of the first arrival is influenced by the offset and the azimuth, as indicated in Fig. 9(c). This can be one of the reasons for a wide range of polarization directions observed by station S04.
As an additional useful result, the modelling demonstrates clearly the near-vertical character of seismic wave arrival to the seafloor even for 11.5 km offset.
Note, that in this model σzz stress component is precisely a vertical principal stress component and, consequently, our principal-stress reference system is not tilted in x–z vertical plane.
Tilting of maximum horizontal stress caused by seamount motion
In the perturbed stress field around the subducting seamount, principal (subhorizontal) stress component σxx is tilted with respect to the Earth's natural reference system (see Fig. 7). The simplest way to evaluate possible consequences of this effect is to rotate the principal symmetry axes of anisotropic layer around y-axis in x–z vertical plane. The synthetic seismograms for a tilt angle of 15° with the source at 0° azimuth are shown in Figs 10(a) and (b) for two offsets of 5 and 11.55 km, respectively. The effect of tilted stress in our initial model was so small that we have to double the thickness of anisotropic layer to see clearly the splitting between the radial and transversal components of S wavefield. Interestingly, the polarization of the fast shear wave depends on the chosen offset value. Obviously, for relatively small offsets no change in polarization pattern occurs. However, for 11.55 km offset, an exchange (flip) of polarizations occurs at certain azimuths. Similar results were observed for 30° tilt. Figs 11(a) and (b) show the dependence of polarizations of first S-wave arrival on the azimuth of seismic source for two different tilt angles. It is clear that prominent tilting of stress principal axes in x–z plane suggested by 2-D numerical simulation (Fig. 7) can obscure the interpretation of observed polarization pattern at given station, when seismic shear waves arrive from a wide range of azimuth directions.
We have suggested that SWS observed on all stations of OBS line is mainly due to the seismic anisotropy of accretionary wedge sediments. Our main argument is in the absence of a polarization—depth relationship, which is a strong indication that the observed SWS is due to the seismic anisotropy in the accretionary wedge. If this is the case, the observed anomalies in fast shear-wave polarization can be a result of deformation of accretionary wedge sediments by subducting seamount. This suggestion is supported by the fact, that the pattern of fast shear-wave polarizations as recorded by remote station S09 is similar to the pattern of shear-wave polarization recorded by inland station KTG. The predominant direction of these polarizations is parallel to the regional maximum stress component.
There are, however, several complications. Meanwhile the polarizations at stations S02 and S04 correspond to the area on top and very close to the big seamount and, consequently, may be related with some reasonable degree of confidence to the perturbation of stress field due to the subduction, the same cannot be said for the station S06. This station is located very close to the small seamount (see Fig. 5), which can, nevertheless, perturb local stress field. Unfortunately, we have very few reliable events in this interesting area. As the earthquakes recorded by S06 are the deepest among our events and the polarizations strike almost N–S, it is possible that they can reflect seismic anisotropy in the subducting plate (Kaneshima & Ando 1989) rather then in the accretionary wedge (see Fig. 12 for the location of events from the lower oceanic crust and the uppermost mantle, as recorded by another station S02).
The anomalies in SWS can also be caused by a complex 3-D geometry of deforming layered sediments, which can be clearly seen in time-migrated MCS reflection profiles (see Fig. 2 in Park et al. 1999). To positively resolve this problem we need additional SWS data and a better understanding of the physics of seamount motion through the pile of accretionary sediments.
In Fig. 13, fast polarizations calculated in a similar way for the off Kii segment of the Nankai Trough are shown together with the western sector data discussed thoroughly above. The elongated ellipse just south of Kii peninsula delineates the location of a formidable asperity caused by the 1944 Tonankai earthquake on the SW rim of Nankai Trough. This is a single asperity with a length scale of 100 km and a depth of 20 km from top of the plate. Its shape is reflected to some degree at the sea bottom topography and it is likely that it has been built from the sequence of fault motions in the area (Kikuchi et al. 2002). Events from the nearby stations KM22 and KM24, recorded within and near the edges of the asperity, show a similar flip of polarizations as seen by S06 station near the off Muroto seamount.
Apart from the anomalies in fast shear-wave polarizations observed in the vicinities of the asperity, the results shown in Fig. 13 are consistent with the E–W and NW–SE compression regimes, believed to exist in SW Japan due to the interaction of the Pacific and Philippine plates. The events at the far end of Shikoku Island, recorded by KM16, show on average the expected E–W direction of polarization. Beneath the Kii channel the pattern of fast shear-wave polarizations seems to be particularly complicated. It is known that the plate subduction gradually changes from low-angle to steep-angle subduction just beneath the Kii peninsula (Yoshioka 1991). The majority of events recorded by KM16 come from depths down to 100 km, whereas a number of events with E–W fast shear-wave polarizations are shallow with depth less then 10 km.
Although the central part of Shikoku Island is controlled by the compression regime parallel to the Median Tectonic Line, the situation in the active accretionary wedge in front of the island is different due to the low strength of the sediments and the high pore-fluid pressure. In this situation, the maximum principal stress in accretionary wedge can deviate to the orthogonal direction (Wang 2000). As indicated by numerous thrust and fold structures, and the borehole breakout experiments off the subduction tip (ODP 2002), in the accretionary prism the maximum compressive stress is aligned in the direction of plate convergence. This agrees with the results of SWS analysis in the Eastern Nankai (stations KM04 and KM15, and the majority of events recorded by KM01). However, towards the west, the polarizations of leading shear waves change to a nearly E–W direction parallel to the Trough (some events at KM01 and predominant direction for S09).
To explain the above changes, one can accept a unique state of stress, which exists in the Nankai Trough (Ukawa 1982). Although the overriding plate is generally under E–W compression, the subducting plate is under lateral extension along its strike. Accordingly, the subducting plate is forced to wrap downward, being constrained laterally at two arc-arc junctions, one between Ryukyu and Nankai–Suruga arcs and the other between Sagami and Nankai–Suruga (Figs 1 and 13, points A and B, respectively). The development of lateral extensional stress may eventually cause occasional ‘tearing’ of the subducting plate, which relaxes the lateral extensional stress and, subsequently, the regional stress may reverse locally into E–W compression. This can be more prominent in the area just in front of the Kii channel, where the change of stress due to the different subduction angle is possible.
The Nankai Trough is one of the best-suited convergent plate margins for studying interplate subduction earthquakes and the formation of accretionary prisms. We invite our reader to have a second look at Fig. 13 showing a complex but systematic spatial pattern of fast shear-wave polarizations recorded by OBS arrays. We believe it is clear that there is a significant amount of valuable information hidden in the SWS records. New results can bridge gaps left on the map of seismic anisotropy of Nankai Trough and bring some invaluable insight in the physics of convergent plate margins.
We are deeply grateful to editor of GJI Raul Madariaga, and an unknown to us referee for sympathetic comments on the first rather incomprehensible version of the paper, which we ourselves found very difficult to understand after the time taken for reviewing. We also thank Toshikaka Baba and Shuichi Kodaira for providing us with their beautiful, colour pictures.