Outer-rise normal fault development and influence on near-trench décollement propagation along the Japan Trench, off Tohoku

Distribution of outer-rise normal faults off Tohoku

The bathymetric and slope maps clearly show normal faults on the Pacific Plate seaward of the Japan Trench (Figure 1). Surface breaks are detected as far as approximately 120 km seaward of the trench axis with the strike of the faults parallel to the trench trend. Sixty kilometers landward from the most seaward surface breaks, the fault patterns become more intricate (Figures 2 and 3), presumably a reflection of these faults’ interacting with the pre-existing fabric of the Pacific Plate and each other as they propagate (e.g., Ranero et al. 2003). The pre-existing fabric appears to limit certain fault orientations and promote others, as active fault strikes coincide with the orientations of fracture zones and abyssal hill fabric (Nakanishi 2011). The southeast region of the survey area is slightly more complex due to potential petite spot volcanism (Site C of Hirano et al. 2008) making it difficult to determine the strike of basement-cutting faults because of more relief in the topography and regional sediment thinning (Figure 1). Fault interactions on the incoming Pacific Plate produce a heavily deformed and spatially complex fault pattern just seaward of the trench.The MCS data only extend approximately 75 km seaward of the Japan Trench but still provide insight into the beginning of faulting. Line D19 intersects both abyssal hill and trench-parallel fabrics (Figure 2). Although the geometry of line D19 is not ideal for imaging those fabrics (its line is not perpendicular to the strike of the fabrics), the seismic line does show continuous horizons above the buried abyssal hill topography, which formed at the spreading center. The continuity of reflectors over the abyssal hill implies that the fault associated with the abyssal hill either is not active or has not slipped enough to offset the youngest sediments. A few kilometers west of the abyssal hill (Figure 2B), a trench-parallel fault (green fault in Figure 2B) offsets the top of igneous crust (TIC) by approximately 30 m. Continued propagation of faults generates three sets of fault orientations: (a) parallel to the abyssal hills, (b) parallel to fracture zones, and (c) parallel to the trench. Faults with zig-zag surface traces containing sharp bends and irregular paths develop where the faults with differing strikes intersect (Figures 1 and 3). This indicates that trench-parallel faulting may control the early stage of brittle deformation with reactivation of pre-existing structures developing later.

Basement-cutting faults

We interpreted 199 normal faults that cut the TIC within the area of the seismic data survey. Throws are based on the vertical offset of the interpreted TIC. In each seismic line, we calculate cumulative throw at the distance to the trench of each fault by summing the throws of all faults seaward of that point (Figure 4). However, the seismic lines’ seaward-termination points are not constant, causing the region >30 km from the trench to be undersampled. Slopes of cumulative throw for each seismic line reveal deformation per distance. Linear regression of cumulative throw versus distance, using a maximum distance of 30 km from the trench to reduce sampling effects, indicates that there are small differences in cumulative throw between the south, central, and north regions. For seismic line groups (south to north) D02 to D06, TH03 to D13, and D15 to D19, the median absolute value of cumulative throw slope is 2.50°, 3.46°, and 3.66°, respectively. Maximum fault throw on a single fault is >500 m but more than half of the faults have throws of <100 m with the average offset being approximately 120 m. Of the 199 faults interpreted, 105 of them dip landward, whereas 94 faults dip seaward. The median number of faults increases towards the trench (Figure 4) with variations from line to line. There is a minor drop in the number of faults at a distance of 30 to 40 km from the trench. This is the result of the termination of multiple seismic lines near 30 km to the trench, which leads to low fault counts for that distance bin. Fault spacing averages approximately 2 km with a minimum and maximum spacing of 200 to 10 km. Analysis of fault throw within our survey region from available seismic (Tsuru et al. 2000) and bathymetry data (Iwabuchi 2012; 2013) is prone to two types of error: variation in sediment thickness in the region complicates the bathymetry-based analysis and limited number of pervious seismic lines in this area is insufficient to provide complete regional coverage.

Incoming Pacific Plate sediments

The Japan Trench margin has long been known for its relatively thin incoming sediment column of approximately 400 m from both seismic and DSDP drill cruises (e.g., Ludwig et al. 1966; Shipboard Scientific Party 1980). Our data provide new insight into regional variation in sediment thickness and internal deformation within the sediments.

Sediment thickness

We evaluated sediment thickness along a series of profiles, based on interpretations of the seafloor and TIC, using both regional and high-resolution lines to plot the stratigraphic thickness of the sediment column (Figure 1B). Thinning of the sediment section is expected above the fault surface at footwalls of normal faults (Groshong 2006) and should not be interpreted as stratigraphic thickness. We find a range of sediment thickness of 0 to 600 m along the incoming Pacific Plate. Thinning in the southeast is largely found near potential ‘petite spot’ volcanoes (Figure 1A; Site C of Hirano et al. 2008) and is likely related to the formation of these small cones, or to erosion around them. Elsewhere, thickness is generally related to initial deposition or reworking of the sediment. Sediment thickness varies from line to line but generally of increases towards the trench.Evidence for variation in sediment thickness and near-trench sediment deposition is found throughout the survey region. As an example, Line D02 in the southern part of the survey area (Figure 5B) illustrates the sedimentary section surrounding the trench axis. The base of the sediment column is marked as the blue dashed line, which has an apparent dip to the west. An unconformity at the top of the sediment section is discernible near the trench axis, with a near horizontal dip, and is marked with a green dashed line. This pair of markers implies that the original sediment column was tilted landward during bending of the subducting plate, followed by deposition of horizontal sediments in the trench. High-resolution line HD21A also images recent sediments that onlap older, landward-tilted strata in a developing graben adjacent to a normal fault, seaward of the trench (Figure 5C). Slight rotation of the lower horizons (in the blue region) indicates syndepositional faulting. Little faulting has occurred after this initial rotation, however, as the upper horizons are near horizontal. Seafloor displacement from this fault is small, indicating that either recent sedimentation has obscured any surface offsets or the fault currently is inactive.

Near-trench, sediment-restricted faults

Faulting and fracturing of the sediment column occurs along the outer rise. Dives with submersible Shinkai 6500 in the 1990s were the first to observe open cracks striking parallel to the horst and graben system at the outer rise of the Japan Trench (Ogawa et al. 1997). Our high-resolution seismic lines image similar structures. Lines HD26B and HD27B (500 m north of HD26B) show an incoming horst approximately 1 km seaward of the trench axis (Figure 6). Lithologic units were correlated from DSDP Site 436 (Shipboard Scientific Party 1980). Faults within the sediment column (units 1 and 2) exhibit offsets of approximately 20 m or less and are densely spaced, with populations of approximately 20 faults occurring within an approximately 2-km-long interval on Figure 6. These faults to do not appear to offset the top of the chert layer (unit 3). Unit 3 reflections are deformed internally, but the upper boundary of the unit is not disrupted. The fault pattern in unit 1 shows poor vertical fault connectivity. Faults do not appear to be confined to a specific sedimentary layer within unit 1. Increased sediment deformation around a nearby basement-cutting fault is common. There is no evidence for growth faulting along these sediment-restricted faults, indicating that the fault system is young. Landward (westward), towards a bounding basement-cutting fault, unit 1 reflections become discontinuous, indicating that the region is heavily deformed. Unit 1 thins in this discontinuous region, potentially from slumping of material into the graben to the west.

The deformation front along the Japan Trench

Our seismic lines reveal the subduction of horst and graben systems and their interactions with the landward trench slope. We define the deformation front as the seafloor trace where either chaotic reflections of the landward slope or the frontal thrust contact coherent, continuous reflections of the incoming plate (Figures 5 and 7). We used the bathymetry grid to interpolate the deformation front between seismic lines, correlating similar topographic features identified on the seismic lines. The deformation front is better defined on the high-resolution lines, which also reveal an imbricate thrust structure in the trench (Figure 7). In many areas, the regional lines display only chaotic reflections around the imbricate thrusts due to their lower resolution.Where a horst block is subducted beneath the base of the landward trench slope, the décollement is localized near the top of the chert layer at the top of the horst (Figure 7). At the seaward edge of the horst block, a low-angle, seaward-dipping reflection shows that the décollement drops down into the adjacent graben, where it localizes within the sediment column, approximately 10 to 20 m above the chert layer.North-south profiles aid in lateral interpretation between horst-graben systems. The high-resolution lines are concentrated over a single trench-graben system. However, strike line S13 images the adjacent section to the north showing both a subducted horst and graben system beneath the landward trench slope (Figure 8). We interpret the décollement as the surface between the continuous (blue) horizons beneath and the deformed (red) horizons above the plane. Across the main part of a horst or graben, approximately 100 m of sediments (blue region) are subducted with minimal variations along the profile, but adjacent to the bounding normal faults, a large amount of sediment that has slumped from the adjacent horst block is being subducted. Furthermore, the seafloor appears to follow the morphology of the TIC indicating that seafloor topography is controlled by the underlying horst-graben morphology.To investigate effects of the subducting plate on the deformation front, we plot the location of the deformation front and the underlying structures (Figure 9). The continuous bathymetric expression of the horst blocks allows us to confidently interpolate our seismic interpretations between the seismic lines. An oblique horst is currently at the deformation front in the southern region, at lines D02 and D03 (Figure 9). The distance from the trench axis to the deformation front is to a maximum of approximately 4 km. At Line TH04 (Figure 9), a graben is being subducted under the deformation front (Figure 8). Here, the deformation front is located further seaward and near the trench axis. Bathymetric highs within the graben reflect the vertical growth of anticlines from the imbricate thrust faults (Figure 7). From line TH04 to D11, the deformation front is located within a graben (Figure 9). However, there is a topographic low at the deformation front at line TH03. This change in topography indicates lateral variation in the vertical growth of the thrust system within a graben, which may be related to previously subducted structures. At line D13 to D15, the deformation front is located above a horst and further west than at line D11. The deformation front at line D17 is above a half-graben (Figure 5). At the northern end of the survey, line D19, the deformation front is on the landward edge of a horst (Figure 9). We generally find that subduction of a horst temporarily impedes seaward propagation of the deformation front.

Fault development

Earthquakes and bathymetric trends demonstrate that bending-related faults form at a maximum of approximately 120 km east of the Japan Trench axis (Obana et al. 2012; Nakanishi 2011). A simple subduction calculation with the Pacific Plate subducting at a rate of approximately 8 to 10 cm/year (DeMets et al. 2010; Apel et al. 2006; Argus et al. 2011; Niitsuma 2004) nearly orthogonally under Honshu, predicts the oldest normal faults about to be subducted to be approximately 1.2 to 1.5 Ma. Assuming steady slip on the normal faults, maximum throws of 500 m divided by the age of the oldest faults gives a maximum throw rate of 333 to 417 m/Ma. Faults with smaller offsets must be much younger or less active. The increase in the density of the faults westward, towards the trench, indicates that new faults continue to form as the plate nears the trench, which indicates that the assumption of steady-state deformation is incorrect. This overall pattern is similar to normal fault development in other trenches, such as the Middle America Trench (Ranero et al. 2003). Furthermore, average fault spacing of 2 km is on the same order as predicted in some modeling studies (Faccenda et al. 2009), although other models generate larger fault spacing of up to 10 km (Naliboff et al. 2013), which is the maximum fault spacing we observe. The increase in the number of faults towards the trench and the range of fault spacing suggests continuous formation of new faults occurs as the Pacific Plate approaches the trench rather than apportioning strain to only major, more widely spaced faults.

The fault traces in our survey area is similar to patterns developed in other locations and in analog models. Onshore three-dimensional (3D) normal fault propagation through basalt shows linkage of surface fractures that produce a fault with an irregular trace (Kaven and Martel 2007). Similar zig-zag fault traces have been generated in analog fault experiments where different stress directions were applied to the sediments (e.g., Henza et al. 2011). The 3D stress field of the Japan Trench outer rise is poorly known. A bend in the Japan Trench trend at 38° N will create 3D effects not accounted for in 2D models. Furthermore, coseismic rupture of thrust segments south and north of 38° N will create a heterogeneous stress field on the outer rise. This is the only region along the Japan Trench where topographic structures on the subducting plate lie collinear with the fracture zone strike (Nakanishi 2011). At approximately 30° from the trend of the northern part of the trench (Figure 1), the fracture zone trend is slightly more than the approximately 25° reactivation cutoff determined by Billen et al. (2007). However, that study considered only reactivation of abyssal hill trends, but this may still imply that fracture zones are still weak with a lower coefficient of friction than the surrounding rock. Following initial surface breaks parallel to the trench, large fault-tip stress concentrations may help reactivate the pre-existing fabrics to develop the fault trace geometry found in the survey area (Figure 1).

Branches, relays, and conjugate faults are found in many normal fault populations (Peacock 2002; Morley et al. 2007; Nicol et al. 1995; Walsh et al. 1999). The presence of such features in our study area indicates poor lateral fault connectivity. Regional north to south differences in cumulative throw (Figure 4) may be related to such lateral effects. Fault interactions, especially in areas with closely spaced faults, may also contribute to variations in slip distributions along a single fault (Martel and Shacat 2006). Therefore, our throw analysis should be viewed cautiously due to effects of nearby faults and location of our MCS lines within individual fault segments.

Incoming plate sediment thickness

Sediment thickness and lithology of the incoming plate plays a dominant role in development and growth of accretionary prisms. Accurate understanding of the geometry of the TIC and overlying sediments is vital to measuring sediment influx and fault throw. Although basement-cutting faults offset the TIC, the TIC is not always a continuous, strong reflector in between basement-cutting faults within the study area, as seen on the eastern section of line D02 (Figure 5A). This discontinuous nature of the TIC in our seismic lines is due to several factors: 1) seismic imaging through complex geology scatters ray paths, resulting in an inadequately imaged subsurface. In our study area, this is largely caused by to the rough seafloor and the chert layer overlying the TIC. The high-resolution survey data likely show greater scattering (due to the higher frequency content), as the regional lines image the TIC more consistently. 2) The velocity of the overlying chert creates little impedance contrast with the TIC. Although no formation velocity for DSDP Site 436 was measured for the chert unit, velocities of porcellanite and chert samples generally range from 2.6 to 4.8 km/s (Shipboard Scientific Party 1980), which may be near the velocity of the TIC. 3) Drilling at Site 436 did not reach the igneous basement, leaving the composition and alteration of the basement unknown near our survey location. Mineralogic analysis of oceanic basalts elsewhere reveals significant alteration with a large potential for fluid incorporation (Kameda et al. 2011), which may lower the acoustic impedance across the TIC. 4) The igneous basement may differ compositionally throughout the survey area. The seismic character of a sill or sheet flow may produce a smoother and stronger reflector than pillows and flows (Abrams et al. 1993), creating variations in reflector strength for the TIC. 5) Continued faulting along the outer rise will further reduce continuous reflections from the TIC. This can be from development of a new fault breach through the TIC, secondary faulting, or off-fault, nonelastic deformation, including fissure and cavity formation, as the fault tip propagates towards the surface (Holland et al. 2006; Martel and Langley 2006). Even with sections of discontinuous TIC, we are confident in our interpretation because we observe relative amplitude and frequency differences between the sediment and underlying basalt along with regions of a prominent TIC horizon.

Our results reveal variations of as much as 600 m in thickness of the sediments overlying the TIC (Figure 1B). Correlation of sediment thickness between regional lines was not possible because the seismic line spacing is too wide and there are complications imposed by variation in fault strikes. Basement topography is a dominant factor in variation of sediment thickness in two ways. 1) Older structures contribute to sediment variations. For example, the abyssal hill imaged in line D19 (Figure 2) has thinner sediments above it than on its flanks (Figures 1 and 2). Interpreted petite spot volcanic fields found in the southeast of the survey (Figure 1) have little to no sediment cover owing to their likely young age, based on ages of nearby fields (Hirano et al. 2008). Similarly, sediment overlying the tops of seamounts is generally much thinner than on the seamounts’ flanks. 2) Basement-cutting normal faults influence sediment thickness. This may be caused by erosion of topographic highs by bottom currents or by mass wasting (Figure 6) and filling of topographic lows, such as along graben systems (Figure 5). Sediment thickness generally increases towards the trench but with fluctuations as large as approximately 400 m on individual lines. No regional trends in sediment thickness are found in other directions. However, the large grid spacing between regional lines will mask small-scale trends. Because the Japan Trench accretes a large portion of sediment at the toe (Figure 7; Kodaira et al. 2012; Nakamura et al. 2013), sediment variations of 600 m may still affect prism morphology because they are a large percentage of the total sediment thickness. However, due to the fluctuations in sediment thickness and lack of regional trends, prism morphology and growth in our survey area owing to the incoming sediments is difficult to constrain.

Décollement propagation

Seismic sections of a near-trench subducting graben (Figure 7; Kodaira et al. 2012; Nakamura et al. 2013) reveal a low-angle, seaward-dipping reflector descending into the graben from the adjacent horst block. Imbricate structures sole into the same plane, which is thus interpreted to be the décollement. The mechanism for this décollement step-down into a local trench graben is poorly understood. Although landslides are a potential mechanism for creating the step-down (Strasser et al. 2013), we propose additional mechanisms for the step-down that involves the subduction and burial of an active normal fault (Figure 10). Surface breaching normal faults, like the majority of incoming faults, have a very different stress field than that of buried faults. Large horizontal and vertical tensile stresses occur at the fault tip and footwall of buried faults (Martel and Langley 2006). As the décollement propagates over a breached normal fault, it buries this fault under the prism. We propose the following situations may occur as a result:

Horst block subduction also affects the up-dip geometry of the décollement (Kimura et al. 2012) and potentially lateral décollement connectivity. Our regional lines image subducted horsts near the trench and show that the décollement ascends over the subducting horst along a fault ramp (Figure 5B). A seafloor bathymetric high above the ramp is an anticline caused by sediments being folded over the fault bend (Figure 9). Fault slip over this 20° dipping décollement fault ramp will create greater seafloor uplift than in sections of shallow dips. Furthermore, the sediment column overlying the décollement ramp shows no continuous reflections, thus indicates internal deformation (Figure 5). In the high-resolution lines, the décollement stops within a graben near the incoming horst block (Figure 7). Seafloor topography within this graben deepens to the south. This morphology change can be attributed to varying internal deformation and uplift along strike, in that the décollement propagates seaward to a horst where the prism topography steepens before creating a ramp over the horst. If décollement propagation does vary in horst-graben systems, subduction of intersecting or linked normal fault sections would create lateral décollement segments that are linked at depth.

Faults within the sediment column may play an additional role in décollement propagation. DSDP Site 436 (Figure 1C) is the only nearby site where sediments on the Japan Trench’s outer rise have been sampled; drilling there only penetrated to the chert layer (Shipboard Scientific Party 1980). At Site 436, a small layer of pelagic clay overlies the chert. In this deep water setting, the overall lithology is not expected to vary much laterally, allowing for a general correlation among our seismic lines. Ogawa et al. (1997) discovered cracks in the seafloor along the outer rise and proposed that a combination of gravitational slope instability and earthquake shaking for the cracks formation. These surface cracks may be from the early stages of sediment-restricted normal faulting at depth, imaged in the high-resolution survey (Figure 6). These sediment-restricted faults may have a basal sliding plane between the basement-chert, chert-clay boundaries, and/or within the upper unit, deforming the sediment along strata. The creation of these sliding surfaces before subduction may play a role during seismic rupture to the trench. The Japan Trench Fast Drilling Project (Integrated Ocean Drilling Program Expedition 343 and 343 T) drilled to the chert layer further landward (Figure 1) and found that the décollement is located within a thin band of pelagic clay (Chester et al. 2012), with localized deformation occurring within 5 m of this pelagic clay. Thus, this lithological unit may be a regional control on rupture to the trench (Chester et al. 2013), as the low coseismic shear stress and low coseismic friction of the clay may have allowed the 2011 Tohoku earthquake to rupture to the trench (Fulton et al. 2013; Ujiie et al. 2013). However, Expedition 343 drilled only above a horst block. Seismic results agree with the décollement’s location above the horst. However, within the graben, the décollement located higher in the sediment column (Figure 7). The pelagic clay within a graben may have a more dominant role in sediment sliding and sediment-restricted faulting (Figure 6) than for décollement propagation. Near-trench faults in sediment may play an important role in weakening the sediment before subduction.

The accretionary prism at the toe of the Japan Trench initially accumulates a major proportion of incoming sediment. Previous models indicated that all of the sediments within a graben were subducted, as the décollement propagated over the top of the entire graben sediment column (e.g., Hilde 1983). This would allow for significant sediment subduction. However, the décollement appears to initially scoop out the majority of sediments within a graben (Figure 7). Furthermore, both seismic and drilling (e.g., Chester et al. 2013) indicate that the décollement is near the base of the sediment column when located above a horst. These results reduce the amount of sediment available for subduction under the toe of the frontal prism. Variations in both the incoming sediment column thickness (Figure 1) and the amount of long-term sediment accretion cause potential difficulties in calculating the amount of sediment subducted (e.g., Clift and Vannucchi 2004). Additionally, instead of underthrust sediment-filled graben blocks being only vertically loaded, initial sediment accretion may affect overall sediment porosity due to the addition of both vertical and horizontal strains on the subducting sediment (e.g., Moore et al. 2011). This may affect the sediment dehydration further landward and, as a result, whether or not subduction erosion might occur at depth. Other water-rich sources may contribute more to the fluid budget including both the igneous basement and chert, where the dehydration range for siliceous sediments starts at approximately 40 km landward of the Japan Trench (Kimura et al. 2012). Sediment cracks and faults along with basement faults may provide necessary pathways for fluid transport and may play a vital role in hydration below the sediment column. These near-trench processes may influence different subduction processes occurring at depth at the Japan Trench subduction zone.