Shear strength of sediments approaching subduction in the Nankai Trough, Japan as constraints on forearc mechanics

Authors


Abstract

[1] The mechanical behavior of the plate boundary fault zone is of paramount importance in subduction zones, because it controls megathrust earthquake nucleation and propagation as well as the structural style of the forearc. In the Nankai area along the NanTroSEIZE (Kumano) drilling transect offshore SW Japan, a heterogeneous sedimentary sequence overlying the oceanic crust enters the subduction zone. In order to predict how variations in lithology, and thus mechanical properties, affect the formation and evolution of the plate boundary fault, we conducted laboratory tests measuring the shear strengths of sediments approaching the trench covering each major lithological sedimentary unit. We observe that shear strength increases nonlinearly with depth, such that the (apparent) coefficient of friction decreases. In combination with a critical taper analysis, the results imply that the plate boundary position is located on the main frontal thrust. Further landward, the plate boundary is expected to step down into progressively lower stratigraphic units, assisted by moderately elevated pore pressures. As seismogenic depths are approached, the décollement may further step down to lower volcaniclastic or pelagic strata but this requires specific overpressure conditions. High-taper angle and elevated strengths in the toe region may be local features restricted to the Kumano transect.

1. Introduction

[2] In subduction zones where large sediment prisms develop, incipient thrusting at the deformation front defines the plate boundary décollement, which partitions accreted and underthrust sediment. Most of the plate convergence is accommodated by this feature, which eventually becomes the location of megathrust earthquakes. However, although large magnitude earthquakes cannot nucleate at shallow depths, coseismic slip can propagate all the way to the trench as evidenced by the 2011 Mw = 9.0 Tohoku earthquake [Ito et al., 2011; Kodaira et al., 2012]. Large slip in the proximity of the trench is also known to generate very large tsunamis, even if the rupture is slower than usual [Polet and Kanamori, 2000]. Identification of the lithologic unit that hosts the plate boundary fault zone is of primary importance, because the geomechanical properties of the constituent sediments control the slip behavior of major faults.

[3] The Nankai Trough subduction zone offshore Japan is known for a long history of great Mw ≥ 8 earthquakes [Ando, 1975; Rikitake, 1976] and has been the subject of extensive scientific drilling, primarily on three borehole transects arranged from southwest to northeast known as the Ashizuri, Muroto, and Kumano transects (Figure 1). These three transects are located within 300 km of each other but comparing structurally similar regions of the wedge along each transect reveals that the taper angle can be significantly different, implying corresponding differences in strength both within the wedge and on the basal décollement [e.g., Kimura et al., 2007]. The décollement was penetrated during drilling at the prism toe on the Muroto transect, but was not reached on the Ashizuri transect. Drilling on the Kumano transect is currently active as part of the Nankai Trough Seismogenic Zone Experiment (NanTroSEIZE), but depths where the décollement was expected to be located were not reached during expeditions to the toe region [Kinoshita et al., 2009]. The exact stratigraphic location of the plate boundary fault near the trench on the Kumano transect, and how its mechanical properties evolve with progressive subduction, therefore remain open questions.

Figure 1.

(a) Map of the Nankai area showing location of drill sites C0011 and C0012 as well as the rupture areas (dashed boxes) and epicenters (stars) of the 1944 Tonankai and 1946 Nankaido earthquakes (modified from Kimura et al., [2008]). (b) Spliced composite seismic reflection data along the Kumano transect in Figure 1, showing IODP Sites C0006, C0007, C0011, and C0012 (Modified from Saito et al., [2010], seismic data from Park et al., [2008] and Moore et al., [2009]).

[4] Because major faults form as planes of weakness, they are expected to develop in strata that is either intrinsically weak, e.g., due to the presence of clays, or weak due to elevated pore pressures that result in low-effective shear stresses. This is supported by observations from ocean drilling, for example in the Barbados subduction zone where the décollement localizes in a smectite-rich package [e.g., Vrolijk, 1990; Deng and Underwood, 2001]. For the Kumano transect, most friction studies have focused on the behavior of major fault zones such as the megasplay and frontal thrust [e.g., Ikari et al., 2009a; Ikari and Saffer, 2011; Tsutsumi et al., 2011; Ujiie et al., 2011], however, these studies utilize residual friction values using remolded samples. In the case of incipient plate boundary fault formation at the toe, the peak strength of previously undeformed, intact sediment is the key parameter, which is sensitive to lithologic variation but also other factors such as consolidation history and diagenesis progression. Here we measure the peak (maximum) as well as residual shear strength of intact core samples from each major lithologic unit of presubduction sediments at in situ conditions. The entire column of sediment approaching the Nankai Trough was successfully recovered from two Kumano transect reference sites (Sites C0011 and C0012, Figure 1) that were drilled seaward of the trench, providing a complete set of possible lithologic units within which the plate boundary fault may form. We incorporate these measurements with observations of forearc geometry from seismic reflection data in critical taper models (or “Coulomb wedge models”) in order to constrain the location of the plate boundary fault and its strength relative to the prism [Davis et al., 1983; Dahlen, 1990]. Using residual friction measurements at elevated effective stresses, we discuss implications for the plate boundary décollement further landward.

2. Geologic Setting

2.1. Nankai Trough Subduction Zone

[5] Off the coast of SW Japan, the Nankai Trough marks the boundary between the landward Nankai accretionary complex, and the seaward Shikoku basin. The subduction zone is formed by convergence of the Philippine Sea plate and the Eurasian plate (or Amurian microplate) at a rate of ∼4–6.5 cm/yr [Seno et al., 1993; Miyazaki and Heki, 2001] (Figure 1). Scientific drilling has focused on this region by targeting the rupture area of the 1944 earthquake, as inferred by tsunami and seismic waveform inversions [Tanioka and Satake, 2001; Kikuchi et al., 2003]. During the first stage of the multiexpedition Nankai Trough Seismogenic Zone Experiment (NanTroSEIZE), several drill sites penetrated two major fault zones at shallow depths (< ∼1 km): the frontal thrust region and a major out-of-sequence thrust fault (megasplay) as well as a site landward in the Kumano basin designed to ultimately reach the seismogenic zone [Tobin and Kinoshita, 2006; Kinoshita et al., 2009]. These sites are located offshore the Kii Peninsula, known as the Kumano transect. Earlier transects drilled during Deep Sea Drilling Project (DSDP) and Ocean Drilling Program (ODP) are located to the southwest of the Kumano transect, these are known as the Ashizuri and Muroto transects (Figure 1). Presubduction reference Sites 1177 (Ashizuri) and 1173 (Muroto) for these transects were drilled during ODP Leg 190 [Moore et al., 2001b].

[6] The accretionary prism along the Kumano transect has a taper angle (décollement dip plus surface slope) of ∼6–16° within 40 km of the trench, which is much wider than along the Ashizuri (∼5–8°) and Muroto transect (∼2–9°) [Taira et al., 1992; Moore et al., 2001a; Kimura et al., 2007; Moore et al., 2009]. Due to the presence of a fossil spreading center, heat flow seaward of the trench on the Muroto transect is high (up to ∼200 mW/m2) compared to the Ashizuri and Kumano transects (∼100–130 mW/m2) [Yamano et al., 1992; Yamano et al., 2003; Kinoshita et al., 2008]. Seismic reflection data reveal complex structure in the near-trench region on the Kumano transect [Moore et al., 2009] (Figure 1). Here the high-taper angle and steeply dipping seafloor contrast with low near-trench taper angles in the Muroto and Ashizuri transects. The main frontal thrust appears as a strong reflector in seismic reflection data and dips 7.6°, with its trace appearing to outcrop in the trench. Interpretation of seismic data in combination with magneto- and biostratigraphy of recovered sediment cores suggest that the trench wedge has been overridden ∼6 km by the frontal thrust since 0.44–0.78 Ma [Screaton et al., 2009]. The main frontal thrust was penetrated at 711.5 mbsf by LWD at Site C0006 during Expedition 314 and was sampled by coring at Site C0007 during Expedition 316 [Kinoshita et al., 2009]; however, drilling just below the fault was hampered by borehole instability caused by thick packages of unlithified turbidite sands that define the trench wedge facies. A second, more shallowly dipping strong reflector that can be traced within the incoming stratigraphy seaward of the trench is located below the trench wedge sediment where drilling was terminated. This deeper reflector may be analogous to the décollement or protodécollement on the Muroto transect, which is located within Lower Shikoku Basin (LSB) facies sediments [Moore et al., 1990; Moore et al. 2001a, 2001b]. It is possible that either the main frontal thrust or the strong LSB reflector could be the incipient décollement, because it has been noted that sediments below the frontal thrust appear to be deformed and thus accommodate strain [Expedition 314 Scientists, 2009; Moore et al., 2009]. The regional stratigraphy of the toe appears to be influenced by a protrusion of oceanic crust that is observable further landward in seismic reflection images. In the outer wedge, the décollement appears to coincide with the LSB facies until the transition zone ∼25–35 km landward from the trench, where it has been argued that it steps down to the sediment-basement interface [Park et al., 2002; Kimura et al., 2007].

2.2. Kumano Transect Reference Sites C0011 and C0012

[7] A primary objective of IODP Expeditions 322 and 333 was to establish the Kumano transect reference drillsites, C0011 and C0012 (Figure 1). Site C0012 is located on the crest of a bathymetric high known as the Kashinosaki Knoll [Ike et al., 2008], while Site C0011 is located landward on the flank of the knoll (Figure 1b). Because of the topographic difference, the deposits are significantly condensed at Site C0012, where the entire sediment section was recovered and the basaltic basement was reached at 538 mbsf (meters below seafloor). At Site C0011, coring did not recover a complete section but reached 881 mbsf during Expedition 322 [Saito et al., 2010], and logging-while-drilling (LWD) reached 952 mbsf during Expedition 319 [Saffer et al., 2010] and terminated approximately 100 m above the basement.

[8] Although recovery of the sedimentary section from Site C0011 over the course of Expeditions 322 and 333 was incomplete, lithostratigraphic units were successfully correlated with C0012 [Henry et al., 2012a]. The major lithologies of Sites C0011 and C0012 are hemipelagic deposits with volcanic ash (Unit I), alternating silty claystone and volcaniclastic tuffaceous sandstone (Unit II), hemipelagic deposits lacking in ash (Unit III), a turbidite facies (Unit IV), and a volcaniclastic unit (Unit V); all are Miocene age or younger. Unit I is interpreted to be the eastward extension of the Upper Shikoku Basin facies (USB). Similarly, the combined Units III and IV represent the Lower Shikoku basin facies (LSB). These facies, along with the volcaniclastic unit, were originally identified at the reference sites for the Ashizuri (Site 1177) and Muroto transects (Site 1173) [Moore et al., 2001b]. Two major sediment packages that were not observed at the Ashizuri or Muroto reference sites but were found at the Kumano transect sites are Unit II, the volcaniclastic sandstone unit between the USB and LSB designated the Middle Shikoku Basin facies (MSB), and red pelagic claystone overlying the basaltic basement at Site C0012 [Saito et al., 2010].

[9] Sites C0011 and C0012 are both characterized by very high-clay mineral content of ∼70%. Lower values are associated with siliciclastic and volcaniclastic turbidites. The dominant clay mineral is smectite, which generally comprises ∼45% of the bulk sediment at both sites. Smectite content is relatively constant with increasing depth at both sites, however at Site C0012 it is lower within USB and increases toward the MSB [Underwood and Guo, 2012]. The amounts of both total clay and smectite are significantly higher for these sites compared to the reference Sites 1173 and 1177. At Site 1173 total clay content is only 40–50% in the USB and generally ∼60% in the LSB, at Site 1177 clay content gradually increases from ∼50 to 65% downsection [Moore et al., 2001b]. Smectite in the bulk sediment is only ∼10–20% in the USB at both Sites 1173 and 1177; however in the LSB, smectite content increases ∼40–50% at Site 1177 compared to 20–30% at Site 1173 [Steurer and Underwood, 2005]. Despite similar amounts of total clay minerals, the lower smectite content at the Muroto transect Site 1173 compared to Sites C0011 and C0012 is likely due to elevated heat flow in the area, which has facilitated the thermally driven reaction of smectite to illite [Steurer and Underwood, 2005; Underwood, 2007].

3. Shear Experiments

3.1. Methods

[10] We conducted two sets of experiments. In the first, we measure shear strength using intact samples under in situ effective stress conditions using a single-direct shear apparatus [Figure S1(a), and Table S1 in the supporting information].1 In situ effective vertical stresses (σv') were calculated from shipboard bulk density measurements made during IODP Expeditions 322 and 333 assuming hydrostatic pore pressure [Saito et al., 2010; Henry et al., 2012b], and the effective normal stresses in these experiments was designed to match these values (Table S1). The samples for these experiments were trimmed from whole round cores from Sites C0011 and C0012. The sample cell consists of two flat-lying, smooth steel plates that hold a cylindrical sample 56 mm in diameter and at least 20 mm in height. The sample holder is flooded with a 3.5 wt% NaCl solution and normal load is then applied vertically (parallel to the cylindrical axis), with the sample being allowed to freely drain through porous metal frits in communication with an open pore fluid reservoir. Because the device is not equipped to measure pore pressure, we assume that excess pressure has dissipated when the sample height reaches a constant value under load. Shear stress, measured by inducing relative displacement between two near-frictionless plates on a plane perpendicular to the cylindrical axis, commonly exhibited a peak followed by a decrease to a residual value [Figure S1(b)]. These experiments were conducted at effective normal stresses of < 3 MPa, and at a constant displacement rate of 0.5 µm/s [see Ikari and Kopf, 2011 for further details]. In this set of experiments, we also measured the cohesion c in order to assess the contribution of diagenetic cementation as a source of sediment strength. This is measured directly by shearing the sample after removing the normal load so that σn' = 0 (equation (1)), either before or after shearing under normal load [Ikari and Kopf, 2011]. These measurements were only made during low-stress single-direct shear experiments, because these tests require intact samples and also because the geometry of the biaxial apparatus used in the second set of experiments is not conducive to measuring cohesion.

Figure 2.

Lithostratigraphy (left column), measurements of maximum and residual shear strength (middle column) and measurements of maximum and residual apparent friction (right column) for intact samples at in situ effective stress conditions (right column) for Site C0011. The lithostratigraphic column is modified from Henry et al. [2012a].

[11] In the second set of experiments, we utilize the double-direct shear configuration within a biaxial testing apparatus with servo-hydraulic control. Normal stress is applied horizontally to three steel forcing blocks that hold two sample layers (dimensions 5.4 cm × 5.7 cm area, ∼8 mm thickness), and we measure the shear stress required to drive the center block past the two stationary side blocks [Figure S1(c)]. The assembly is jacketed so that the effective normal stress during the experiment represents the combined effects of confining pressure (Pc), externally applied normal load, and pore pressure (Pp). During shear, Pc was held constant at 6 MPa; for each individual sample specimen the pore pressure at the inner sample boundary was held constant at 5 MPa, and a no-flow (undrained) condition was set at the outer boundary in order to monitor pore pressure in the layer throughout the experiment, following Ikari et al. [2009b]. Here we measured the shear strength at effective normal stresses of 5 (which approximates the in situ condition in most cases), 15, 25, and 35 MPa in order to extrapolate our results to deeper in the subduction zone system [Figure S1(d), Table S2]. For these experiments, three samples were trimmed from intact whole round cores, the rest were remolded wet (cold pressed into the sample assembly) (Table S1). 3.5 wt% NaCl solution was also used as pore fluid for these experiments [see Ikari et al., 2009b; Samuelson et al., 2009 for further details].

[12] During shear deformation of geologic material, shear strength is commonly described by the Coulomb-Mohr failure criterion:

display math(1)

where τ is the measured shear stress, µ is the coefficient of internal friction, σn' is the effective normal stress, and c is the cohesion [Handin, 1969; Byerlee, 1978]. In cases where enough deformation has accumulated to form a well-defined slip zone, c is often assumed to be negligible and the friction coefficient becomes µs, the coefficient of sliding friction. However, it has been demonstrated that in materials containing a significant proportion of clay minerals, c exists throughout the shearing process and is also dependent on the effective normal stress [Ikari and Kopf, 2011]. Cementation from advanced diagenesis would also cause elevated cohesion. Therefore, we use an apparent coefficient of friction µa, calculated from the measured shear stress:

display math(2)

which includes both frictional and cohesive strengths and does not distinguish between them. We report both maximum (peak) and residual (steady-state) values of τ (and µa), however, some of the deeper samples did not exhibit a clear peak so residual values were taken as maxima in these cases. Using apparent friction defined in this manner also facilitates comparison with previous studies in which cohesion is assumed to be negligible. In the cases where we measured cohesion directly, we normalize it by the maximum effective vertical stress and report it in terms of the parameter χ:

display math(3)

3.2. Results

3.2.1. Maximum Strength

[13] In Figures 2 and 3, we show maximum values of shear strength and apparent friction coefficient measured for intact samples at in situ conditions for Sites C0011 and C0012. Shear strength generally increases as a function of depth for both reference sites, however this increase is nonlinear which results in a decrease in µa with depth. At C0011, τ increases from ∼30 kPa to ∼1.5 MPa, while µa declines from ∼0.6 at the top of the section to nearly 0.2 at the bottom. Minimum µa of ∼0.2 is observed for samples from Units II, VI and V; no intact samples from Unit III at Site C0011 were tested. At C0012, the shallowest sample tested (35.7 mbsf) has a τ = 13 kPa and µa = 0.85, and for the deepest intact sample tested (405.8 mbsf) τ = 1.03 MPa and µa = 0.36. High values of µa at both sites generally coincide with the volcanic ash-bearing Units I and II. Very high strength was observed for a sample of volcaniclastic sandstone from the MSB (161.0 mbsf, Unit II), with µa = 1.06.

Figure 3.

Lithostratigraphy (left column), measurements of maximum and residual shear strength (middle column) and measurements of maximum and residual apparent friction (right column) for intact samples at in situ effective stress conditions (right column) for Site C0012. The lithostratigraphy is modified from Henry et al. [2012b].

Figure 4.

Residual µa as a function of effective normal stress for samples from Unit II and lower. Values for the décollement zone on the Muroto transect at Site 1174 [Ikari and Saffer, 2011] and on the Ashizuri transect at Site 1177 [Kopf and Brown, 2003] shown for reference.

3.2.2. Residual Strength

[14] In experiments performed at in situ effective normal stresses, we observe a pronounced peak in shear strength followed by a decay toward a lower residual value. While indentification of the peak strength is straightforward, in the single-direct apparatus the subsequent decay occurs over long displacements and in some cases did not reach a steady value by the end of the experiment (up to ∼10 mm displacement). Therefore, the residual µa values from in situ stress experiments reported here should be considered to be upper estimates. These values range from 0.19 to 0.57 at Site C0011 and 0.19–0.61 at Site C0012, with the exception of the volcaniclastic sandstone from the MSB (Unit II) at C0012 which exhibited residual µa = 0.72. At both sites, residual µa generally decreases as a function of depth (Figures 2 and 3).

[15] We were able to obtain values of steady-state shear strength from biaxial experiments at elevated effective normal stresses for samples from Units II through VI (Figure 4). At σn' = 5 MPa, residual µa ranges from 0.11 to 0.32, with the lowest value of µa exhibited by a specimen from Unit III. At higher effective normal stresses, µa does not exceed ∼0.2 at both sites and the lowest strength is consistently observed in the sample from Unit V (856.2 mbsf). Once again, distinctly higher friction values (µa = ∼0.5–0.55) are observed for a sample with sand-rich composition, in this case a turbidite from Unit IV.

Figure 5.

Seismic reflection data from Figure 1 showing the toe area and outer wedge, showing differing values of the surface slope angle α, and décollement dip β of the two major reflectors (frontal thrust and LSB). Also note thrust fault in the oceanic basement displacing the regional stratigraphy.

3.2.3. Cohesive Strength

[16] The cohesive strength, represented by the cohesion coefficient χ, ranges from 0.02 to 0.06 at Site C0011 and are slightly higher at Site C0012, where values range from 0.02 to 0.08. The volcaniclastic sand at Site C0012 has a very high-cohesion coefficient of 0.24. Postshearing cohesion measurements were mostly conducted for Site C0012 samples and χ ranges from 0.01 to 0.09, nearly identical to the preshear values. The whole-round cores were considered appropriate for cohesion measurements based on their “intactness” from shipboard X-ray Computed Tomography (XCT) scans, and by careful observation and handling during the sample preparation. Some cores recovered during Expedition 322 experienced significant drilling-induced damage, which in our case would lead to apparently low values of cohesion. However, the very high cohesion of the volcaniclastic sand suggests that drilling-induced damage is not a concern for these samples. Furthermore, we note that susceptibility to drilling damage is also a function of low-cohesive strength, so minor damage in samples with low in situ cohesion would have a minimal effect on our data and interpretations.

4. Critically Tapered Wedge Model

4.1. Model Description

[17] In order to constrain the position and the properties of the plate boundary décollement, we incorporate our friction measurements into a critical taper (Coulomb wedge) model for accretionary prisms [Davis et al., 1983; Dahlen, 1990]. In this model, the frictional strengths of the wedge and the basal thrust (décollement) are related to the taper angle, i.e., the sum of the surface slope α and the décollement dip β. The wedge itself is assumed to be at critical state, meaning that it is on the verge of experiencing shear failure at any location. Although there is some evidence that this may not be the case in Nankai [e.g., Conin et al., 2012], we consider the critical state case as a first approximation. Fluctuations in stress state can bring the wedge closer to or farther from critical state, so we assume here that the wedge, in its present-day geometry, likely experienced critical state at some point in its history [Wang and Hu, 2006]. A simplifying assumption employed here is that all sediments are “noncohesive”, which is likely appropriate in this case because most measured cohesion values are low. The taper angle then depends on the (assumed uniform) pore pressure ratio and the coefficient of sliding friction of the wedge and décollement, represented by the parameters ψo and ψb:

display math(4)

where:

display math(5)
display math(6)

[18] Since the frictional strengths are modulated by pore fluid pressure via its effect on effective (normal) stress, the quantities α' and ϕb' are the surface slope angle and inverse tangent of the friction coefficient in the décollement which both account for the fluid pressure ratio λ:

display math(7)
display math(8)

where λ is the ratio of total pore pressure to total overburden stress. The model is further simplified by assuming that λ in the wedge is equal to that in the décollement (λb), thus making ϕb' = tan−1µb. The fluid density ρf and bulk sediment density ρ are assumed to be 1.0 and 2.2 g/cm3, respectively. The value for bulk density assumes a porosity of ∼30%, which is an average value for the first ∼2 km depth in accretionary prisms [Bray and Karig, 1985].

4.2. Model Results and Interpretation

4.2.1. Toe Region

[19] From seismic reflection data, the observed surface slope angle α in the toe region of the Kumano transect is 10° [Moore et al., 2009] (Figure 5), which is significantly steeper than similar areas along the Ashizuri and Muroto transects [Kimura et al., 2007; Moore et al., 2009]. The toe area is considered to be well drained, based on lack of observed focused fluid flow and the presence of the trench wedge turbidites [Screaton et al., 2009], therefore we use λ = 0.5 representing approximately hydrostatic conditions. However, numerical modeling results integrating fluid flow and fault slip suggest that in the underlying Shikoku Basin sediments beneath the frontal thrust, overpressures of 45–77% of the overburden may exist [Rowe et al., 2012]. Therefore, we also test the case where λ = 0.65, as a lower estimate for overpressured conditions. Based on drilling results suggesting that the material accreted at the toe consists of the USB and the trench wedge facies [Screaton et al., 2009], and our measurements that show consistently high-friction values in Units I and II, we employ an average value of µa = 0.58 for the wedge. Because the toe is the location of incipient deformation, we use maximum (peak) values of apparent frictional strength for calculations in this particular region.

Figure 6.

Calculated basal friction (µb) values using critical taper theory. For the toe area, wedge friction is 0.58 and pore pressure ratios (λ) of (a) 0.5 and (b) 0.65 are tested. For the outer wedge, λ = 0.7 and wedge strengths (µw) of (c) 0.58 and (d) 0.40 are tested. FTR = Frontal Thrust Reflector, LSBR = Lower Shikoku Basin Reflector.

[20] In seismic reflection data, there are two candidates for the plate boundary fault zone: (1) the main frontal thrust which dips 7.6°, and (2) a stratigraphically lower reflector that is consistent with the LSB facies and dips 2.5° (Figure 5). The calculated décollement dip angle β matches the dip of the frontal thrust reflector for a given α of 10° when we employ décollement friction values of µa = 0.58–0.55 together with λ = 0.5–0.65, respectively (Figures 6a and 6b). These values are consistent with a plate boundary being located in frictionally stronger material, such as the USB facies or the overlying sandy trench wedge facies. In order for the dip of the décollement to be 2.5°, matching the LSB reflector, µa is predicted to be 0.50 to 0.56 which is inconsistent with our measured value of 0.46. We note that if the décollement friction is 0.46, the dip must be horizontal or nearly horizontal (Figures 6a and 6b). Our results thus suggest that the plate boundary décollement in the toe region is located on the main frontal thrust, which is hosted in frictionally strong material (Figure 7). This is in contrast with the toe regions on the Ashizuri and Muroto transects where the formation of the décollement occurs within the LSB, probably facilitated by very high-clay mineral contents and low-frictional strength [Brown et al., 2003] or excess pore pressures that reduce shear strength at depth [LePichon et al., 1993].

Figure 7.

Schematic illustration of the prism toe area, with friction coefficients assigned to major lithologic units based on experimental results at in situ conditions. The location of the plate boundary fault is expected to coincide with the main frontal thrust, which is the top of the USB at the toe and within the LSB further landward.

4.2.2. Outer Wedge

[21] Landward of the toe region, the surface slope decreases to 3.3° resulting in a narrower taper angle. Furthermore, only one candidate reflector remains, which is likely located within the LSB based on its dip angle of 2.5° and position relative to basement (Figure 5). Based on our observation that residual shear strength generally increases with depth, significant excess pore pressures would be required in this unit to make it weak enough to be mechanically favorable for hosting a major thrust fault [Le Pichon et al., 1993]. Recent estimates of in situ pore pressure at the base of the outer wedge determined by relationships between P wave velocity, porosity, and effective mean stress document excess pore pressures of 45–91%, or λ = ∼0.7–0.95 [Kitajima and Saffer, 2012]. We employ these values of λ, α, and β in a Coulomb wedge calculation for the outer wedge area. We assume the wedge frictional strength ranges between ∼0.4 based on friction measurements near the megasplay fault zone [Ikari and Saffer, 2011], and 0.58 based on our measurements of USB samples. The resulting friction of the décollement ranges from 0.23 to 0.33, generally consistent with our residual µa values of 0.11–0.31 (Figures 6c and 6d). We note that a weaker wedge is more consistent with both our measurements and previously published results [e.g., Kopf and Brown, 2003; Ikari and Saffer, 2011]. This is especially true if pore pressures are elevated such that λ > 0.7 as suggested by Kitajima and Saffer [2012], because in this case the predicted décollement friction would be above our range of measurements for a wedge strength of 0.58. This suggests that both the USB and LSB facies may be represented within the accreted wedge sediments. Thus, available data from this and previous studies are all consistent with the décollement having migrated from the USB and trench wedge in the toe region, into the LSB further landward, coincident with the presence of elevated fluid pressure. The décollement is also hosted by the LSB in the Muroto and Ashizuri transects, [Moore et al., 2001a, 2001b; Underwood, 2007], suggesting that this formation controls the location of the plate boundary on a regional scale.

4.2.3. Transition Zone

[22] Kimura et al. [2007] define the transition zone as a region of high-surface slope marking the transition between outer and inner wedges, which also coincides with the updip limit of the seismogenic zone (Figure 1). In the Nankai area, seismic reflection data indicate that the décollement steps to a lower stratigraphic position at ∼25–35 km landward from the trench, within the transition zone on all three transects [Moore et al., 2001a, 2001b; Park et al., 2002; Bangs et al., 2004; Kimura et al., 2007]. Such downstepping is facilitated by weakening in progressively deeper strata, however it is important to note that this requires a net decrease in shear strength, not just the friction coefficient. Although we observe that the coefficient of friction decreases with depth at the input sites, the shear strength continues to increase downsection, so frictional strength variations alone do not appear to be capable of facilitating any downstepping. For the Muroto transect, it was suggested that downstepping may be caused by gradual drainage of accumulated pore fluid overpressure in the underthrust sequence, thus raising the effective and shear stresses within this specific package causing the décollement to migrate to more energetically favorable strata [Skarbek and Saffer, 2009; Tobin and Saffer, 2009]. Comparison of friction measurements for samples from Units IIB through VI from Sites C0011 and C0012 shows that while lowest µa occurs in Unit III at approximately in situ conditions, at elevated effective stresses of 15–35 MPa the frictionally weakest sample is from Unit V (Figure 4). Therefore, this unit requires the lowest relative amount of overpressure (lowest λ) in order to be the weakest sediment package. This suggests that the décollement is likely to step down into the volcaniclastic unit, bypassing the turbidite unit containing very strong, sand-rich sediment, but less likely to step directly into the pelagic clay layer (Unit VI). However, our results show that the pelagic clay is also frictionally weak in general (consistently the second-weakest sample) and therefore is the next likely candidate to host the décollement if fluid pressure becomes further elevated relative to Unit V. Although sand-rich lithologies in Unit V exhibit high permeability and could act as fluid conduits [Hüpers and Kopf, 2012], the transition zone is located ∼25–35 km from the trench so that the path length of fluid escape is large, a condition favorable for maintaining overpressure. Additionally, shearing of sediment can significantly reduce fault-perpendicular permeability [e.g., Brown and Moore, 1993; Faulkner and Rutter, 2000; Kopf, 2001; Ikari and Saffer, 2012], so it is possible that permeability reduction driven by shearing in Unit V could limit vertical flow in Unit VI and build overpressure. Since the strengths of the volcaniclastic and pelagic samples are very similar, only a small amount of pore pressure increase may be necessary for further décollement step-down. If we assume a moderately elevated fluid pressure [Skarbek and Saffer, 2009; Tobin and Saffer, 2009; Kitajima and Saffer, 2012] resulting in a depth-dependent effective stress gradient of ∼6–7 MPa/km, our experimental results are relevant for depths up to ∼5–6 km. This coincides with the inferred depth of the seismogenic zone updip limit in Nankai [Satake, 1993; Moore and Saffer, 2001], therefore we suggest that Units V and/or VI may host the plate boundary fault at seismogenic depths.

5. Discussion

5.1. Lithologic/Mineralogic Controls on Observed Strength Patterns

[23] We observe that our values of µa broadly correlate with bulk mineralogy; low values of residual µa are attributable to the clay-rich lithology of the majority of these samples. This is consistent with a wealth of experimental data, which shows that both clay-rich standards [e.g., Lupini et al., 1981; Logan and Rauenzahn, 1987; Morrow et al., 1992; Brown et al., 2003; Ikari et al., 2007, 2009a; Behnsen and Faulkner, 2012] and natural fault samples [e.g., Smith and Faulkner, 2010; Carpenter et al., 2011] are frictionally weak. This includes previous work documenting residual friction values of 0.32–0.46 for samples from the vicinity of the megasplay fault at Site C0004 and the frontal thrust zone at C0007 [Ikari et al., 2009a; Ikari and Saffer, 2011], and measurements on samples from the Muroto transect [Brown et al., 2003; Kopf and Brown, 2003; Bourlange et al., 2004; Ikari and Saffer, 2011; Saffer et al., 2012]. High values of residual µa (> 0.4) are observed for two clay-poor sand packages: volcaniclastic sand from 161 mbsf at Site C0012, and turbidite sand from 781 mbsf at Site C0011.

[24] There are two likely explanations for maximum µa values that are significantly higher than residual values, for example the maximum µa of 0.46 measured for the LSB. The first is overconsolidation of the samples, which has been documented in silty claystones at Sites C0011 and C0012 [Hüpers and Kopf, 2012]. The second is the presence of large amounts of smectite throughout the sediment column approaching the trench along the Kumano transect, which has not reacted to form illite due to lower temperatures compared to the Muroto transect. Previous work has shown that smectite-rich gouges, even when normally consolidated, can sometimes exhibit a peak in strength that is significantly higher than residual values [Saffer et al., 2001; Saffer and Marone, 2003; Ikari et al., 2009b].

[25] However, variations in lithology and mineralogy are not completely consistent with our experimental results, and therefore require other explanations. As noted by Underwood [2007], in the Muroto transect the location of the décollement does not coincide with abnormally high-bulk clay or smectite contents, therefore clay mineralogy alone cannot always explain patterns in mechanical behavior. On the Kumano transect, there are no obvious packages of anomalously high-bulk clay or smectite that would represent abnormally weak layers. Rather, there are small intervals that are abnormally strong due to sand-rich lithologies. In general, we observe that maximum and residual µa decrease as a function of depth/effective stress, but both the bulk clay content and bulk smectite content at Sites C0011 and C0012 are relatively constant downsection [Underwood and Guo, 2012]. One explanation could be the presence of significant amounts of silica in the form of volcanic ash within the upper portion of the sediment column, which is difficult to identify in XRD analysis. In some studies where the friction of clay-rich sediment is measured over a range of effective normal stress, it is observed that the coefficient of friction tends to decrease with increasing pressure, consistent with our results [e.g., Saffer et al., 2001; Saffer and Marone, 2003; Ikari et al., 2007; Behnsen and Faulkner, 2012].

[26] We also note that the reason for the volcaniclastic sample from Unit V being consistently the weakest sample at effective normal stresses greater than in situ is difficult to explain by mineralogy. Strangely, XRD analysis shows that this sample is composed of only ∼35% phyllosilicates, while the amounts of zeolite (15%) and iron oxides (e.g., hematite, 7%) could be significant. Studies investigating the frictional strength of zeolites are rare; however, the friction of the zeolite species laumontite and clinoptilolite was measured to be 0.66 to 0.80 under water-saturated conditions and thus would not be a source of weakness [Morrow and Byerlee, 1991; Morrow et al., 2000]. We are currently unaware of studies measuring the frictional behavior of iron oxides. The cause of the unusual weakness in the sample from the volcaniclastic unit merits further investigation.

5.2. Implications of Basement Thrusting for Fault Behavior

[27] Modification of wedge structure and position of the plate boundary fault zone suggests that the subducting oceanic basement thrust (see Figure 5) essentially functions similarly to a subducting seamount. Where such features interact with plate boundary fault zones, they are commonly aseismic and act as barriers to earthquake rupture [Wang and Bilek, 2011]. The basement thrust coincides with the convergence of the frontal thrust and the LSB reflector and thus appears to be disrupting the local lithostratigraphy (Figure 5), suggesting that this feature causes the plate boundary to be located on the frictionally stronger frontal thrust. Because it has been suggested that rupture propagation in shallow accretionary prisms is facilitated by weak, clay-rich sediment [Faulkner et al., 2011] or high-overall pore pressures [Bangs et al., 2009], the high-friction, well-drained deposits on the frontal thrust make the toe of the Kumano transect less conducive for earthquake propagation than adjacent areas (e.g., Muroto). This is supported by earlier experiments documenting that friction coefficients on the frontal thrust are ∼0.1 higher than on the Muroto décollement [Ikari and Saffer, 2011].

[28] If the geometry in the Kumano transect toe is driven by the subducting oceanic basement thrust, it follows that once the feature subducts sufficiently landward that the taper angle would narrow and the plate boundary would return to the LSB. This is supported by the development of the protothrust zone below the trench wedge which suggests that more typical imbrication patterns may be resuming, as well as by recent slumping and mass transport which may suggest that the taper is becoming supercritical [Moore et al., 2009; Screaton et al., 2009]. In this case, the LSB would host the décollement across the entire Nankai margin, functioning as a large, homogeneous slip plane that could favor earthquake rupture as a single asperity and/or large amounts of slip that reach the seafloor. The magnitude of the effect of local strengthening of the toe region on earthquake rupture and slip propagation should be the focus of future investigations.

6. Conclusions

[29] Shear strength measurements on sediments approaching the Nankai Trough along the NanTroSEIZE Kumano transect show that while the apparent friction coefficient of the upper sediments (Units I and II) is elevated compared to deeper sediments (Units III and below), shear strength generally increases with depth. A volcaniclastic sand layer represents a significant departure from this trend and is observed to be among the strongest specimens tested. Results of experiments conducted at in situ effective normal stress conditions combined with critical taper analysis suggests that the plate boundary fault is located at the base of the frontal thrust near the trench, however this feature may be a local anomaly and that the LSB controls the megathrust margin-wide. As subduction progresses landward, the plate boundary would be expected to be hosted in progressively lower units, migrating to the LSB in the toe region and likely reaching Units V and VI at seismogenic depths based on observations that the friction coefficient decreases in deeper stratigraphic units. However, since the reduction in friction coefficient is not sufficient to cause a depth-dependent decrease in shear strength, downstepping of the décollement probably requires the presence of moderately elevated pore pressures

Acknowledgments

[30] We thank Editor Thorsten Becker, an anonymous Associate Editor and three anonymous reviewers for helpful comments that greatly improved the clarity and content of this article. We also thank Chris Marone for use of the biaxial shear apparatus, and Demian Saffer and Rob Skarbek for helpful discussions. This work was supported by the U.S. Science Support Program via a Post Expedition Activity (PEA) award for IODP Expedition 322 to M. Ikari, and by the Deutsche Forschungsgemeinschaft (DFG) via MARUM Center of Excellence and grant HU 1789/2-1 to A. Hüpers.

Ancillary