Significant along-strike changes in the protothrust zone at the toe of the Nankai Trough accretionary prism were imaged in new high-resolution seismic reflection data. The width of the protothrust zone varies greatly along strike; two spatially discrete segments have a wide protothrust zone (∼3.3–7.8 km, ∼50–110 protothrusts), and two segments have almost no protothrust zone (∼0.5–2.8 km, <20 protothrusts). The widest protothrust zone occurs in the region with the widest and thickest sediment wedge and subducting turbidite package, both of which are influenced by basement topography. The trench wedge size and lithology, the lithology of the subducting section, and the basement topography all influence the rate of consolidation in the trench wedge, which we hypothesize is an important control over the presence and width of the protothrust zone. We conclude that protothrusts are fractures that form from shear surfaces in deformation band clusters as the trench fill sediment is consolidated. Strain localization occurs at sites with a high density of protothrusts, which become the probable locations of future frontal thrust propagation. The frontal thrust may propagate forward with a lower buildup of strain where it is adjacent to a wide protothrust zone than at areas with a narrow or no protothrust zone. This is reflected in the accretionary prism geometry, where wide protothrust zones occur adjacent to fault-propagation folds with shallow prism toe surface slopes.

Convergent margins host a wide range of deformation processes that vary greatly both between different margins as well as along individual margins. In some cases, a protothrust zone extends seaward from the frontal thrust to the deformation front, which is the furthest seaward expression of faulting. The protothrust zone is an area with parallel shear structures that have relatively small offsets (e.g., Nankai Trough—Karig, 1986; Cascadia—Wang et al., 1994; Hikurangi—Barnes et al., 2018). However, protothrust zones only occur in discrete locations, and it is unclear how they relate to other subduction zone characteristics. The Nankai Trough subduction zone in SW Japan (Fig. 1) is one of the few places where protothrusts are well developed. It has also been the site of many geological and geophysical surveys in the last few decades, making it an ideal laboratory for studying the relationship between protothrusts and other subduction margin characteristics (e.g., Taira et al., 1991; Moran et al., 1993; Bangs et al., 2004; Yamashita et al., 2018). Drill cores at Ocean Drilling Program (ODP) Site 1174 (landward of the deformation front) revealed deformation bands in the Nankai protothrust zone that were interpreted to correlate with the protothrusts in seismic images (Karig and Lundberg, 1990; Ujiie et al., 2004). However, the relationship between core-scale deformation bands and seismic-scale features in the protothrust zone is poorly understood. As a result, the nature of protothrusts and the physical conditions that promote their formation have not been well established.

Karig and Lundberg (1990) suggested that protothrusts are narrow panels of slightly steeper-dipping beds that host deformation bands and that deformation bands initially form at shallower angles in subhorizontal beds but are rotated to steeper dips within these panels. Deformation bands can include compaction bands, shear bands, and dilation bands or hybrids of these bands (Karig and Lundberg, 1990; Bésuelle, 2001). The specific deformation mechanism depends on the mineralogy, grain size and shape, sorting, cementation, porosity, and stress state of the sediment (Fossen et al., 2007). However, deformation bands form preferentially in coarser, high-porosity sediment (Fossen et al., 2018), which is commonly found at accretionary margin trench wedges.

The evidence for the steeper-dipping panels comes from cores that show localized bedding dips of 10°–15°, which do not agree with the shallow dips in seismic sections. These dips led Morgan et al. (1994) to postulate that the protothrusts could represent narrow angular folds caused by local buckling.

Lateral variations in the incoming crust and the overlying sediment type and thickness are thought to control along-strike variations in the Nankai Trough accretionary prism. However, the relationship between the incoming material and the deformation processes is still poorly understood. This study addressed the role of subducting topography and sediment properties in producing variations in the accretionary prism along the Nankai Trough subduction zone by examining the protothrust zone. We investigated how the protothrust zone is affected by changes in the incoming plate and sediment, as well as how protothrusts develop from deformation bands into the frontal thrust. We also considered how protothrust zones are related to the style of frontal thrust propagation. We used high-resolution seismic reflection data that cover a wide section of the margin to infer that variations in the protothrust zone are caused by variations in the incoming sediment, which in turn influence strain localization of the sediment and thus its behavior during accretion.

Tectonic History

The heterogeneous basement structure of the subducting Shikoku Basin (Kobayashi et al., 1995) and its highly variable sediment cover (Ike et al., 2008) strongly influence deformation at the Nankai Trough. The Shikoku Basin lithosphere was formed during back-arc rifting and seafloor spreading in the eastern part of the Philippine Sea plate between 30 and 15 Ma (Watts and Weissel, 1975; Kobayashi and Nakada, 1978; Taylor and Fujioka, 1992; Okino et al., 1994; Sdrolias et al., 2004). The extinct Shikoku Basin spreading center lies just to the east of our study area, so the age of the basin increases to the SW. The Kinan seamount chain (Fig. 1), which trends roughly NNW-SSE down the center of the basin, formed near the extinct spreading axis. Although organized spreading ceased ca. 15 Ma, seamount eruptions continued to as recently as 8–7 Ma (Ishii, 2000; Sato et al., 2002; Ishizuka et al., 2009). The superposition of seamount topography on the subducting Shikoku Basin topography has had profound effects on the structure of the Nankai accretionary prism.

An important event in the history of the Shikoku Basin that greatly affected the sedimentation patterns in the basin was the apparent cessation of subduction at the Nankai Trough at ca. 12 Ma, followed by resurgence of subduction at ca. 6 Ma (Kimura et al., 2014, 2018; Wu et al., 2016). During this gap in subduction, sediment delivered from the Japanese Islands flooded the trench and poured far out onto the floor of Shikoku Basin (e.g., Pickering et al., 2013). Subsequent reinitiation of subduction after ca. 6 Ma led to the formation of the bathymetric low that defines the present Nankai Trough. Collision of the Izu-Bonin arc with Honshu and convergence between SW Japan and NE Japan at ca. 2 Ma caused uplift in eastern Honshu that supplied sediments to the trough through submarine canyon systems (Kimura et al., 2018).

Sedimentation History

The large variations in basement relief of the subducting Philippine Sea plate have resulted in variations in overlying sediment thickness and type. Few of the regional and local variations in basement topography are reflected in the seafloor bathymetry because the basement topography is almost completely buried beneath the sediment. Our seismic-stratigraphic interpretations were based on core-log-seismic correlations at several Deep Sea Drilling Project (DSDP), ODP, and Integrated Ocean Drilling Program (IODP) sites (Fig. 2). We followed the facies designations defined by Underwood and Pickering (2018), who used core data from the IODP Kumano transect (Underwood et al., 2010) to redefine the earlier facies terminology, which was based on results from previous ODP drilling on Legs 131 and 190 (e.g., Ike et al., 2008).

The oldest stratigraphic unit is a volcaniclastic facies that overlies the oceanic basement. It is much thicker at the western ODP Site 1177 along the Ashizuri (western) transect than in the Muroto (eastern) transect (Fig. 2; ODP Sites 808, 1173, and 1174; Moore et al., 2001; Steurer and Underwood, 2003). This facies also has a higher proportion of siliciclastic silt at Site 1177 than at Site 1173. Overlying the basal volcaniclastic unit at Site 1177, there is a package of siliciclastic turbidites with abundant organic matter. These are interpreted to be Lower to Middle Miocene Kyushu Fan deposits (LSB-T of Ike et al., 2008), which were supplied by a mainland source area, possibly between Kyushu and Shikoku, and transported along a westerly supply route (Pickering et al., 2013). These deposits are not present at Site 1173 (Underwood, 2007) due to the deflection of turbidity current flow paths around the Kinan seamount chain.

The Shikoku Basin facies consists of hemipelagic and pyroclastic/hemipelagic mudstones with ages of 12.5–9.1 Ma and 7.8 Ma–recent, respectively (Underwood and Pickering, 2018; Upper Shikoku Basin [USB] and Lower Shikoku Basin [LSB] of Ike et al., 2008). The USB/LSB boundary is diagenetic, and we refer to it as the “diagenetic boundary.” The pyroclastic/hemipelagic–hemipelagic boundary is lithological and was identified in drill cores as the onset of ash beds. The mudstones in the western Shikoku Basin contain a higher percentage of smectite than those in the central Shikoku Basin (Steurer and Underwood, 2003). Additionally, the cessation of near-trench volcanism ca. 11 Ma (Kimura et al., 2005) resulted in a gradual change from smectite- to illite-dominated clay mineral assemblages (Underwood and Fergusson, 2005; Guo and Underwood, 2011). This temporal change in mineralogy supports a mainland terrigenous source, even though the Izu-Bonin arc was closer to the eastern part of the basin during the Miocene (Underwood et al., 2010). Significant lateral transport of volcanic lithic sediment along the trough axis from the Izu-Honshu collision zone during the Quaternary led to the development of the axial trench sediment wedge on top of the older Shikoku Basin deposits (e.g., Underwood and Fergusson, 2005; Underwood and Pickering, 2018).

Seventeen two-dimensional (2-D) high-resolution seismic reflection lines were collected across the Nankai Trough trench axis (Fig. 3) on the R/V Kaiyo of the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) in 2013 (KY13–11) and 2014 (KY14–07). For high-resolution shallow seismic imaging, a 380 in.3 (6.23 L) air-gun array was used with a 1.2-km-long hydrophone streamer that had 192 channels at a 6.25 m interval. Standard seismic processing techniques were applied, including noisy trace editing, 20–200 Hz band-pass filtering, normal moveout (NMO) correction, 16-fold common midpoint (CMP) binning with a 3.125 m interval, stack, and Kirchhoff time migration. We then performed time-to-depth conversion using velocity functions based on a three-dimensional (3-D) P-wave velocity model derived from the JAMSTEC regional ocean-bottom seismometer (OBS) surveys (Nakanishi et al., 2018). Observations of along-strike variations in sediment properties and oceanic basement relief were supplemented by several longer-offset 2-D seismic surveys (e.g., Moore et al., 1990, 2001; Ike et al., 2008; Park et al., 2014) that were oriented both parallel and perpendicular to the protothrust zone, as well as the Muroto transect 3-D survey (Fig. 3; Bangs et al., 2004).

Seismic Stratigraphy

The deepest mapped unit is the volcaniclastic facies, which overlies the oceanic crust. The Kyushu Fan siliciclastic turbidite sequence, the next oldest mapped unit (Fig. 2), is characterized by continuous, parallel-bedded, high-amplitude reflections. The sequence is not ubiquitous along strike but instead occurs mostly in basement lows (e.g., Fig. 4), and it onlaps the basement topography, most notably the basement scarp in lines 121–123 (Fig. 5). The siliciclastic turbidite package is continuous along lines 107–117, where the basement is smooth with small variations in topography (<150 m; Figs. 4 and 5). There is a thin siliciclastic turbidite package along lines 121–125, although it has fewer distinct layers than the siliciclastic turbidite package along lines 107–117. Along line 127, there is a siliciclastic turbidite package seaward of the frontal thrust, which onlaps a basement high at its landward boundary. Lines 119 and 120 are located above a basement high, and the siliciclastic turbidite sequence is absent beneath the trench wedge. Along line 120, the package is only present seaward of the trench axis, and along line 119, there is a possible siliciclastic turbidite package beneath the accretionary prism. The siliciclastic turbidite unit is also absent in lines 129–134 (Fig. 4), which have a smooth basement with gradual changes in basement depth.

The overlying hemipelagic mudstone and pyroclastic/hemipelagic Shikoku Basin units both occur throughout the entire study area. The boundary between the hemipelagic and the pyroclastic/hemipelagic units has been debated following drilling in the Kumano transect (Underwood and Pickering, 2018), and the more recent definitions place the top unit boundary at the first occurrence of ash in the cores, which is difficult to interpret seismically. Therefore, we considered the hemipelagic and pyroclastic/hemipelagic sediment as one unit (pyroclastic/hemipelagic–hemipelagic) that has a downward decrease in ash layers.

The pyroclastic/hemipelagic–hemipelagic Shikoku Basin sediments are buried beneath an upward-coarsening wedge of turbidites that forms the trench-fill unit. The trench-fill unit seaward of the frontal thrust makes up the trench wedge. The trench wedge varies along strike in thickness, width, and bounding characteristics (Fig. 6). The trench wedge width is the horizontal distance between the seafloor expression of the frontal thrust and the point of trench-fill onlap onto the pyroclastic/hemipelagic–hemipelagic sequence at the seafloor. The trench wedge thickness is measured between the seafloor and the base of the trench fill at the frontal thrust. The trench wedge is wide (∼10–20 km) and thick (550–890 m) in the west and center (lines 107–117 and 121–127; Table 1; Figs. 6, 7, and 8). In contrast, lines 119–120 and 129–134 have a narrow (∼8–14 km), thin (270–480 m) trench wedge (Figs. 9 and 10). Individual strongly reflective turbidite packages within the trench wedge become thinner and less reflective to the west (Fig. 4; Underwood and Pickering, 2018). We traced four marker horizons in the trench wedge throughout the area to use for structural analysis, which is discussed below.

Seismic Structure

We define the frontal thrust as the furthest seaward accretionary thrust fault that offsets the seafloor and has >30 m offset in all the sediment packages above the décollement. Incipient frontal thrusts in some lines have >30 m offset in parts of the sediment, but they have not fully propagated from the décollement to the seafloor (Fig. 10).

The protothrust zone is defined as the zone between the frontal thrust and the deformation front (Karig, 1986). Steep discontinuities with reverse-sense displacement within the protothrust zone that strike subparallel to the frontal thrust are interpreted as protothrusts. Polygonal normal faults in the Shikoku Basin sediment (Fig. 10; Heffernan et al., 2004) reduce the continuity of horizons, making it difficult to identify protothrusts in the Shikoku Basin facies. The polygonal faults in the incoming Shikoku Basin facies occur across the entire region and become slightly degraded by compaction and deformation of sediment beneath the trench wedge. The faults have a mean dip of 74°, and there are an equal number of landward- and seaward-dipping faults in our 2-D seismic lines. The complexity and regional extent of these faults are the subject of work in progress and are beyond the scope of this paper. However, the polygonal normal faults do not appear to have a spatial relationship with the protothrusts that would suggest they significantly influence the protothrust zone characteristics.

The protothrust zone width ranges from 0.5 to 7.8 km. Along-strike variations in the protothrust zone width exist between four distinct segments (I–IV; Fig. 5). A relatively wide protothrust zone exists between lines 107–117 (segment I) and 121–127 (segment III). Segments I and III have similar trench wedge widths but have different protothrust zone widths (Table 1; Fig. 6). In segment I, there is a moderate-sized, 3–5-km-wide protothrust zone, and there is a gradual transition from the protothrust zone into a largely undeformed trench wedge (Fig. 7). In segment III, which contains the widest section of the protothrust zone (3.5–8 km), the deformation front occurs furthest seaward, above a basement scarp (Fig. 8). Lines 119–120 (segment II) and 129–134 (segment IV) have narrow protothrust zones (0.5–3 km) with relatively few protothrusts (Figs. 9 and 10). The few (<20 per line) protothrusts in these lines occur preferentially in the deepest part of the trench sediment near the toe of the frontal thrust.

Following Barnes et al. (2018), we measured the protothrust dip, downdip extent, separation, and spacing (Fig. 11). We traced four trench wedge horizons (A–D) along strike to compare variations in deformation along strike and with depth (Figs. 710). A key horizon was chosen out of the interpreted horizons (A–D) for each line. The key horizon was the one that was intersected by most of the protothrusts, and it was used to measure the spacing of the protothrusts and the distance seaward of the frontal thrust. In lines 107–109, 115–127, and 134, it was horizon C, and in lines 111–113 and 129–133, it was horizon D.

Protothrust Dip

The dip of every interpreted protothrust was recorded at each of the four horizons to quantify variations in dip with depth (Fig. 6). The protothrusts strike northeasterly, subparallel to the deformation front, and have predominantly northwest dips. The protothrusts steepen upward and show distinct changes in dip at some shallow, strongly reflective packages (Fig. 8). These packages correspond to the thick, sandy turbidites identified in ODP cores (Moore et al., 2001). The mean dips of the individual protothrusts in segment I range from 37° to 64°; however, all of the steepest dips are recorded at horizon A. The mean dip is 37° to 53° if horizon A is excluded. The mean protothrust dips in segment III are 40° to 67° (40°–54° excluding horizon A), which is similar to protothrust dips in segment I and is consistent with high-angle, compactive shear bands associated with porosity reduction recorded in ODP drill cores (Karig and Lundberg, 1990; Ujiie et al., 2004). These dip angles are higher than predicted for Coulomb shear, which is ∼30° to the subhorizontal maximum compressional stress (Morgan et al., 2007). The dips of the protothrusts in segments II and IV have more variability than those in segments I and III (Fig. 6). Segment II has mean dips of 41°–81°, and segment IV has mean dips of 17°–64°. The variability of the dips in narrow protothrust zones may be the result of the fractures forming due to bending stresses associated with fault-bend fold propagation (Fig. 10D) rather than the same mechanism that produced the approximately parallel protothrusts in segments I and III.

Protothrusts also occur in the hanging walls of the frontal thrusts (Figs. 7 and 8). These protothrusts correlate with the deformation bands in drill cores at DSDP Site 583, which dip between 25° and 60° (Karig and Lundberg, 1990). The accreted protothrusts were measured at approximately the same stratigraphic horizon as horizon C. The mean dip of the accreted protothrusts is 54°, which is higher than the mean protothrust dip at horizon C within the protothrust zone (48°; Table 1).

Protothrust Spacing

The spacing between protothrusts was measured as the horizontal distance between two protothrusts along the key horizon. The mean protothrust spacing increases with the protothrust zone width but shows a different trend for wide (segments I and III) versus narrow (segment II and IV) protothrust zones (Fig. 12). This difference supports our interpretation that the mechanisms that produce protothrusts at wide and narrow protothrust zones are different.

The spacing within each protothrust zone predominantly decreases landward (Fig. 13), which is consistent with an increase in strain. Interestingly, the landward reduction in protothrust spacing is not linear; clusters of closely spaced protothrusts occur in the areas that are associated with seafloor deformation and lower-angle protothrusts (e.g., Figs. 8 and 13). These locations are interpreted as the locations of incipient frontal thrusts.

Protothrust Downdip Extent

The downdip extent was recorded as the distance between the upper and lower tips of the fault (Fig. 11). Protothrust downdip extents are likely to be minimum values because the resolution decreases from ∼3 m in the shallow sediment to ∼15 m in the deepest sediment, so the deepest part of the protothrusts may be below the resolution of the seismic image. Additionally, the protothrusts were more clearly imaged in sediment with strong reflectivity. Therefore, lithological variations may result in a variable ability to image protothrusts, especially within the Shikoku Basin sediment. Protothrusts were difficult to interpret in the Shikoku Basin facies in the eastern protothrust zone because it is acoustically weak. Furthermore, the presence of polygonal normal faults within the subducting sediment made it hard to discern overprinted protothrusts. The range of protothrust downdip extents is similar at both the western and eastern wide sections of the protothrust zone (segments I and III), but the mean is less at the protothrust zone in segment III as compared to segment I (Table 1). This is despite a similar maximum trench wedge thickness in both segments. In both segments, some protothrusts penetrate from the décollement to the seafloor, but most protothrusts are confined to the trench wedge (e.g., Fig. 7). The thickness, extent, and lithology of the Shikoku Basin facies vary locally (Ike et al., 2008), which might cause differences in the likelihood of protothrusts penetrating beneath the trench fill, or the degree to which protothrusts are imaged. The protothrusts that penetrate deeper tend to have a shallower dip at depth (Table 1), and changes in dip commonly occur at strongly reflective horizons, which may represent a change in lithology or sediment properties (e.g., Fig. 8).

Factors Controlling Protothrust Zone Size and Characteristics

Basement Topography

The protothrust zone width is clearly influenced by the subducting basement topography. The wide protothrust zone in segment I is above the smoothest portion of the oceanic crust, whereas the protothrust zone narrows significantly in segment II, where there is a subducting seamount (line 119–120). Additionally, the wide protothrust zone in segment III is above a smooth portion of the oceanic crust and terminates at the deformation front above the location of a basement scarp (Fig. 8). The relationship between the deformation front and basement scarps in segments III and IV is shown in Figure 14. In segment III, the basement topography influences the location of an area of high strain identified by greater protothrust displacement and associated seafloor uplift (Fig. 8), which is the probable location of a future frontal thrust. Previous studies show that steps in basement topography can effectively pin the deformation front to a particular location (Thomas, 1977; Bahroudi and Koyi, 2003; Vidal-Royo et al., 2009). Therefore, the basement scarp may explain why the protothrust zone is wider in segment III than in segment I despite a similar trench wedge width.

Basement topography also influences the thickness of sediment cover (Fig. 14), which is discussed in the following sections. Models show that the deformation front propagates further seaward of the frontal thrust where the sediment cover is thicker (e.g., Marshak et al., 1992). Therefore, where basement topography causes a sudden lateral change in the sediment thickness, the deformation front may bend along strike as a result of differential propagation of deformation (Farzipour-Saein et al., 2013). As a result, steps in the basement topography may explain the abrupt changes in deformation front location and thus the width of the protothrust zone in our survey area.

Subducting Sediment Section

The seismic lines with a relatively narrow protothrust zone at the Nankai Trough are the same lines in which the siliciclastic turbidite unit is absent from the subducting hemipelagic facies (Fig. 6). The only exception is line 127, which has a relatively wide protothrust zone but does not have a siliciclastic turbidite unit. However, diffraction noise above the oceanic basement (Fig. 5) may be the result of this line on the very edge of a basement feature and could mark the transition between areas with and without the siliciclastic turbidite unit. The protothrusts do not penetrate the siliciclastic turbidite unit, so if the siliciclastic turbidite sediment controls protothrust formation, its influence must be indirect, such as by modifying the local stress state. At the Nankai Trough, lithologic variability (the presence of siliciclastic turbidite) is the primary factor controlling the décollement pore pressure (Saffer and Bekins, 2002). High pore pressures are common because the low-permeability hemipelagic sediment cap prevents pore-fluid escape at a rate comparable to the loading during sedimentation and subduction (Moore et al., 2001; Gamage and Screaton, 2006). However, the siliciclastic turbidite facies is more permeable than the hemipelagic/pyroclastic–hemipelagic facies, and the siliciclastic turbidite allows fluid to be expelled as it is compressed, reducing the pore pressure along the décollement (Karig and Morgan, 1994; Underwood and Pickering, 2018). The inferred décollement pore-pressure variations are consistent with the differences in wedge taper measured along the Ashizuri and Muroto transects (Moore et al., 1990), which correspond to the western and eastern parts of our survey area, respectively. The differences in wedge taper are interpreted to reflect a high basal strength along the Ashizuri transect due to better drainage along the décollement (Saffer, 2010). In our data, we see that changes in the prism geometry are consistent with changes in the subducting lithology and basement topography (Fig. 6).

Simple analytical models have been used to explain the influence of décollement pore pressure on deformation in the trench wedge (Wang et al., 1994; Weiss et al., 2018). Wang et al. (1994) hypothesized that an incipient décollement with high pore pressure is required seaward of the frontal thrust for stress to transfer to the footwall and a protothrust zone to form. Their model also showed that the magnitude of pore pressure within the incipient décollement may control the width of the protothrust zone. However, this model predicts that the protothrust zone would be wider in the eastern part of our study area, where the décollement pore pressure is higher and the wedge taper is lower than in the west, which is not the case. Therefore, although the entire Nankai margin must have sufficiently high incipient décollement pore pressure to transmit stress a few kilometers seaward of the frontal thrust, the presence of the subducting siliciclastic turbidite package and its reduction in the décollement pore pressure cannot explain the differences in the maximum width of the wide protothrust zone in segments I and III because the relationship between the protothrust zone width and the presence of the siliciclastic turbidite unit is the opposite of what the Wang et al. (1994) model predicts (Fig. 6).

Trench Wedge

The Nankai protothrust zone is widest where there is a wide and thick trench wedge (Fig. 6; Table 1). This observation agrees with previous studies that show that deformation at accretionary margins can be influenced by the thickness and type of accreting sediment (e.g., Marques and Cobbold, 2002). The thickness of the trench wedge sediment may influence the consolidation rate and thus the mechanical behavior of the trench wedge sediment. For example, turbidite deposition within the trench wedge can drive loading and consolidation of sediment within and beneath the wedge (Hüpers et al., 2018). As a result, a wide, thick trench wedge is likely to have more consolidated trench-fill sediment seaward of the frontal thrust than a narrow, thin trench wedge.

The size of the Nankai Trough trench wedge is controlled by the terrigenous sedimentation rate, which in turn is controlled by the magnitude, frequency, and pathways of turbidity current flows (Mountney and Westbrook, 1996). Along-strike variations in the sedimentation rate are largely due to the subducting fossil Shikoku Basin spreading center axis, which forms a bathymetric high (Underwood and Pickering, 2018). The easternmost lines (segment IV) are adjacent to the higher topography of the extinct spreading center, which limits the amount of sediment that reaches the site from the axial turbidity currents that travel from the northeast (Pickering et al., 1993). As a result, a narrower trench wedge of 7–13 km has existed over the last 0.5 m.y. (Mountney and Westbrook, 1996). In contrast, near the westernmost lines (segment I), the trench wedge has been much wider (13–21 km) over the last 1 m.y. Segment I is located above a closed basement low, which has been a site with high sedimentation rate due to sediments derived both along the axis from the Izu-Bonin collision zone, as well as from the proximal Ashizuri Canyon (Pickering et al., 1992). These sediments extend as far as segment III. However, segment II is located along a bathymetric high, which acted as a barrier to the turbidity currents. The turbidity currents were diverted around the highs in segment II before being deposited in the lows of segment I.

Although the segments with a wide trench wedge correlate with those that have wide protothrust zones, the width of the trench wedge does not correlate directly with the maximum width of wide protothrust zones (Fig. 6). For example, segments I and III have similar trench wedge widths but different protothrust zone widths. Lithological variations in the trench wedge are also likely to influence the protothrust characteristics. At the Nankai Trough, the protothrusts steepen upward at the base of the strongly reflective sandy turbidites, highlighting a lithological control over protothrust characteristics. Strongly reflective turbidite packages occur throughout the entire study area, but they become thinner and less reflective to the west. These packages correlate with sand and silt turbidites that are more abundant and thicker in the trench wedge at the Muroto transect drill sites (segment IV) than the Ashizuri transect drill sites (segment I), which has a greater proportion of hemipelagic mud (Moore et al., 2001). The gradual along-strike changes in the trench wedge lithology do not correlate with the distinct segments that have wide protothrust zones at the Nankai Trough, so they are likely not a primary control over the protothrust zone width. However, they may have a secondary control that could explain the difference in protothrust zone width between segments I and III.

Protothrust Formation

The high-resolution data in this study revealed many protothrusts that have displacements greater than the seismic resolution (∼3 m), whereas deformation bands in cores mostly have displacements of <1 mm and thicknesses of 1 mm to 1 cm, which are below seismic resolution (Karig and Lundberg, 1990). This leads us to suggest that the protothrusts are fractures that may evolve from deformation bands.

Deformation bands can reduce the porosity of sediment by up to an order of magnitude, and once a rock becomes cohesive with reduced porosity, deformation tends to occur by fracture propagation. Patches of slip surfaces can form in deformation band clusters (Aydin and Johnson, 1978), which can connect and develop into a continuous slip surface or fault (Shipton and Cowie, 2003). The protothrusts appear to nucleate in the lower half of the trench wedge turbidites, where the sediment is more consolidated. This depth is where the protothrust offset and the frequency of protothrusts are greatest. The protothrusts then grow updip and downdip, which is consistent with protothrust growth models (Weiss et al., 2018) and a previous study at the Hikurangi margin (Barnes et al., 2018). Therefore, although deformation bands form in underconsolidated sediment in the seaward portion of the trench wedge, they only evolve into protothrusts when the strata hosting the bands have undergone sufficient consolidation.

The timing of consolidation within the trench wedge may be a key control over the formation of wide protothrust zones. If consolidation occurs too far from the frontal thrust, such as by gravitational stresses, by the time it is subjected to compressive tectonic stresses in the vicinity of the frontal thrust, it is too consolidated for deformation bands to form. Conversely, if consolidation within the trench wedge is delayed or insufficient as the sediment approaches the frontal thrust, then deformation bands will not form shear surfaces or evolve into protothrusts. Variations in consolidation rate and strain localization could be due to a number of factors discussed above, including the subducting lithology and basement topography and the trench wedge geometry and lithology. Previous studies have shown that sediment consolidation within the trench wedge increases gradually in both the vertical and horizontal directions where there is a wide trench wedge in western Nankai (Morgan et al., 1994). This is the location of a wide protothrust zone (segment I). Where the trench wedge is narrower in eastern Nankai (segment IV), lateral increases in consolidation only occur much closer to the frontal thrust (Morgan and Karig, 1995). The delayed, rapid increase in consolidation in narrow trench wedges is not conducive to significant protothrust formation. These studies agree with our hypothesis that the trench wedge size may influence the consolidation rate of the sediment and thus control the width of the protothrust zone at the Nankai Trough.

Mean protothrust spacing in wide, well-established protothrust zones is greater where the protothrust zone is wider (Fig. 12). However, the protothrust spacing at wide, well-established protothrust zones cannot be compared with the protothrust spacing at narrow or barely existent protothrust zones. This is because the protothrusts appear to be formed by fundamentally different mechanisms (Barnes et al., 2018). The relationship between protothrust spacing and the width of the protothrust zone could be explained by a self-similar growth model of protothrust zones. This would mean that the protothrust zone grows by a constant balance between the development of new protothrusts within the protothrust zone and the creation of new protothrusts seaward of the current deformation front. Analogous behavior occurs in deformation bands in high-porosity sandstones, which show a logarithmic decrease in deformation band frequency away from a fault core (Anders and Wiltschko, 1994; Schueller et al., 2013). The distribution of deformation bands remains self-similar as the width of the fault zone increases with time. The relationship between protothrust spacing and protothrust zone width is different between areas with a wide and narrow protothrust zone. This implies that the growth of protothrust zones occurs in a fundamentally different way in narrow and wide protothrust zones, and it is not simply a continuum where narrow protothrust zones will evolve into wide protothrust zones.

Implications for Strain Localization and Frontal Thrust Propagation

Protothrust spacing generally decreases landward, which may be due to the increase in consolidation and strain in the vicinity of the frontal thrust. However, clusters of closely spaced protothrusts with relatively higher cumulative displacement introduce heterogeneities in the seaward reduction in protothrust spacing (Fig. 6). We hypothesize that these clusters reflect strain localization and create zones of weakness along which the frontal thrust may propagate. This strain localization is well displayed in Figure 8D, where a cluster of protothrusts and growth strata occur in a concentrated area above a subducting basement scarp. The growth strata and seafloor uplift are interpreted to represent an incipient thrust fault. Once strain becomes localized seaward of the current frontal thrust, the strain rate on the current frontal thrust decreases significantly (Adam et al., 2005), and the incipient thrust fault becomes the new frontal thrust.

Most of the accreted protothrusts are likely to be inactive protothrusts that have been rotated to a steeper dip as the horizons underwent folding. However, it is plausible that some protothrusts may form landward of the frontal thrust in areas where the frontal thrust is young and the sediment within the thrust fault block is still undergoing rapid consolidation.

Areas with a well-established protothrust zone often have frontal thrusts that are listric, fault-propagation folds, with relatively low fault displacement (e.g., Figs. 6 and 9; Suppe and Medwedeff, 1990). In areas that are not adjacent to a wide protothrust zone, greater deformation along the frontal thrust occurs (Fig. 6), which is often accompanied by fault-bend folds (Suppe 1983). Fault-bend folds can evolve from fault-propagation folds where the frontal thrust has accommodated a large amount of strain (Marshak et al., 2019). In segment IV, where there is no significant subducting basement topography and no wide protothrust zone, the first few thrusts are fault-bend folds with large seafloor topography (130–240 m; Figs. 5 and 10). Conversely, in segment I, a wide protothrust zone occurs adjacent to a frontal thrust slice that has little displacement (40–140 m) and is largely associated with a fault-propagation fold (Figs. 5 and 7). Additionally, in segment III, where the protothrust zone is widest, the topography of the frontal thrust does not exceed 70 m (Figs. 3, 5, and 8). One possible explanation for the relationship between protothrust zone width and strain accommodation of the frontal thrust is that zones of weakness within well-established protothrust zones may promote forward propagation of the frontal thrust at a lower strain rate than at frontal thrusts that are not associated with a wide protothrust zone. Where there is no wide protothrust zone, strain at the accretionary prism toe is accommodated by greater displacement along the frontal thrust and within the thrust slice. Alternatively, the steep dip of the protothrusts may promote deformation by fault-propagation folding. Hughes et al. (2014) showed that fault-bend folding is favored at low fault ramp dips, whereas fault-propagation folding is favored at high fault ramp dips. If the frontal thrust evolves from a cluster of high-angle protothrusts, then the initial dip of the frontal thrust is steeper than would be predicted for Coulomb shear, which is ∼30° to the subhorizontal maximum compressional stress (Morgan et al., 2007).

We can infer temporal variations in accretionary processes along individual lines from the relationship between protothrust zone width and frontal prism geometry. The frontal 5–10 km of the accretionary prism in segment III has a very shallow slope (<3°), which is consistent with a wide protothrust zone. However, the prism steepens abruptly (6°–9°) above an increase in basement topography. The siliciclastic turbidite package is nonexistent or much thinner landward of the basement high. We determine that there was not a wide protothrust zone at the time the steeper portion of the wedge was accreting. Additionally, line 127 has the narrowest protothrust zone in segment III and appears to mark the present transition from narrow to wide protothrust zone. The sediment thickens seaward at the step in basement topography, which is directly beneath the frontal thrust (Fig. 5).

Abundant accreted protothrusts and a plateau in the prism slope in segment II may indicate that there was a wide protothrust zone prior to the recent subduction of a basement high (Fig. 9). There also appears to be a thick turbidite sequence (siliciclastic turbidite facies) beneath the accretionary prism, which onlaps the basement high. The coexistence of these characteristics further supports the suggestion that there is a relationship among sediment thickness, protothrust zone width, and frontal thrust geometry at the Nankai Trough subduction zone. Along line 121, the accretionary style may be transitioning from that with a wide to a narrow protothrust zone. The protothrust zone in line 121 is less than half as wide as the nearest line within segment III (Table 1), and the deformation front is pinned above the basement scarp. Seaward of the basement scarp, the sediment is considerably thinner, and there is no siliciclastic turbidite package. We predict that a fault-bend fold will form above the basement scarp in the absence of a wide protothrust zone, and the frontal wedge taper will increase, similar to the present accretionary processes occurring along lines 119 and 120.

Comparison with Other Margins

The only other subduction zone where comparable structural analysis of protothrust zones has been applied is the central Hikurangi margin of New Zealand (Barnes et al., 2018). The central Hikurangi margin has a convergence rate and accretionary processes that are similar to those in our study area. The dip angles of the protothrusts are also similar to those at Nankai and are consistent with compactive shear deformation bands. Additionally, the frontal thrust is a high-angle fault-propagation fold (Ghisetti et al., 2016; Barnes et al., 2018), which is consistent with those adjacent to a wide protothrust zone at Nankai. Barnes et al. (2018) also observed an increase in strain localization seaward of the frontal thrust and associated reduction of strain rate on the current frontal thrust. However, significant differences exist between the two subduction zones. Protothrusts in Hikurangi initiate further seaward of the frontal thrust than at the Nankai Trough (25 km at Hikurangi vs. <8 km at Nankai). The difference in protothrust zone width can be explained by the thickness of the trench wedge siliciclastic sequence. In central Hikurangi, the trench wedge is 2.2–2.5 km (>4 km total sediment), which is 2–3 times the thickness at Nankai. At Hikurangi, the protothrusts initiate within and are largely confined to the turbidite sequence, which implies that rheology-related differences due to lithology and/or consolidation state vary with depth (Barnes et al., 2018). The lithology beneath the turbidite sequence is not conducive to protothrust formation and limits the downdip extent of the protothrusts. Lithological differences between the Nankai Trough and Hikurangi margin may explain the differences between the two protothrust zones.

Differences in trench wedge lithologies may explain why some accretionary margins that have thick incoming sediment sections do not have protothrust zones. At the Barbados margin, the youngest sediment is calcareous ooze and mud (Moore et al., 1982), and there is no wedge of sandy trench turbidites and thus no protothrust zone. However, at the Makran accretionary margin and southern Hikurangi margin, the incoming sediment is exceptionally thick (7 km and 9 km, respectively) and consists of sandy turbidites (Kopp et al., 2000; Plaza-Faverola et al., 2012), but there is no protothrust zone. The thick sandy wedges are highly permeable and lose most of their fluid during normal compaction prior to accretion. As a result, the sediment is well consolidated when it approaches the deformation front (Fruehn et al., 1997), which inhibits the formation of protothrusts.

The relationship between protothrusts and consolidation is exemplified at the Cascadia and Chile margins (Cochrane et al., 1994; Han et al., 2017; Olsen et al., 2020). At the Cascadia margin, along the Washington transect, where the sediment is overconsolidated from ∼20 km seaward of the frontal thrust, there are no protothrusts. In contrast, along the Oregon transect, where a thick underconsolidated sediment sequence is being both subducted and accreted, the upper part of the sediment hosts protothrusts and becomes progressively consolidated (Han et al., 2017). The landward increase in velocity is more rapid in the protothrust zone than in the rest of the basin, which implies that consolidation is more rapid in the protothrust zone than in the rest of the basin (Cochrane et al., 1994). Along the Chile margin, protothrusts also only form in areas that have thick (∼2.8 km) underconsolidated incoming sediment (Diaz, 1999; Olsen et al., 2020) and show greatest consolidation within the protothrust zone.

The sediment consolidation at Cascadia is related to the décollement depth. Where the décollement lies near (<0.6 km above) the basement, the entire incoming sediment section is subject to horizontal compression from the accretionary wedge, leading to overconsolidation. Where there is a thick section of subducted sediment (∼1.4–1.7 km), the subducted sediment does not experience the horizontal compression from the wedge, and high porosity is maintained, resulting in a weak basal décollement (Han et al., 2017). A weak décollement results in less consolidation in the trench wedge (Davis et al., 1983), which is required for the formation of a protothrust zone (Wang et al., 1994). Consistent with the observations at Cascadia, the absence of a protothrust zone at the northern Hikurangi margin correlates with a thin (∼500 m) layer of subducting sediment (Pedley et al., 2010) as well as a thin trench wedge, rough crust, and subducting seamounts (Barnes et al., 2020), similar to segment II.

Along most of the Sumatran margin, little sediment is subducted (McNeill et al., 2017), and the trench fill is more consolidated than many other subduction settings (Hüpers et al., 2017). As a result, there is no protothrust zone. However, variations in the depth of the décollement and the amount of sediment subducted are still shown to influence the frontal thrust propagation and fault vergence (Bradley et al., 2019).

A protothrust zone is largely absent along the Alaska-Aleutian margin. Along most of the margin, thin incoming sediment is the primary reason for no protothrust zone (Bécel et al., 2017). One exception is offshore the Kenai Peninsula, in the eastern portion of the margin, where there is sufficient sediment thickness to produce an ∼12-km-wide protothrust zone, although less than five protothrusts were identified (von Huene et al., 1998). Seismic transects immediately to the northeast of that line do not have a protothrust zone despite having a wide and thick trench wedge and thick subducting sediment. These lines also have a narrow accretionary wedge and a steeper frontal wedge taper (Fruehn et al., 1999), which reflect the progressive evolution of the margin from erosion to accretion following the subduction of the Yakutat terrane (Fruehn et al., 1999). Recent tectonic erosion also occurred further west in the Semidi segment, where there is up to 3 km of trench fill and >900 m of subducted sediment but no protothrust zone (von Huene et al., 2012; Li et al., 2018). The recent subduction of major basement relief alters the local stress state, and there is commonly diminished lateral compression in the wake of subducted topography (Sun et al., 2020). Insufficient compressive stress in the trench wedge will inhibit the formation of a wide protothrust zone and/or significantly limit the number of protothrusts within the protothrust zone.

From the variations in protothrust formation between subduction margins, we conclude that thick trench wedge sediments must be underconsolidated and consist of sandy turbidites for protothrusts to form. We also find that consolidation increases more rapidly where protothrusts form. Furthermore, the margin must be in a well-established accretionary phase, as recent erosional phases appear to inhibit the formation of protothrust zones.

The protothrust zone at the toe of the Nankai Trough accretionary prism shows great along-strike variability. Two spatially discrete segments have a wide protothrust zone (∼3.3–7.8 km, ∼50–110 protothrusts), but these are bounded by segments with almost no protothrust zone (∼0.5–2.8 km, <20 protothrusts). At the Nankai Trough, the primary control over the presence of a wide protothrust zone seems to be the trench wedge size, which influences the progression of sediment consolidation. Secondary controls over the maximum width of wide protothrust zones include the accreting lithology and the subducting basement topography.

The majority of the Nankai protothrusts have a high dip angle, consistent with compactive shear bands in drill cores. We hypothesize that the protothrusts are fractures that form from shear surfaces in deformation band clusters that have undergone porosity reduction where the trench-fill sediment is more consolidated. Closely spaced protothrusts are associated with strain localization and may become the locations of frontal thrust propagation. Forward propagation of frontal thrusts as fault-propagating folds may occur at a lower buildup of strain than where mature protothrusts are absent. Where there is a narrow or no protothrust zone, there is greater displacement along the frontal thrust, which is probably due to greater strain buildup along the frontal thrust. Alternatively, the protothrusts may dictate the dip at which the frontal thrust forms and thus whether the frontal thrust is associated with a fault-bend or fault-propagating fold. Either way, it is clear that the protothrust zone influences deformation along the frontal thrust and thus the geometry of the accretionary prism.

We thank the captain and crew of the R/V Kaiyo for assistance during seismic data acquisition, and we greatly appreciate the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) for providing us with the seismic data. All seismic surveys were supported by the “Research Project for Compound Disaster Mitigation on the Great Earthquakes and Tsunamis around the Nankai Trough Region” of the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan. This paper was improved by the reviews of earlier drafts by Roland von Huene, Julia Morgan, Helen Janiszewski, and Garrett Ito. Special thanks go to Phil Barnes for his extremely thorough review and suggestions. We also thank Guest Associate Editor Dave Scholl for his timely handling of the manuscript. This work was supported by National Science Foundation grant OCE-1658580. This is University of Hawaii School of Ocean and Earth Science and Technology (SOEST) contribution 11176.

Science Editor: Shanaka de Silva
Guest Associate Editor: David W. Scholl
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.