Analysis of normal fault populations in the Kumano forearc basin, Nankai Trough, Japan: 2. Principal axes of stress and strain from inversion of fault orientations


  • Alison Sacks,

    1. Department of Geosciences, The Pennsylvania State University, University Park, Pennsylvania, USA
    2. Now at BP Exploration Alaska, Inc., Anchorage, Alaska, USA
    Search for more papers by this author
  • Demian M. Saffer,

    Corresponding author
    1. Department of Geosciences, The Pennsylvania State University, University Park, Pennsylvania, USA
    • Corresponding author: D. M. Saffer, BP Exploration Alaska, Inc., 900 East Benson Blvd., Anchorage, AK 99508-4254, USA. (

    Search for more papers by this author
  • Donald Fisher

    1. Department of Geosciences, The Pennsylvania State University, University Park, Pennsylvania, USA
    Search for more papers by this author


[1] We use a high-resolution 3-D seismic survey to map a population of recent normal faults within the Kumano Basin of the Nankai subduction zone, in order to quantify patterns of strain and stress state over the last 0.44 Myr. We identify distinct fault populations that define three phases of extension. Phases 1 and 2 comprise NW-SE striking faults located along the western basin edge and in the northwestern portion of the study area, respectively. The NE-SW striking faults of phase 3 comprise the largest population, and extend ~20 km landward from the basin's seaward edge. Phase 2 faults typically terminate within a few reflectors of the seafloor, whereas most phase 3 faults form seafloor scarps. Inversion of the fault populations documents NE-SW extension during phases 1 and 2 and NW-SE extension during phase 3, consistent with both core-scale structures and horizontal stress orientations observed at Integrated Ocean Drilling Program (IODP) boreholes. Slip on phase 3 faults accommodates strain of up to ~1–2%, concentrated near the basin's seaward edge. Inversion for a best-fit stress tensor yields a subvertical σ1 and subhorizontal σ2 and σ3 for all faulting phases. We find that during phase 3 in most portions of the basin, σ2 = σ3 (SHmax = Shmin), reflecting widely varying fault strikes. This contrasts with distinct SHmax and Shmin magnitudes inferred from IODP borehole data; these observations may be reconciled if the orientation of maximum horizontal stress fluctuates due to variation of subduction parallel compression through the seismic cycle.

1 Introduction and Background

[2] In subduction zones, strain and stress in the upper plate reflect conditions and mechanical properties along the underlying plate boundary fault, and the material properties of the overriding and downgoing plates [e.g., Davis et al., 1983; Wang and Hu, 2006]. Forearc basins, common features at many convergent margins [e.g., Berglar et al., 2008; Contardo et al., 2008; Dickinson, 1995], provide a record of sedimentation and deformation on the upper plate that has been linked to upper plate and subduction megathrust mechanical behavior. For example, Wells et al. [2003] and Song and Simons [2003] suggest that large forearc basins correlate with asperities on the plate boundary that may act as source regions for great earthquakes. On the basis of a mechanical model of upper plate stresses, Wang and Hu [2006] suggest that forearc basins should form above a stable inner wedge, which overlies the interseismically locked portion of the plate interface. In the model of Fuller et al., [2006], sedimentation in the forearc basin stabilizes the wedge through the maintenance of a horizontal slope above a steepening decollement. These models predict that large forearc basins and the underlying inner accretionary wedge should undergo little permanent deformation during the seismic cycle, and may be characterized by extensional stresses following large earthquakes [e.g., Wang and Hu, 2006]. However, direct observations of strain and stress in these settings are usually limited to the expression of deformation at the seafloor, and to regions accessible to drilling.

[3] Offshore southwest Japan, the Nankai Trough accretionary complex is the focus of the Nankai Trough Seismogenic Zone Experiment (NanTroSEIZE), a major Integrated Ocean Drilling Program (IODP) drilling project, and of a 3-D seismic survey acquired in support of drilling [Kinoshita et al., 2009; Moore et al., 2009] (Figure 1). The high-resolution seismic volume images a complete transect of the margin, and structures within 2 km of the seafloor are imaged in particularly excellent detail [e.g., Moore et al., 2009] (Figures 2 and 3). A large population of recent and active normal faults in the forearc basin provides an opportunity to document basin-wide patterns of both strain and stress. In this study, we (1) investigate the timing, orientation, and magnitude of strain and stress in the Kumano Basin over the last 0.44 Myr through structural and kinematic analyses of this fault population, and (2) discuss these results in the context of borehole and core-scale indicators of deformation and stress. This work follows from a companion paper by Moore et al. [2013], which characterizes the orientations and timing of normal faults in the basin using seismic coherency analysis.

Figure 1.

(a) Regional map of the Nankai Trough. Inset shows regional tectonic setting. Area of Figure 1a indicated in red on inset map. PSP = Philippine Sea Plate; EP = Eurasian Plate; PP = Pacific Plate. The red box in Figure 1a outlines the region shown in Figure 1b. The Kumano 3-D seismic survey (black outline) and IODP drill sites (black circles) are also shown. (b) Map of the 3-D seismic survey area showing survey boundary (dashed white line), borehole locations (black circles), and major structural features and interpreted bathymetry (labeled, after Moore et al., [2009] and Kimura et al., [2011]). White lines show the seismic lines used for our detailed strain analysis (Figures 3, 5 and supporting information). Longest inline L2455 is the transect shown in Figure 2. Traces of faults mapped in the seismic volume are shown for a depth of 150 mbsf (dark gray lines). Maximum horizontal stress (SHmax) orientations (red bars) are defined from wellbore failures at IODP drill sites during Expeditions 314, 315, 316 and 319 [Lin et al., 2010]. At drill sites C0002 and C0009, SHmax rotates clockwise below 936 and 1285 mbsf, respectively (blue bars). Yellow arrows indicate far-field convergence of the Philippine Sea Plate and Japan [Miyazaki and Heki, 2001].

Figure 2.

Interpreted cross-section along central inline (L2455) shown in Figure 1b, showing NanTroSEIZE boreholes (after Moore et al. [2009]). Interpretation of major structural features after Moore et al. [2009]. Faults interpreted as part of this study are shown in grey in forearc basin. Grey box indicates area shown in Figure 3.

Figure 3.

Example (a) uninterpreted and (b) interpreted portions of seismic inline 2455 in the Kumano Basin (location shown in Figure 1b, area shown in Figure 2), showing examples of mapped faults. An older generation of normal faults is interpreted in the landward-most part of the survey (grey); these are shown for the sake of completeness, but are not included in our inversions.

1.1 Geologic Setting and Previous Work

[4] The Nankai Trough is formed by the subduction of the Philippine Sea Plate beneath the Eurasian continental plate at an azimuth of 300°–315° and a velocity of 4.1–6.5 cm yr−1 [Seno et al., 1993; Miyazaki and Heki, 2001] (Figure 1). Along the NanTroSEIZE transect, the sediment section on the subducting plate includes ~1.1 km of Miocene pelagic and hemipelagic sediments of the Shikoku Basin, overlain by a ~1.3 km thick package of trench deposits [e.g., Moore et al., 2009] (Figure 2). The modern accretionary prism incorporates the entire thickness of the trench deposits and at least half of the underlying Shikoku Basin sequence [Park et al., 2002]. The Nankai Trough, with its long record of great earthquakes and tsunami [Ando, 1975], has been the focus of numerous studies investigating subduction zone processes, including geodetic surveys [Aoki et al., 1982], seismic reflection surveys [Park et al., 2002; Moore et al., 2009], and IODP and Ocean Drilling Program (ODP) drilling [e.g., Karig et al., 1975; Kagami et al., 1986; Taira et al., 1991; Moore et al., 2001; Kinoshita et al., 2009; Saffer et al., 2010]. As part of the NanTroSEIZE project, recent IODP expeditions have drilled 11 boreholes spanning from seaward of the frontal thrust to 25 km landward of the forearc high in the Kumano Basin (Figures 1 and 2).

[5] The Kumano forearc basin is bounded at its seaward edge (~30 km from the trench) by a prominent bathymetric “notch” associated with a network of transpressive faults termed the Kumano Basin Edge Fault Zone (KBEFZ) [Martin et al., 2010]. Just seaward of the KBEFZ, a major out of sequence splay fault, termed the “megasplay”, separates the active outer accretionary wedge from an inner wedge that is characterized by higher seismic velocity [Park et al., 2002; Moore et al, 2009] (Figure 2). The 3-D seismic survey covers the seaward-most ~25 km of the basin (Figures 1 and 2). The basin sediments consist primarily of mudstones with sandy intervals. The section ranges from ~1 to >2.5 km thick, and is <2 Myr old [Expedition 314 Scientists, 2009; Saffer et al., 2010]. The landward dip of strata increases progressively with proximity to the forearc high, and bed dips are progressively gentler in younger strata, indicating successive tilting associated with slip on the megasplay fault [Strasser et al., 2009; Gulick et al., 2010]. The forearc basin sediments unconformably overlie Miocene age sediments of the accretionary prism [Expedition 314 Scientists, 2009; Saffer et al., 2010]. The basin sediment package thickens significantly in the landward portion of the basin, generally due to an increased depth to the top of the prism, but also enhanced locally by the infilling of several major depressions. The entire section is cut by normal faults estimated to have initiated from ~1.0 to 0.3 Ma [Gulick et al., 2010]. Many of these faults are recently active, penetrating to within a few reflectors of the seafloor or forming seafloor scarps (Figure 3; see also supporting information).

[6] NanTroSEIZE drilling has recovered cores, collected logs, and measured a suite of downhole properties across a transect from the incoming plate to ~55 km landward of the trench [Kinoshita et al., 2009] (Figure 2). The orientations of maximum and minimum horizontal stresses (SHmax and Shmin, respectively) defined from wellbore failures (borehole breakouts and drilling-induced tensile fractures) in these holes demonstrate that in the outer wedge, the azimuth of maximum horizontal compression is subparallel to the subduction direction [Lin et al., 2010]. At Site C0002, ~5 km landward of the basin's seaward edge, SHmax is subparallel to the trench (rotated 90° relative to that in the outer wedge). At Site C0009, 20 km landward, SHmax is again subparallel to the subduction direction. This suite of results is consistent with horizontal stress orientations inferred from seismic velocity anisotropy within the upper plate, which indicates a gradual rotation of SHmax with distance landward from the area of Site C0002 [Tsuji et al., 2011].

2 Methods

2.1 Seismic Interpretation and Fault Mapping

[7] The Kumano 3-D seismic survey was acquired in April 2006 by Petroleum GeoServices (PGS). Acquisition geometry yielded nominal 30-fold data [Moore et al., 2009]. Pre-stack (3-D) depth migration (PSDM) was performed at Japan Agency for Marine Earth Science and Technology (JAMSTEC)-Institute for Research on Earth Evolution (IFREE) [Moore et al., 2009]. The resulting seismic volume, 12 km wide by 56 km long, spans from the toe of the accretionary wedge to ~55 km landward. The estimated vertical depth resolution is ~ 5–7 m near the seafloor, and decreases to ~10–20 m in the lower basin strata 2–3 km below the seafloor [Moore et al., 2009].

[8] In this study, we mapped faults in the depth-migrated seismic volume using Landmark SeisWorks 3-D, and correlated the faults through the volume on every inline. We computed strike and dip for each fault from best-fit planes. Maximum displacement (assuming pure dip-slip), and the along-strike location of maximum displacement for each fault were also documented [Sacks, 2011]. Crosscutting and stratigraphic relationships define the relative ages of fault activity. To analyze spatial and temporal variations of fault characteristics, stress state, and strain, we divided the area of the forearc basin covered by the 3-D seismic volume, in which faults were observed, into eight subregions of approximately equal size (Figures 1b and 4).

Figure 4.

Map showing our study region, divided into eight subregions as described in the text. Lines labeled A and C represent locations of transect lines shown in Figure 5. Traces of all faults are shown at 150 mbsf, with stereonets of poles to fault planes in each subregion. Faults strike in nearly every direction, but form two significant clusters striking NW-SE and NE-SW. NE-SW striking faults are concentrated in the seaward subregions (1 through 6). Planar NW-SE striking faults are primarily in the western subregions (2, 4, and 6). Fault strikes in every subregion span nearly the entire range of strikes observed in the entire basin. In the landward region of the survey, there are fewer faults with more varied orientations. Fault dips average 58° (σ = 5°), and range from 45° to 82°. Landward-dipping faults are more common and generally exhibit a narrower range of dips than seaward-dipping faults.

2.2 Strain Analysis

[9] We inverted the fault population for maximum extension direction using T-TECTO 3.0 [Žalohar and Vrabec, 2008]. This program uses the Moment Tensor Summation method for determining maximum extension direction and rotation [Molnar, 1983; Marrett and Allmendinger, 1990, 1991]. For each individual fault, shortening and extension axes are defined and weighted by the maximum displacement. These axes are then summed to calculate the direction of maximum extension for a given fault population. We calculated maximum extension direction for the entire fault population, and for subpopulations defined by relative age and geographic subregion. In addition, we selected six transects, three parallel and three perpendicular to the survey axis, to document absolute extension and its distribution (Figure 3; see also supporting information). We compute the total horizontal extension along each transect by summing the heaves of individual faults. In a companion paper, Moore et al. [2013] determine total extension by balanced cross-section restoration, and obtain similar values to those we report here.

2.3 Stress Inversion

[10] We inverted each fault subpopulation defined by age and geographic subregion for a best-fit effective stress tensor using T-TECTO 3.0 [Žalohar and Vrabec, 2007], which follows the method of Angelier [1977, 1979, 1984]. The basic assumptions governing this inversion are that (1) slip occurs in the direction of maximum shear stress resolved on each fault; (2) slip on faults grouped in a given population resulted from a single stress tensor; and (3) the distribution of fault planes is independent of stress, thereby allowing for reactivation of faults and yielding a stress tensor solution more strongly dependent on slip direction than on fault plane orientation. Without clearly identifiable piercing points in the seismic data, we assumed pure dip-slip motion on all faults. This assumption is consistent with the observation that conjugate fault pairs typically exhibit horizontal intersections.

[11] T-TECTO 3.0 uses two mechanical parameters, internal and residual friction angle, to define the range of slip directions—if any—that are consistent with a given stress tensor. The range of friction angles was defined using conjugate angles of faults in the population we mapped [Twiss and Moores, 1992; Sacks, 2011]. Based on the four dominant orientations of conjugate faults in our dataset, with bisecting angles of 63° and 70°, we define an internal friction angle of 35° (internal friction coefficient, μi = 0.7), and a residual friction angle of 30° (μr = 0.58). A detailed sensitivity analysis verifies that the choice of friction angles has a negligible effect on the resulting stress orientations, but does have a modest effect on the relative magnitudes of the principal stresses [Sacks, 2011].

[12] To determine a best-fit stress tensor, the sum of angular misfit between the observed slip direction and the direction of maximum resolved shear stress on every fault in a given population is minimized, for a suite of trial stress tensors. Delineation of individual fault populations is necessary to obtain meaningful results from this inversion method. Faults and features resulting from separate events must be differentiated, on the basis of geologic or statistical data, in order to resolve distinct stress tensors for each event. In this study, the primary distinction is made using relative ages determined through crosscutting and stratigraphic relationships. Geographic association is a secondary criterion by which subpopulations are defined.

3 Results

3.1 Fault Orientations and Age Relationships

[13] We mapped 435 normal faults cutting the uppermost basin strata (Figures 35 and supporting information). The seaward portion of the survey area (subregions 1 through 6, Figure 4) contains the largest number of NE-SW striking faults. The western region (subregions 2, 4 and 6) also contains a planar NW-SE striking population. Fault strikes in every subregion span nearly the entire range of strikes observed in the entire basin, but strikes in the seaward portion of the basin cluster more tightly about NE-SW and NW-SE. In the landward region of the survey, there are fewer faults, but more varied orientations. Observable NE-SW striking faults do not occur farther than 25 km from the basin edge.

[14] Fault dips average 58° (σ = 5°), and range from 45° to 82°. Landward-dipping faults comprise 70% of the total population, and generally exhibit a narrower range of dips than seaward-dipping faults, particularly in the seaward-most part of the basin. The faults strike in nearly every direction, but form two significant clusters striking NW-SE and NE-SW (Figure 4). The NE-SW striking cluster contains 66% of all faults mapped in the basin; these are distributed across the faulted portion of the basin. In this population, strikes of conjugate planar faults cluster in three groups: 045/225, 065/245, and 080/260 (reported as azimuthal strikes of the conjugate pairs, using the right-hand rule). Within each of the three groups, the strikes vary by ± 15°. A fourth group of NE-SW striking faults is characterized by arcuate geometry. These arcuate faults, localized to the southeastern corner of the mapped region, span a 60° range of strikes from 060/240 to 120/300, with mean of 085/265.

[15] Faults striking NW-SE comprise the remaining 34% of the population, falling into two subsets. One subset is planar, with average strikes of 155/335 ± 21°, occupying the SW corner of the survey area. The other subset contains arcuate faults with an 80° range of strikes that range from 085/265 in the seaward portion of the basin, to 165/345 in the landward-most part of the 3-D survey. This subset of arcuate faults exhibits the smallest maximum displacements of any subset, averaging ~8 m. Average maximum displacement for all other faults is ~11 m, with a range of 5–45 m.

[16] Crosscutting and stratigraphic relationships define the relative ages of fault activity. The NW-SE striking faults are always crosscut by NE-SW striking faults. The two groups of NW-SE striking faults (arcuate and planar) occupy distinct regions and do not intersect; we use stratigraphic relationships to define their relative timing (Figure 5). Planar faults in the SW corner of the survey do not penetrate above a horizon that is clearly offset by the arcuate faults in the NE corner. This horizon is dated at 0.44 Ma at IODP Site C0009 [see also Moore et al., 2013]. These crosscutting and stratigraphic relationships define three phases of faulting: (1) an oldest generation of planar NW-SE striking faults (phase 1; corresponding to “phase 2” described in the companion paper by Moore et al. [2013]); (2) arcuate faults striking approximately NW-SE that cut the 0.44 Ma horizon (phase 2; corresponding to Moore et al.'s 2012] “phase 3”); and (3) NE-SW striking faults that crosscut the other faults (phase 3; corresponding to Moore et al.'s [2012] “phase 4”). Our phase 2 and 3 faults cut almost the entire forearc basin sediment package, typically to within a few seismic reflectors of the seafloor. Most (61%) of phase 3 faults form scarps at the seafloor, and scarps are associated only with this fault population.

Figure 5.

Examples showing crosscutting and stratigraphic relationships between the three phases of faults and a horizon dated at 0.44 Ma used to define relative fault ages. (a) Uninterpreted seismic data, and (b) interpreted section showing older phase 1 (blue) faults cut by phase 3 (red) faults. (c, d) Uninterpreted and interpreted seismic data showing phase 2 (green) faults cut by younger phase 3 (red) faults. (e) Interpreted crossline demonstrating stratigraphic relationships between phases 2 and 3 (green and red) faults cutting the 0.44 Ma horizon (purple), and older phase 1 (blue) faults that do not offset it.

3.2 Extension Axes and Magnitude

[17] Basin-wide maximum horizontal extension direction has rotated from NE-SW to NW-SE through time, from N068°E–N248°E during phase 1, to N038°E–N218°E during phase 2, to N158°E–N338°E during phase 3 (Figure 6). Furthermore, populations of faults defined by subregion and relative age indicate clear directions of maximum extension in time and space (Figure 6). Where present, both phases 1 and 2 record NE-SW extension. Phase 1 accommodates maximum extension directions of N059°E–N239°E, N073°E–N253°E, and N069°E–N249°E in subregions 2, 4, and 6, respectively. Phase 2 accommodates maximum extension directions ranging from N027°E–N207°E in subregion 3 to N041°E–N221°E in subregion 7 on the eastern side of the survey, and orientations of N005°E–N185°E and N041°E–N221°E in subregions 6 and 8. The wide range of extension directions corresponds to fanning strikes exhibited by these arcuate faults, which rotate from NW-SE to nearly N-S with distance landward from the basin edge. Phase 3 reflects NW-SE extension throughout the basin, ranging from N165°E–N345°E in the southeast corner (subregion 1) to N153°E–N333°E at the landward extent of active faulting (subregion 7).

Figure 6.

Base maps of study area with fault populations separated by age and subregion. Fault traces are shown for a depth slice at 150 mbsf. Small stereonets show great circles and slip vectors for faults in each subregion. Arrows in each subregion on the map indicate the direction of maximum extension defined by our inversions. Orientations of SHmax at IODP sites C0002 and C0009 are shown for reference. SHmax in upper part of borehole indicated in red, and in lower part of borehole indicated in purple. Arrows on stereonets beneath each map indicate the direction of overall maximum extension for all faults in each phase, and illustrate the rotation of maximum extension over time.

[18] Detailed analysis of extension along selected inlines and crosslines in the seismic survey (which are subparallel and subperpendicular to the dominant extension axes; cf. Figure 6) provides insight into the spatial distribution of extensional strain (Figure 7). Parallel to the direction of subduction, the total extension associated with mapped faults ranges from 130 to 280 m, yielding an average of 1.4% and a maximum of 1.8% extensional strain in the inline direction. Perpendicular to the direction of subduction, total extension ranges from 36 to 180 m, corresponding to an average of 1.0% and a maximum of 1.6% extensional strain along the width of the survey. Maximum extension along the inline direction, normal to the margin, is 280 m over a total faulted shown length of 15 km. It is important to note that the strains and total extensions we report should be considered minima, because our observations are restricted to structures with displacements of > ~5–7 m (limited by the resolution of the seismic data), and any small-offset faults below the resolution of the seismic data would contribute to the total extension.

Figure 7.

(a) Distribution of strain along selected inlines and traces as described in the text. Strain was measured as horizontal separation (heave) of terminations of reflections cut by faults along the inlines and crosslines indicated. Total length is shown by green bars, and strain is shown by red arrows. For reference, an arrow scaled to 1% strain is shown. (b) Symmetric rose diagram showing cumulative extension in each phase of faulting.

3.3 Stress

[19] We inverted faults from each phase and subregion to obtain best-fit effective stress tensors as a function of location and age (Figure 8). In all cases, σ1 is subvertical and σ2 and σ3 are subhorizontal, as expected for a normal faulting regime [Anderson, 1951]. Inversions of all phase 1 faults, and phases 2 and 3 faults in some subregions, yield distinct orientations and magnitudes for σ2′ (≈ SHmax) and σ3′ (≈ Shmin) (Figure 8). The orientation of σ2 for phase 1 faults, which are restricted to the SW side of the basin, ranges from 08/148 (dip angle/dip direction in degrees) in subregion 1, to 02/167 in subregion 6 (Figure 8a). Phase 2 faults on the eastern side of the survey indicate σ2 orientations ranging from 01/137 in subregion 3 to 02/118 in subregion 7; our inversion yields equal values of σ2 and σ3 for subregions in the NW portion of the basin (Figure 8b). For phase 3 faults, variations in slip direction are sufficiently large that our inversion yields equal values of σ2 and σ3, with the exception of subregion 7, where we obtain an σ2 orientation of 03/238 (Figure 8c).

Figure 8.

Stress orientations for phases (a) 1, (b) 2, and (c) 3 within each subregion of the surveyed portion of the basin. Maximum principal stress (σ1) is subvertical in all cases. Ellipses show the orientations of SHmax and Shmin, scaled such that the lengths of principal axes reflect the relative magnitudes of the two horizontal effective stresses. Orientation of SHmax from borehole breakouts at C0002 and C0009 is shown by red and purple arrows at the borehole locations.

[20] In cases where our inversion yields distinct minimum and intermediate principal stresses, the ratio of effective stresses is: σ1′:σ2′:σ3′ = 1.0:0.35:0.27 (Figure 9). This condition applies primarily in phases 1 and 2, and to subregion 7 in phase 3 (Figure 8). In cases where σ2′ = σ3′, which describes the majority of inversion results for phase 3 faults, the ratio σ1′:σ2′:σ3′ = 1.0:0.27:0.27 (Figure 9). In all cases the effective stress ratio (σ3/σ1) is 0.27, satisfying the failure criterion for normal faulting given the internal and residual friction angles inferred from conjugate faults. Previous studies have shown that pore fluid pressures are near hydrostatic through much of the Kumano Basin [Guo et al., 2013; Saffer et al., 2010, 2013]. This allows us to quantify the magnitude of effective stresses, based on our inversion results and the total overburden stress (σv) defined by integration of density logs. These data define an effective vertical stress gradient (σv′) of 6.14 MPa/km, and maximum (σ2′) and minimum (σ3′) effective horizontal principal stress gradients of 2.15 MPa/km and 1.66 MPa/km, respectively.

Figure 9.

Mohr circles summarizing stress states (normalized to the effective vertical stress, σv′) for the two outcomes of our inversion: (a) cases where σ2 > σ3; and (b) cases where σ2 = σ3. Our results indicate a near-vertical σ1 and near-horizontal σ2 and σ3 in all cases (i.e., σv′ = σ1′). All fault planes are plotted as black points. Most fall between the failure envelopes defined by the angles of internal and residual friction (ϕ = 35° and ϕ = 30°, respectively). Using a vertical stress σv computed from density logs, normalized magnitudes of σ3 obtained from downhole pressure tests (Modular formation dynamics tester = MDT, Leak-off test = LOT) at site C0009 (red points) [Saffer et al., 2013] are shown for comparison.

4 Discussion and Implications

4.1 Spatial and Temporal Extension Patterns

[21] Our results show that the maximum overall horizontal extension direction in the Kumano Basin has rotated from NE-SW to NW-SE through time, from N068°E–N248°E during phase 1, to N038°E–N218°E during phase 2, to N158°E–N338°E during the most recent phase of fault activity (phase 3) (Figure 6). This is highly consistent with analysis of faults and shear zones in core samples from Site C0002, which documents two distinct generations of extension at ~ N060°E–N240°E and ~ N170°E–N350°E (Figure 10) [Expedition 315 Scientists, 2009]. The directions and timing of extension indicated by these structures correlate closely with those derived from the normal fault populations in our phases 1 and 3 (Figures 6 and 10). Our phase 2 extension (N038°E–N218°E) is not represented by observed structures in the core; this is consistent with the observation that Phase 2 faults mapped in the seismic data do not occur near Site C0002. Byrne et al. [2009] determine a maximum extension direction of ~ N140°E–N320°E from anelastic strain recovery (ASR) of core samples from Site C0002. The results of anelastic strain recovery represent the most recent deformation in the forearc basin, and are generally consistent with the results of our inversion for phase 3 faults, and with the analyses of core-scale structures from site C0002.

Figure 10.

Great circles and slip directions for features observed in cores at C0002 [after Expedition 315 Scientists, 2009]. (a) Shear zones with observed normal offset (red) indicate maximum extension at N060°E–N240°E, assuming dip-slip. Shear zones in grey have no observable offset. (b) Faults (black) crosscut shear zones, and indicate maximum extension at N170°E–N350°E. Measured striae (red) indicate oblique-normal slip directions on features oriented NNE-SSW. All other faults are assumed to undergo dip-slip. (c) Black and grey arrows show maximum extension directions indicated by core faults and shear zones, respectively, and extension directions (colored arrows) defined by our inversion.

[22] Although our stress tensor solutions for the region around site C0002 do not yield discrete orientations of maximum and minimum horizontal stress (i.e., σ2 = σ3), it is useful to compare the maximum extension directions from inversion of the fault populations with the orientation of maximum horizontal stress inferred from borehole breakouts and tensile failures [e.g., Lin et al., 2010]. At site C0002, an SHmax (≈ σ2) orientation of N030°E–N210°E [Expedition 314 Scientists, 2009] is generally consistent with the maximum extension directions we define for the most recent (phase 3) faults that cut the seafloor (Figure 6). At site C0009, the maximum extension direction indicated by the nearest faults (phase 2) is N041°E–N221°E. This is also consistent with the current SHmax orientation of N148°E–N328°E derived from borehole breakouts (Figure 6) [Lin et al., 2010]. However, the faults nearest site C0009 are not associated with the most recent phase of extension in the basin and do not form scarps at the sea floor. The fact that SHmax orientations at site C0009 are consistent with an older phase of faulting suggests that current stress conditions are likely similar to those under which the nearby phase 2 normal faults formed. Although these faults do not exhibit scarps, they often penetrate to within a few seismic reflectors beneath the seafloor.

[23] Park et al. [2002] attribute normal faulting in the forearc basin to gravity-driven slumping behind the forearc high due to uplift along the megasplay. If this is the case, deformation in the basin expressed as tilting, faulting, or extension, should be largest near the forearc high. During the deposition of forearc basin strata, landward tilting was the dominant mode of deformation, particularly near the seaward basin edge [Gulick et al., 2010]; however, there is no bias toward steeper landward-dipping faults across the basin or near the forearc high, as would be expected if any landward tilting of the basin had occurred contemporaneously with—or postdated—fault formation.

[24] Summing heaves for each phase of faulting as a function of extension direction provides insight into the relationship between the magnitude, direction, and timing of extension (Figure 7). Extension parallel to the direction of subduction is accommodated by phase 3 faults, and far outweighs the extension during earlier phases of faulting. As noted above, this recent extension is concentrated near the seaward edge of the basin. Several mechanisms have been proposed to explain recent trench-normal extension focused in the outer Kumano Basin. On the basis of mechanical models, Conin et al. [2012] show that slip on a frictionally weak megasplay fault should result in patterns of both stress and strain consistent with observations. Bangs et al. [2009] suggest that an increased thickness of subducted sediment has locally steepened the megasplay fault and caused uplift at the forearc high, leading to the observed extension.

[25] Alternatively, Kimura et al. [2011] describe a series of bathymetric lineaments to the southeast of the survey area, which they interpret as normal faults accommodating upper plate extension driven by seamount subduction. It is possible that the same process has driven uplift and extension of the Kumano Basin. This is supported by the fact that faults mapped in the seismic volume are parallel to, and in some cases extensions of, the lineaments described by Kimura et al., [2011] [see also Sacks, 2011]. Our data provide some additional evidence supporting the idea that extension in the SW portion of the basin may be related to basement uplift; in this region, phase 1 (NW-SE striking) faults exhibit a bias toward steeper dips in the northeast-dipping subpopulation, and gentler dips in the southwest dipping subpopulation. This suggests recent uplift and tilting of the SW corner of the basin, as also argued by Gulick et al., 2010 and Moore et al. [2013].

[26] To investigate the detailed spatial distribution of strain, we separated each inline transect into four segments divided internally at subregion boundaries, and at each end by the extent of faulting mapped in the seismic volume (Figure 7). The seaward end of each transect is the edge of the forearc basin; the landward end of each transect is the seismic trace beyond which we observe no NE-SW striking faults. Trace-parallel transects were divided at the central octant boundary and extend to the edges of the survey. The highest strains and absolute extensions occur in the central and southeastern portion of the survey area [Sacks, 2011] (Figure 7). Absolute extensions of 55 m (L2235), 114 m (L2455) and 101 m (L2645) in the segment nearest to the forearc high correspond to the highest strains along each inline transect, ranging from 2.0% (L2235) to 2.6% (L2455). Strain generally decreases landward and westward, reaching a minimum of 0.2% (L2235). Despite the uncertainties associated with the limited resolution of the seismic data, our estimates of total extension and strain are similar to values of ~2% reported in a companion paper by Moore et al. [2013], based on balanced cross-section restoration.

4.2 Stress Magnitude and Orientation

[27] Our stress inversion yields a ratio of effective principal stresses (σ1′:σ2′:σ3′) of 1.0:0.35:0.27 for most portions of the basin during phases 1 and 2 (Figures 8 and 9a), where σ1 is subvertical, and σ2 and σ3 are subhorizontal (i.e., the maximum and minimum horizontal principal stresses). Our inversions yield an SHmax azimuth ranging from N148°E–N328°E to N167°E–N347°E during phase 1, and from N118°E–N298°E to N137°E–N317°E during phase 2 (Figure 8). For phase 3, our inversion indicates a stress ratio of 1.0:0.27:0.27 throughout the basin, with the exception of the northeastern-most subregion near site C0009 (Figures 8 and 9b). This result reflects the wide range of fault orientations in the most recent population.

[28] Horizontal stress magnitudes and orientations at sites C0002 and C0009 have also been estimated by previous studies [Chang et al. 2010; Saffer et al., 2013]. At site C0002, Chang et al. [2010] report the magnitude of minimum horizontal stress from borehole breakout widths, which define a range of effective stress ratios (σhmin′/σv′) from 0.28 to 1.39 and indicate a predominantly normal faulting regime in the basin sediments. The lower end of this range is comparable to the values we report. As discussed above, the orientation of SHmax defined by wellbore failures at site C0002 is consistent with the dominant extension axis defined by phase 3 faults, but our stress inversion yields no clear orientation for maximum horizontal stress. At site C0009, leak-off tests (LOT) and hydraulic fracturing measurements of the minimum principal stress document effective stress ratios (σhmin′/σv′) of 0.44 and 0.77, respectively [Saffer et al., 2013] (Figure 9). These ratios document a normal faulting regime, but with a differential stress below the failure criterion. This is consistent with the observation that site C0009 is ~10 km landward of any mapped active normal faults (cf. Figures 4 and 6).

[29] The stress state from our inversion of the most recent fault population that cuts the seafloor (phase 3) is the most appropriate to compare with present day stresses inferred from wellbore failures and ASR on core samples [Byrne et al., 2009; Lin et al., 2010]. As noted above, our inversion yields σ2 = σ3 (the two horizontal stress magnitudes are equal), whereas the indicators of present day or very recent stress from borehole and core data indicate distinct magnitudes and orientations of the two. This discrepancy may be explained by differences in the timescale sampled by the different stress indicators. Borehole breakouts record only instantaneous stresses at the time of drilling, and ASR documents only recent and recoverable deformation. In contrast, our inversion reflects all stress states that drove faulting, over the entire timespan represented by the population.

[30] During the period spanned by phase 3 faulting in the seaward portion of the basin, subtle changes in the orientation of the horizontal principal stresses would allow slip on normal faults having the range of observed strikes, if the two stresses were close in magnitude. One possible explanation is that stress states within the basin vary throughout the seismic cycle [e.g., Wang and Hu, 2006] (Figure 11). In the late interseismic period, strain accumulation along the décollement and megasplay will drive compression in the upper plate, such that SHmax would be rotated toward the subduction direction (Figure 11a). As a result, slip on normal faults striking at orientations other than trench-parallel would be permitted. Depending on the degree of stress rotation, this could include the arcuate faults striking obliquely to the trench, or perhaps those striking nearly normal to the trench (e.g., phase 2 and/or faults in the landward portion of the basin). Following a great earthquake, post-seismic extension associated with the release of stress along the plate boundary fault, possibly accompanied by uplift along the megasplay, should drive normal faulting in the hanging wall and allow dip-slip motion along normal faults striking parallel to the megasplay [e.g., Conin et al., 2012] (Figure 11b). In this set of scenarios, the orientations of SHmax and Shmin (σ2 and σ3) would switch in response to fluctuations in the magnitude of stress parallel to plate convergence, modulated by the seismic cycle.

Figure 11.

Schematic block diagram illustrating stress changes during the seismic cycle that could allow faults of widely varying orientations to slip. (a) Late in the interseismic period, convergence-parallel stress in the upper plate driven by strain accumulation along the décollement and megasplay leads to a stress regime in which normal faults striking normal or obliquely to the trench undergo slip (purple planes). (b) Following a large earthquake, post-seismic extension reduces the trench-normal compression, leading to a rotation of horizontal stresses and inducing slip on normal faults parallel to the trench (orange planes).

5 Summary and Conclusions

[31] Mapping of a large recent and active normal fault population in the Kumano Basin documents three phases of extension. The oldest phase ended ~0.44 Ma, whereas the two later phases are currently or recently active. Inversion of these groups for the direction of maximum horizontal extension indicates that it has changed through time from NE-SW, approximately perpendicular to plate convergence, to NW-SE, subparallel to plate convergence. This finding is consistent with: (1) observations of core-scale structures from IODP Site C0002 that indicate a similar rotation of extension axes; and (2) modern horizontal principal stress orientations determined from wellbore failures and ASR at sites C0002 and C0009 [Chang et al., 2010; Byrne et al., 2009; Lin et al., 2010]. Extensional strain is concentrated in the outermost ~15 km of the forearc basin, where fault density and throw on individual structures are both largest.

[32] Inversion of the fault orientations for stress tensors indicates a nearly vertical σ1 and subhorizontal σ2 and σ3, consistent with a normal faulting regime. All fault populations define a minimum to maximum effective stress ratio (σ3/σ1) of 0.27. Inversion of the phase 1 and some phase 2 faults yields relative stress magnitudes of 1.0:0.35:0.27 (Figure 8), whereas the phase 3 inversion yields equal magnitudes of σ2 and σ3, reflecting the widely varying strikes of this fault population and exhibit. In situ measurements of the least principal stress carried out at site C0009 indicate a minimum effective stress ratio higher than that derived from our inversion, but consistent with a normal faulting regime. Faults in the vicinity of site C0009 are recently, but not currently active, and exhibit no seafloor scarps; stress magnitudes derived from inversion of these fault data may reflect a former stress state. Active normal faults are characterized by a wide range of orientations, both parallel and perpendicular to the trench, implying that horizontal stresses may be modulated over the earthquake cycle.


[33] This work was supported by NSF Award #0451602 to D.S. We thank Tom Canich and Heather Nelson for substantial computing and software support. Landmark Graphics Seisworks 3-D was the primary software tool used in this study. We thank Ros King and an anonymous reviewer for comments that improved the clarity of the manuscript.