Direct constraints on in situ stress state from deep drilling into the Nankai subduction zone, Japan

The Nankai Trough Seismogenic Zone Experiment (NanTroSEIZE) is a comprehensive investigation of subduction zone faulting and stress conditions (Tobin and Kinoshita, 2006; Tobin et al., 2019). NanTroSEIZE has combined seismic imaging with Integrated Ocean Drilling Program (IODP) drilling for direct sampling, in situ measurements, and long-term borehole monitoring to better understand the nature of the megathrust seismic cycle, fault locking, and the spectrum of fault slip. A transect of boreholes (Fig. 1) was drilled on a series of IODP expeditions from 2007 through 2019 (Tobin et al., 2019). The centerpiece is ultradeep drilling at IODP Site C0002 (Fig. 1B), which was targeted to cross and sample a seismic reflector interpreted as the main plate-boundary fault (the “megathrust”), which lies at ~5000 m below the seafloor (mbsf) based on reflection and refraction depth imaging (Moore et al., 2007; Bangs et al., 2009; Kamei et al., 2012).

The principal borehole (Fig. 2) at this site (Hole C0002F/N/P) was drilled by the riser drilling vessel Chikyu to 3058 mbsf, with steel casing cemented in place to a depth of 2922 mbsf (Tobin et al., 2015a, 2019; Strasser et al., 2014). One key objective at Site C0002 is to characterize the present-day state of stress in the inner accretionary wedge, which forms the upper plate of the primary plate-boundary fault (Fig. 1B). The orientations and absolute magnitudes of the three principal stresses—and their temporal evolution—drive fault strength, slip, and earthquakes (e.g., Scholz, 1998; Brodsky et al., 2020) throughout the seismic cycle (e.g., Magee and Zoback, 1993; Wang and Hu, 2006). Decades of effort to measure stress and pore-fluid pressure conditions, either directly or indirectly, however, have met with limited success (summarized in Saffer and Tobin, 2011), and the quantitative state of stress at depth in any subduction setting is not known with confidence. Previous NanTroSEIZE work using borehole breakouts and induced tensile fracture orientations from resistivity log imaging has established the orientation of the principal stress axes but not their magnitudes (Chang et al., 2010; Lin et al., 2015).

Unlike nearly all other scientific ocean drilling, this hole was drilled with a riser system, using a closed loop of drilling mud designed to clean the hole and provide pressure support to the borehole, permitting a number of otherwise impossible stress- and pore pressure-related observations (e.g., Saffer et al., 2013). We analyzed a suite of data sets collected during drilling that, taken together, provide quantitative constraints on the in situ stress tensor, including all three principal stresses and pore-fluid pressure.

We determined the vertical stress (S_v) directly using bulk density measurements from core samples and borehole cuttings returns (Kitajima et al., 2017). Values of bulk density ranged from ~1500 kg/m³ near the seafloor to ~2400 kg/m³ at 3 km depth. Integration of the bulk density profile yielded the vertical total stress:

where ρ_b is bulk density, g is gravitational acceleration, and z is depth below the seafloor.

To define the least principal stress, leak-off tests (LOTs) were performed at 872 and 1954.5 mbsf as the borehole was drilled and cased (Strasser et al., 2014; Tobin et al., 2015b). In LOTs, a short section of open borehole is pressurized by injecting drilling fluid into the sealed wellbore at a constant rate. The leak-off point (LOP) corresponds to the fluid pressure and volume at which the relationship deviates from a linear-elastic trend, and this point is interpreted to indicate fracture volume creation when the pressure in the sealed hole reaches a value comparable to the minimum stress (Zoback, 2007). Thus, the LOP provides a robust estimate of the least principal stress (σ₃). Inadvertent (but illuminating) mud losses during drilling, and accompanying pressure responses in a borehole observatory ~110 m to the southwest, were interpreted as additional indications of σ₃ at 872 mbsf and yielded values consistent with the LOT result (Tobin et al., 2015b).

We also performed step-rate injection tests (SRIT) at 2908–2920 mbsf that provided an additional measurement of σ₃ (Fig. S1 in the Supplemental Material¹). In this test, fluids were pumped into the closed borehole at a series of constant rates, and borehole pressure response was observed. This test yielded both a leak-off pressure during the injection phases and an indication of the instantaneous shut-in pressure (ISIP), interpreted to represent subsequent closure of fractures by the minimum stress after the pumps were shut down (e.g., Zoback, 2007). The LOP and ISIP yielded consistent estimates of σ₃.

Finally, several drill-string “pack-offs” occurred near the base of the borehole (~3000 mbsf) without indications of any losses of the circulating mud, suggesting that the accompanying spikes in fluid pressure in the borehole remained lower than the minimum tangential stresses at the borehole's circumference (Fig. 3A). Because the tangential stresses around a wellbore are a function of the differential stress in the horizontal plane (Zoback, 2007), the pack-off pressures sustained repeatedly by the wellbore provided a key constraint on the maximum horizontal stress (S_Hmax) magnitude (Fig. 3B).

To construct a depth profile of stress state, we augmented the data sets from drilling at Site C0002 with previously reported pressure and stress values at a second nearby riser drill site (C0009) that also penetrated the forearc basin and into the inner accretionary prism ~20 km landward of C0002, to a depth of 1594 mbsf (Fig. 1). These include in situ pore pressure measurements, mini-fracture tests to define σ₃ (Saffer et al., 2013), and analyses of wellbore failures that constrained S_Hmax (Ito et al., 2013; Saffer et al., 2013; Huffman et al., 2016).

The observations described above provided quantitative information on the magnitude of all three principal stresses at Sites C0002 and C0009 (Fig. 4). We assumed that the principal stresses were vertical and horizontal or very nearly so, as is dictated at shallow depths by the free surface at the seafloor (e.g., Anderson, 1951; Davis et al., 1983; Chang et al., 2010) and corroborated by the low dip of the décollement and narrow taper. We report all values of pressure and stress referenced to the seafloor (we subtracted a common hydrostatic pressure of 19.78 MPa from all values, corresponding to the ocean depth of 1939 m below the rig floor).

The vertical principal stress, S_v, was quantitatively well determined from integrated bulk density based on well logs and cores. It increased monotonically with depth and reached a value of 63–64 MPa at 3000 mbsf. The LOTs, mud losses during circulation, and SRIT provided measures of the least principal stress (σ₃) at several depths (Fig. 4). The LOT at 872 mbsf (Hole C0002F) yielded a LOP of 12.2 MPa (Strasser et al., 2014). Mud losses interpreted to reflect inadvertent hydraulic fracturing of the formation a few meters shallower are consistent with this value (Tobin et al., 2015b). The LOT at 1954 mbsf (Hole C0002P) yielded a value of 32.4 MPa. The SRIT at 2919 mbsf yielded a value of 47.1–47.3 MPa. These measurements, each separated in depth by ~1000 m, defined a linear gradient in the minimum principal stress of ~16.2 MPa/km and are in excellent agreement with values reported at nearby Site C0009 to a depth of 1534 mbsf (Saffer et al., 2013). These values of σ₃ are consistently less than S_v, indicating that σ₃ = S_hmin. Therefore, one fundamental and robust result is that the inner accretionary prism lies in a normal or strike-slip faulting regime throughout the entire drilled depth; a thrust-faulting regime requires S_v = σ₃, a condition not observed at any depth at Site C0002 or Site C0009.

Next, we analyzed the borehole pack-off events to obtain constraints on the maximum horizontal stress (S_Hmax; Fig. 3). Pressure spikes in the mud circulation system caused by these blockages corresponded to peak pressures of 61–63 MPa at the bottom of the hole over a 10 h period. During this time, no mud losses or pressure leaks were observed, as would be expected if the associated tangential (hoop) stresses around the borehole wall were sufficient to drive tensile fracturing (Zoback, 2007). The hoop stresses acting on the wall of a subvertical wellbore increase with the differential stress acting in the horizontal plane; hence, the containment of the pressure spikes provides a constraint on the upper bound of S_Hmax. Values greater than this bound would reach or exceed tensile failure conditions, and mud losses would have taken place (Fig. 3B).

There is some uncertainty in the exact depth of the pack-offs: the drill bit and bottom of hole were at 3034 mbsf, but the hole was in an open condition from 2922 mbsf (the base of the steel casing) to the bit depth (Fig. 2). We report S_Hmax as a range, including both this uncertainty in depth and a range of assumed values of tensile strength from 0 to 10 MPa (orange parallelogram in Fig. 4) based on laboratory experiments on core samples (Takahashi et al., 2013; Kitamura et al., 2019). The resulting S_Hmax lies between 49.9 MPa (shallowest assumed depth and zero tensile strength) and 63 MPa (greatest assumed depth and tensile strength of 10 MPa). For the full range of scenarios, S_Hmax remains less than S_v, and the value we report is consistent with the trend at shallower depths reported in previous studies (Ito et al., 2013; Chang and Song, 2016; Huffman et al., 2016; Fig. 4).

We did not have direct measurements of the formation pore-fluid pressure but inferred that it was close to the hydrostatic value based on three fundamental observations. First, resistivity logs collected during drilling showed mud invasion throughout much of the drilled interval, indicating that the mud weight we used (1.12–1.28 g/cm³) slightly exceeded formation pore pressure (Tobin et al., 2015b). Second, although shipboard data showed that the formation is gas-bearing, there were no significant gas or fluid volume “shows” during drilling, even when circulation of mud was stopped for pipe connections. Third, direct measurements of pore pressure at Site C0009 indicated hydrostatic pressure conditions to at least 1534 mbsf (Saffer et al., 2013). We concluded that the pore-fluid pressure was not substantially elevated above hydrostatic over the depth interval of drilling.

In summary, at 3000 m depth, the stresses were: S_v ≈ 64 MPa; S_Hmax ≈ 52–63 MPa; and S_hmin ≈ 49 MPa. These values define a normal faulting stress regime where S_v = σ₁ > S_Hmax = σ₂ > S_hmin = σ₃. If S_Hmax is close to the upper end of its permissible range, then the stress regime may be transitional to a strike-slip regime, with σ₁ ≈ σ₂. The total differential stress remained modest throughout the entire depth interval, reaching a maximum value of ~14–18 MPa at 3000 mbsf.

These observations deep in the inner accretionary wedge carry several key implications. Our analysis suggests that the present-day state of stress to 3 km depth is in the normal faulting to strike-slip—but not thrust—stress regime. Differential stresses are low, and the wedge is apparently not near Mohr-Coulomb failure in lateral compression (Fig. 4); in fact, it is far from this limit. This further suggests that the depth accessed by Hole C0002 is not in a critical state near failure. This is consistent with regional tectonic models for subduction, suggesting that the presence of an actively filling forearc basin indicates a stable elastic inner wedge (Fuller et al., 2006; Wang and Hu, 2006; Ruh, 2020). This may be in contrast to the outer wedge, which exhibits a tapered upper surface and numerous imbricate thrust faults (Fig. 1B).

Although differential stresses may remain low all the way to the plate boundary at ~5 km below the seafloor, S_Hmax must transition to become greater than the vertical stress, either spatially between 3 and 5 km depth or temporally during the interseismic period, in order to drive thrust motion along the plate boundary. In fact, 5–10 km seaward of Site C0002, Sugioka et al. (2012) reported a cluster of very low-frequency earthquakes (VLFEs) with thrust mechanisms at plate-interface depth, implying that the fault is in fact in a thrust regime. Intriguingly, To et al. (2015) found additional, smaller VLFEs at shallower depths within the accretionary wedge, which they concluded have normal faulting mechanisms. These VLFEs may lend support to the idea of depth-dependent variations in stress regime. Given that the last great (M8 class) earthquake here occurred in 1944, it is also possible that the upper plate exhibits temporal variations in S_Hmax during the seismic cycle. For either scenario, the low differential stresses we observed, the fact that S_Hmax < S_v, and the observation of recent and ongoing plate-interface slip events (Sugioka et al., 2012; Wallace et al., 2016; Araki et al., 2017) all require that the megathrust system operates and fails at small horizontal compressive and shear stresses.

The combination of observed thrust motion along the plate interface and the low values of S_Hmax and differential stress in the deep interior of the hanging wall likely requires variations in stress over short scales of time (years to decades) and space (thousands of meters), as well as a low-strength megathrust. These inferences from in situ stress data enabled by deep drilling provide new constraints on subduction seismogenic locking processes.

We thank the members of Integrated Ocean Drilling Program Expeditions 338, 348, and 358 and shipboard scientific parties, Japan Agency for Marine-Earth Science and Technology (JAMSTEC) Center for Deep Earth Exploration (CDEX) and MarineWorks Japan staff, and Chikyu drilling and ship crew and staff for their hard work on the Nankai Trough Seismogenic Zone Experiment (NanTroSEIZE) mission. We thank Jonas Ruh, John Platt, and one anonymous reviewer for helpful suggestions.