In situ stress magnitude and rock strength in the Nankai accretionary complex: a novel approach using paired constraints from downhole data in two wells

Understanding in situ tectonic stress and rock strength at seismically active margins is important for characterizing fault strength and failure mechanics, yet direct measurements of these quantities are rare. Hydrofracturing and leak-off tests are commonly employed in drilling to measure the minimum principal stress, but provide information only at specific depths and are costly (e.g., Bell 1990; Engelder 1993; Lin et al. 2008; Saffer et al. 2010; Hung et al. 2009). The direct measurement of maximum horizontal stress magnitude is more difficult, and few tools for this exist (e.g., Moran et al. 1993). As a result, indirect approaches have been widely used to estimate in situ horizontal stress orientation and magnitude from wellbore failures and from fault or fracture orientations measured in borehole logging data (e.g., Moos and Zoback 1990; Zoback and Healy 1992; Brudy et al. 1997; Zoback 2007; Chang et al. 2010; Ito et al. 2013; Lin et al. 2015) or from anelastic strain recovery measurements on core samples (e.g., Amadei and Stephansson 1997; Lin et al. 2006; Byrne et al. 2009). Similarly, in situ strength must be inferred indirectly from downhole drilling parameters (e.g., Kerkar et al. 2014) or defined from laboratory measurements (e.g., Horsrud 2001; Song et al. 2011; Chang et al. 2006). Laboratory measurements require core samples, are time-consuming, and must be upscaled to apply them in the field.

Recent work at subduction margins, including the Integrated Ocean Drilling Program (IODP) Nankai Trough Subduction Zone Experiment (NanTroSEIZE) (Fig. 1), has used the width of compressional wellbore breakouts to estimate the magnitude of the maximum and minimum horizontal principal stresses (S _Hmax and S _hmin) (e.g., Chang et al. 2010; Lin et al. 2013). Far-field stress is distorted in a predictable way at the circular free surface introduced by borehole excavation, such that the principal stresses are normal and tangential to the borehole wall, with the latter varying as a function of azimuth (e.g., Kirsch 1898). Breakouts form only at azimuths around the borehole where local differential stress at the borehole wall exceeds the rock’s failure criterion (Fig. 2a) (Zoback 2007). Thus, if rock strength is known, breakout widths measured by downhole logging can provide an indication of the far-field principal stress magnitudes (e.g., Moos and Zoback 1990; Chang et al. 2010). However, large uncertainties in assumed or estimated rock strength (e.g., unconfined compressive strength, UCS; friction coefficient, µ) lead to large uncertainties in the estimated stress magnitudes, because there is a direct trade-off between the two in determining the stresses required for rock failure (Huffman and Saffer 2016). In the absence of direct laboratory measurements, UCS is commonly assumed or estimated from P-wave velocity (V _p), using empirical fits for various rock types (e.g., Lal 1999; Horsrud 2001; Chang et al. 2006, 2010). However, it is unclear whether these empirical relations, even for shale, are representative of unlithified mudstones with porosity of 10–20 % which are typical of the upper few km of the subsurface, and for which few laboratory data are available.

The availability of downhole logging data from two adjacent (offset 70 m; Fig. 1) boreholes at Integrated Ocean Drilling Program (IODP) Site C0002 presents a rare opportunity to simultaneously constrain both rock strength and stress magnitudes through an in situ experiment. This differs from previous analyses, in which UCS has been assumed (or estimated from V _p) in order to then compute a range of S _Hmax consistent with observed wellbore failures. The two holes were drilled with the same diameter into the same sedimentary sequences and were affected by the same far-field stresses, but were drilled with different downhole annular pressures (APRS). Here, we combine observations of wellbore breakouts at the two sites to place bounds on both the in situ horizontal stress magnitudes and UCS.

Although previous work has focused on the in situ stress magnitudes (e.g., Chang et al. 2010; Lin et al. 2013; Yamada and Shibanuma 2015), the use of two holes allows us, for the first time, to concurrently constrain formation strength in situ. To accomplish this, we develop a forward model to predict breakout width in each borehole using direct measurements of APRS from measurement while drilling (MWD) tools and of S _hmin from leak-off tests. We then find combinations of S _Hmax and UCS that best match the paired observations of borehole breakouts measured in resistivity images from the two holes. From this, we define a vertical profile of in situ stress and strength, which constrains the full stress tensor in the lowermost Kumano forearc basin and uppermost accretionary prism (Fig. 1). We discuss the implications of these in situ stress magnitudes and UCS values for the faulting regime and for empirical relations between rock strength and P-wave velocity that are commonly used to constrain stress magnitudes using borehole breakouts.

The Nankai Trough is formed by northwestward subduction of the Philippine Sea Plate (PSP) beneath the Eurasian Plate offshore southern Honshu at ~4.1–6.5 cm/year (Fig. 1a) (Seno et al. 1993; Miyazaki and Heki 2001). The subduction zone is typified by Mw > 8 earthquakes that recur every ~100–200 years (Ando 1975). As the PSP subducts, part of the incoming sedimentary succession is offscraped at the trench, forming a ~100-km-wide accretionary prism (Fig. 1) (e.g., Moore et al. 1990). Major research initiatives to study the Nankai Trough include active and passive source seismic surveys (e.g., Moore et al. 1990, 2009), drilling, and analysis of associated core, cuttings, and geophysical data (e.g., Karig et al. 1975; Moore et al. 2001; Tobin et al. 2009). Scientific drilling in the Nankai Trough has included Ocean Drilling Program (ODP) Legs 131, 190, and 196 (Hill et al. 1993; Moore et al. 2001; Mikada et al. 2002) offshore of Shikoku Island, and several IODP Expeditions offshore the Kii Peninsula as part of the NanTroSEIZE experiment (Expeditions 314–316, 319, 322, 326, 332–333, 338, and 348) (Kinoshita et al. 2009; Tobin et al. 2009, 2015; Saffer et al. 2010; Saito et al. 2010; Kopf et al. 2011; Henry et al. 2012; Moore et al. 2013). These efforts have provided important information about the plate boundary and outer forearc, including its overall structure, fault zone properties, and hydrogeology.

We focus on IODP Site C0002, located approximately 30-km landward of the trench along the NanTroSEIZE drilling transect, above the inferred seismogenic portion of the plate boundary (Fig. 1) (Tobin et al. 2015; Moore et al. 2013). Operations at Site C0002 included the drilling of multiple holes (Fig. 1). We analyze data from Holes C0002A and C0002F, two vertical boreholes separated by 70 m in map view (Fig. 1). The boreholes penetrated ~918 m of the Kumano forearc basin fill, composed of Pliocene-to-Pleistocene mudstones and turbidites, and into the uppermost several hundred meters of the underlying accretionary wedge, composed of Miocene silty claystones with lesser amounts of sandstone and sandy siltstone (Ashi et al. 2008, 2009; Moore et al. 2013). Holes C0002F and C0002A provide data with overlapping coverage from 872 to 1401 meters below the seafloor (mbsf).

During Expedition 314 in 2007, Hole C0002A was drilled as a dedicated logging-while-drilling (LWD) hole and penetrated from the seafloor to 1401 mbsf in a riserless drilling mode (Tobin et al. 2015). In riserless drilling, the annular wellbore pressure is not actively controlled, but is governed by the hydrostatic pressure plus any additional pressure associated with cuttings in the well and dynamic circulation during drilling. Azimuthal electrical resistivity images from Hole C0002A document well-developed and continuous breakouts throughout the hole, oriented at azimuths of 135°–315° ± 45° (Fig. 2b). This indicates that S _Hmax is oriented at ~45°–225°, subparallel to the trench (Lin et al. 2010; Chang et al. 2010). The observed breakout angular widths range from 20 to 140°, with an average of 70° (Ashi et al. 2009). There is little change in S _Hmax orientation from the Kumano basin into the accretionary prism below (Lin et al. 2010; Chang et al. 2010). Previous analyses of the breakout widths to estimate far-field stress magnitudes suggest a normal faulting regime (overburden stress or S _v = σ ₁) within the Kumano basin (<918 mbsf), which transitions to a strike-slip or thrust regime (maximum horizontal stress or S _Hmax = σ ₁) in the accretionary prism below (>918 mbsf) (Chang et al. 2010).

Hole C0002F was drilled to 872 mbsf in riserless drilling mode in 2010 with a limited set of LWD measurement tools (Expedition 326 Scientists 2011). It was deepened to 2006 mbsf in a riser drilling mode in 2012 (Strasser et al. 2014) and to 3058.5 mbsf in 2014 (Tobin et al. 2015) with a full suite of LWD tools. In riser drilling, annular pressure is controlled by the use of weighted drilling mud and return circulation in a riser pipe that connects the borehole to the drilling platform at the sea surface through the water column. For Expedition 338, mud with a density of 1100 kg/m³ was used. Leak-off tests (LOTs) conducted in Hole C0002F at 874 and 1954.5 mbsf indicate a minimum principal stress, defined from leak-off pressures (e.g., Zoback 2007), of 12.6 and 33.6 MPa, respectively, at these depths (pressures are referenced to zero at the seafloor) (Strasser et al. 2014; Tobin et al. 2015). In both cases, the leak-off pressures are less than the total vertical stress (S _v), indicating that S _hmin is the minimum principal stress. This indicates that the stress state lies in either a normal or strike-slip faulting regime throughout the section, which is broadly consistent with observations of abundant normal faults in the Kumano basin in seismic reflection data (e.g., Sacks et al. 2013) and with previous estimates of stress magnitudes from wellbore breakouts (Chang et al. 2010). In the underlying accretionary prism, seismic data document thrust splay faults that were active until ~0.4 Ma, but these thrust faults do not disrupt younger sediments (Boston et al. 2016). If these structures remain active at rates less than resolvable by the seismic data or are near-failure, the stress state at depth must transition to a thrust faulting regime (i.e., S _v is the minimum principal stress, and S _Hmax is the maximum principal stress). Such a transition is also expected at greater depth closer to the active plate boundary megathrust.

Our approach identifies combinations of UCS and far-field horizontal stress magnitude that are consistent with observed breakout widths at the two adjacent boreholes (i.e., measured breakout widths where they are documented, and zero width where there are no breakouts). We use a forward model to define expected breakout widths for a wide range of boundary conditions and rock properties and then identify the combinations of far-field stress magnitude and UCS that best reproduce measured breakout widths at Hole C0002A and the lack of breakouts at Hole C0002F, which were drilled with different annular pressures. We report all pressures and stresses referenced to zero at the seafloor.

Because breakout widths are strongly sensitive to the values of S _Hmax and UCS, and because these two quantities are the least constrained of the parameters that affect breakout formation, we treat them as unknowns and specify values for the other parameters on the basis of existing data. In previous studies of wellbore breakouts in the Nankai accretionary complex, a friction coefficient of 0.6 was assumed, but few direct laboratory tests were available to constrain this. Triaxial tests on core samples of Miocene mudstones from Sites 1174 (outboard of the trench offshore Shikoku Island) and C0002 indicate values of µ = 0.48 (Olcott and Saffer 2013). Based on these data, we assume µ = 0.48 in our analysis, but also explore the effect of this choice by considering a case where µ = 0.6 for an example at 1150 mbsf. We assume Poisson’s ratio is 0.45, following Peacock et al. (2010) and Tsuji et al. (2011) from inverted P- and mode-converted S-wave velocities from active-source ocean bottom seismometers and from the time lag between P-refracted waves and PPS-converted waves from active-source ocean bottom seismometers, respectively.

We define the overburden stress (S _v) by integrating density data obtained from core samples (Ashi et al. 2009; Strasser et al. 2014). We assume that formation pore pressure (P _p) is hydrostatic based on sedimentation models for the Kumano basin, consolidation tests on core samples, and direct measurements conducted during drilling at nearby IODP Site C0009 (Guo et al. 2013; Saffer et al. 2013). Values of annular pressure (APRS) in both holes were recorded continuously during drilling (Table 1) (Ashi et al. 2009; Strasser et al. 2014). We define S _hmin using the results of the two LOTs, one in Hole C0002F and one in Hole C0002P. Within the Kumano basin (<918 mbsf), we fit a linear gradient from S _hmin = 0 MPa at the seafloor to S _hmin = 12.6 MPa defined by the LOT at 874 mbsf in Hole C0002F (Strasser et al. 2014). For the accretionary prism (>918 mbsf), we define a second linear gradient between the LOT at 874 mbsf and a LOT at 1954.5 mbsf, where S _hmin = 33.6 MPa in Hole C0002P (Table 1) (Tobin et al. 2015). To account for uncertainty in the leak-off tests themselves and in extrapolation along assumed linear gradients between the measurements, we consider a range of S _hmin of ±2 MPa around the values defined by these gradients.

In our forward models, we evaluate a range of potential S _Hmax values based on the limits for frictional failure along pre-existing faults (Zoback 2007) and assuming that the principal stresses are horizontal and vertical (e.g., Chang et al. 2010). For our lower bound of S _hmin, this defines a range of S _Hmax from 11.5 to 32.2 MPa at 900 mbsf and 20.5 to 42.9 MPa at 1386 mbsf. Due to the absence of reliable data on strength of poorly lithified mudstones, we consider a broad range of possible UCS values, from 0 to 20 MPa, consistent with laboratory tests on shale (e.g., Lal 1999; Horsrud 2001).

At all depths, there is a narrow range of best-fit UCS–S _Hmax pairs for a given value of S _hmin (Fig. 3). As an illustrative example of residuals at a given S _hmin within the range defined by LOT estimates ±2 MPa, we show two example sets of residuals: one at 900 mbsf in the Kumano basin where S _hmin = 12.9 MPa (Fig. 3a) and one at 1150 mbsf well within the prism (>918 mbsf) where S _hmin = 19.6 MPa (Fig. 3b). Similar analyses for the six depths where breakouts are observed in Hole C0002A but not in Hole C0002F yield a depth profile of S _Hmax, ranging from 11.9 to 22.1 MPa at 900 mbsf to 20.8–26.6 MPa at 1386 mbsf (Table 1; Fig. 4a). The corresponding profile of best-fit in situ UCS values ranges from 5.6 to 20 MPa at 900 mbsf to 3.2–8 MPa at 1386 mbsf (Table 1; Fig. 4b). Uncertainty in S _hmin is the primary cause for the large ranges in both S _Hmax and UCS (Eq. 1; Fig. 4). In contrast, the coefficient of friction has a relatively minor effect. Varying µ from 0.48 to 0.6 results an increase in best-fit UCS and S _Hmax values of only ~1 MPa (~10 % of UCS and ~2.5 % of S _Hmax).

Previous estimates of stress magnitude from borehole breakouts at Site C0002 were computed using rock strength values defined from empirical relationships between P-wave velocity (V _p) and UCS for shales and were formulated without constraints on S _hmin, which are now constrained by leak-off tests (e.g., Chang et al. 2010). These previous analyses suggested a stress state that lies in the normal faulting regime within the Kumano basin and transitions to a strike-slip or reverse faulting regime in the accretionary prism. This interpretation is broadly consistent with observations of normal faults in the Kumano basin (Sacks et al. 2013) and with major thrust splay faults within the prism (Moore et al. 2009). However, the LOTs show that at least to ~2000 mbsf, S _hmin is the least principal stress (i.e., S _hmin is less than S _v), limiting stresses to a normal or strike-slip regime in this depth range both at Site C0002 (Strasser et al. 2014; Tobin et al. 2015) and at Site C0009 located ~20-km landward (Ito et al. 2013; Saffer et al. 2013). The inconsistency between our result and the imaging of thrust faults in the accretionary wedge could indicate that these structures are no longer active (Boston et al. 2016), that S _Hmax and S _hmin increase with depth to become the maximum and intermediate principal stresses, or that the stress regime changes through the earthquake cycle and will build toward a thrust regime as strain accumulates during the interseismic period (e.g., Wang and Hu 2006). The hypothesis that horizontal stress magnitudes fluctuate during the seismic cycle is broadly consistent with temporal variations in stress interpreted from analyses of calcite crystallography in samples from the accretionary prism (Takeshita et al. 2014).

Compared to previous constraints on stress from breakouts that do not include data from two boreholes, our analysis narrows the range of S _Hmax to 1400 mbsf and also indicates either a normal or strike-slip faulting regime in this depth range. Uncertainties remain too large to distinguish between the two faulting regimes, because the range of S _Hmax we predict overlaps with S _v (Fig. 4a) and the range of S _hmin we consider (±2 MPa from LOT gradients). In total, the available observations suggest that if thrust faults in the wedge are active, then the stress regime must transition to one of thrusting (S _hmin > S _v) at depths >2000 mbsf or must evolve in time to drive thrust faulting later in the seismic cycle (e.g., Wang and Hu 2006; Sacks et al. 2013; Hardebeck 2012; Lin et al. 2013; Huffman and Saffer 2016).

The values of UCS determined from our analysis are in general agreement with, but fall slightly below, empirical relations between V _p and UCS for shale (e.g., Lal 1999; Horsrud 2001) and are in good agreement with the trend in rock strength defined by laboratory experiments on a limited set of core samples from the Nankai margin (Song et al. 2011; Olcott and Saffer 2013) (Fig. 4c). Sonic logs spanning the depth range we analyzed document V _p values that cluster near 2400 m s⁻¹ (for depths 900–1150 mbsf) and 2800–3000 m s⁻¹ (depths 1300–1386 mbsf) (Table 1). Our forward model results indicate a range of UCS that is broadly consistent with empirical relations in the shallower zone, but lower than predicted by these relations for the deeper interval. In combination, the values of UCS defined by our analysis and by direct laboratory tests (Olcott and Saffer 2013) suggest that empirical relations between V _p and UCS for shale may overestimate rock strength slightly; this would result in an overestimate of far-field tectonic stress derived from breakout widths (e.g., Chang et al. 2010; Lin et al. 2013).

Quantification of in situ horizontal stress magnitudes and rock strength is challenging, typically owing to a lack of data needed to constrain models. The problem is further complicated by the fact that most methods employed to estimate in situ stress magnitudes require independent datasets to constrain rock strength. Here we address this problem by using data from two adjacent boreholes drilled with different boundary conditions (annular pressures) that sample the same volume of the shallow crust in an active tectonic setting, and where the minimum horizontal stress is measured independently by leak-off tests. Our approach simultaneously defines the best-fit in situ maximum horizontal stress magnitude (S _Hmax) and rock unconfined compressive strength (UCS), using a forward model of compressional borehole breakouts that is constrained by observations of wellbore failure, or lack of wellbore failure, at the two adjacent holes. This ultimately allows estimation of the full stress tensor at several depths in the Kumano forearc basin and underlying upper accretionary prism. We find that S _Hmax ranges from 11.9 to 22.2 MPa at 900 mbsf to 20.8–26.6 MPa at 1386 mbsf. Predicted values of in situ UCS range from 5.6 to 20 MPa at 900 mbsf to 3.2–8 MPa at 1386 mbsf.

Our results suggest that the in situ stress state lies in a normal to strike-slip faulting regime. This is consistent with a range of independent observations, including widespread recent normal faulting in Kumano basin (e.g., Sacks et al. 2013), and in situ wireline hydraulic fracturing measurements conducted while drilling ~20-km landward (Saffer et al. 2013). Our results also indicate that empirically derived relations commonly used to estimate rock strength from P-wave velocity for shale may slightly overestimate UCS in mudstone and could lead to overestimates of far-field stresses determined from breakout widths.