## Abstract

We present regional *in situ* stress analyses based on publicly available log and pressure data from coal seam gas developments in the Permian Bowen basin, Australia. Together with earlier data from the eastern part of the Jurassic Surat basin, our results show a broad, but systematic, rotation of *S*_{Hmax} azimuths in this part of eastern Australia as well as systematic changes in stress state with depth. Overall, the geomechanical state of the region appears to reflect the interplay between basin-controlling structures and a complex far-field stress regime. At the reservoir level, within and between Permian coal seams, this stress complexity is reflected in highly variable stress states both vertically and laterally. Stress data, including direct pressure measurements and observations of borehole failure in image logs, have been used to calibrate sonic-derived one-dimensional wellbore stress models that consistently exhibit a change in tectonic stress regime with depth. Shallow depths (<600 m) are characterized by a reverse-thrust stress regime and deeper levels are characterized by a strike-slip regime. Changes in the stress state with depth influence the mechanical stratigraphy of rocks with widely contrasting mechanical attributes (coals and clastic sediments). Our results highlight the interdependency between regional tectonic, local structural and detailed rheological influences on the well scale geomechanical conditions that have to be taken into consideration in drilling and completion designs.

**Supplementary material:** Database of additional wells with image log data are available at https://doi.org/10.6084/m9.figshare.c.3785849

The Australian continent is characterized by significant variability in the magnitude and orientation of *in situ* stresses (Coblentz *et al.* 1995, 1998; Hillis *et al.* 1999; Hillis & Reynolds 2000, 2003; Reynolds *et al.* 2002, 2003) at both the continental and regional scale. This is well documented from oilfield data, particularly in the eastern–central interior basins (Reynolds *et al.* 2005; Nelson *et al.* 2007). The *in situ* stress distribution in the Australian plate is controlled by plate boundary forces acting on the Australian plate (Fig. 1). The key plate tectonic elements bracketing the Australian plate include the divergent southern margin between Australia and Antarctica, transpressional convergence at the southeastern plate margin, compression along the northern and northwestern plate margin (particularly the Papua New Guinea fold–thrust belt and the Himalayan collision zone) and subduction at the northeastern margin (Indonesian Arc; Reynolds *et al.* 2002, 2003; Sandiford *et al.* 2004).

The Bowen basin (Figs 1 & 2) is interpreted as a Permian to Triassic back-arc basin (Holcombe *et al.* 1997*a*, *b*; Korsch & Totterdell 2009) and is one of a series of rift basins that developed across eastern Australia from the Early Permian (Korsch *et al.* 2009*a*). Between *c.* 265 and 230 Ma, the Bowen basin and the New England Orogen were subjected to contractional and strike-slip deformation known as the Hunter Bowen Orogeny (Holcombe *et al.* 1997*b*; Korsch *et al.* 2009*b*). This deformation initiated basin-bounding fault systems (Fig. 2). The Bowen basin is broadly characterized by two north–south-trending depocentres, the Denison and Taroom troughs (Fig. 2), with internal half-graben structures that initiated during the Early Permian. The early Permian basin-fill is dominated by fluvio-lacustrine clastic successions.

The deposition of thick, mid–late Permian coal measures (2–15 m), particularly in the eastern part of the basin, occurred during thermal sag, which was followed by significant late Permian inversion of local half-graben. In the Bowen basin, one to three coal seams of economic interest can be developed at the well scale. Late Permian and Triassic contraction led to complex reactivation structures and the deposition of late Permian–Early Triassic fluvio-marine clastics; deposition ceased during a mid–late Triassic contractional event. The development of permeability in coal seam gas fields is loosely associated with structural highs.

The southeastern part of the Bowen basin is overlain by the Surat basin (grey coloured area in Figs 2 & 3), which formed a broad intracontinental depression during the Jurassic. Surat basin sediments are dominated by siltstones and mudstones with minor sandstones, all of which were deposited in a fluvio-lacustrine depositional environment. All the clastic lithologies contain significant volcanic components, resulting in low porosity and permeability. The Surat basin contains multiple coal seams (on average 60) with an average thickness of 0.4 m. These coal seams form the Walloons fairway (Fig. 2).

Structuring in the Surat basin is subtle and reflects the more intense deformation of the underlying reactivated inversion structures in the Permian Bowen basin. Triassic structural highs set up gentle, low-amplitude highs in the Surat basin, which are characterized by exceptional permeabilities ranging from hundreds of millidarcies (mD) to multidarcies (e.g. Undulla Nose, Fig. 2). High permeability regions are characterized by coals with multiple fracture orientations readily identifiable on image logs (i.e. no preferred fracture orientation).

Coal seam methane has been produced in SE Queensland for over 20 years. The industry saw a marked acceleration of drilling and broad data acquisition in the early 2000s, when several world-scale coal seam gas to liquefied natural gas projects commenced construction and production. Early production was mainly from vertical wells with or without hydraulic fracture completions. Ongoing optimization of both drilling design and hydraulic fracture completions has fostered a wealth of diagnostic data acquisition and some design changes (Flottmann *et al.* 2013; Kirk-Burnnand *et al.* 2015). Initial work on the geomechnical framework (Brooke-Barnett *et al.* 2015) has shown the variability of stress magnitudes and the potential influence of fundamental basement and basin structures.

The focus of this study is the *in situ* stress state in the context of the geology and geomechanics of the Permian Bowen basin and the Roma Shelf region of the Surat basin (Fig. 2). The wireline data utilized in this study are publicly available at QDEX (2016). The results presented here complement earlier studies conducted in the eastern part of the Jurassic Surat basin (excluding the Roma Shelf region) to the south of the Bowen basin (Figs 2 & 3; Flottmann *et al.* 2013; Brooke-Barnett *et al.* 2015). The data presented here show significant three-dimensional complexity and granularity in the stress tensor (the relative magnitude of the principal stresses) and their plan view orientation in an area with hitherto sparse datasets; the World Stress Map (Heidbach *et al.* 2008, 2010) shows a comparatively uniform distribution of the maximum horizontal stress (*S*_{Hmax}) in eastern Queensland. Similarly, both the magnitude of differential stresses and the Andersonian stress state (reverse/strike-slip/normal; Anderson 1951) varies significantly with depth. The geomechanics of the Bowen and Surat basins are uniquely influenced by the geological setting because the stratigraphic column vertically juxtaposes lithologies with starkly contrasting rheological properties (e.g. clastic sediments v. coals).

This paper has four objectives:

Documenting the significant plan view variability of the

*S*_{Hmax}orientation in an intracontinental basin setting utilizing a comprehensive and regionally extensive*in situ*stress dataset based on >180 wells, of which 145 wells present new data.Establishing systematic variations of stress geometry with depth from representative examples of one-dimensional wellbore stress models based on log-derived strain-based stress calculations.

Discussing the implications of both the lateral and vertical stress variability on the bulk mechanical stratigraphy which, in turn, influence completion strategies, such as hydraulic fracture stimulation and inclined and horizontal drilling.

Assessing the

*in situ*stress and geomechanical implications based on data from the Permian Bowen basin (presented here) with existing data and interpretations from the Jurassic Surat basin.

## Data, conditioning and calibration

The basic stratigraphic correlations are from standard well logs (gamma ray, density, resistivity, two-arm caliper; for stratigraphic overview, see Cook & Jell 2013). Integrating the density log and extrapolating the trend to the surface allows calculation of the vertical stress magnitude. The integral of the density logs with reference to depth gives the gradient of the vertical stress (*S*_{V}), usually around 1 psi/ft (*c.* 19.2 ppg, *c.* 22.6 kPa m^{−1}) (equation 1).

**Vertical (overburden) stress at depth z (Pa):**
(1)
where

*z*is the depth below ground level, ρ is the density in kg m

^{−3}and

*g*is the acceleration due to gravity (assumed to be 9.81 m s

^{−2}).

The orientation of *S*_{Hmax} (the maximum horizontal stress) is derived from borehole breakouts and the drilling-induced tensile fractures (DITFs) observed in image logs. Breakouts form due to conjugate microfracturing and shear failure at the wellbore wall in response to hoop stresses (Kirsch 1898). Breakout is a product of far-field stresses interacting with the wellbore wall and elongating the wellbore in the direction of *S*_{hmin} (the minimum horizontal stress; see Bell 1990, 1996*a*, *b*), leading to an overall oval shape of the wellbore. DITFs form in the azimuth of *S*_{Hmax}; they form sharp, usually linear, fractures where the tensile hoop stresses are greater than the tensile rock strength at the wellbore wall. Both breakouts and DITFs are sensitive to mud-weight changes and the magnitude of the differential stresses (in vertical wellbores, *S*_{Hmax}−*S*_{hmin}) applied to the wellbore wall. For a given far-field stress state, elevated mud-weights can suppress the initiation of borehole breakout, whereas high mud-weights can, in turn, initiate DITFs. The wells used here are usually drilled either under balance or slightly over balance with respect to the hydrostatic gradient of.433 psi/ft (*c.* 8.3 ppg; 9.8 kPa m^{−1}). Wireline logging is undertaken after the wellbore is filled with a 3% KCl brine under slightly overbalanced conditions. All logs are referenced and corrected to true north. Stress measurements were ranked according the classification scheme of the World Stress Map (Tingay *et al.* 2008). The stress data are displayed on a basin structure map generated using the SEEBASE method of integrating various datasets to generate a best approximate of a ‘depth-to-basement’ structure image (SEEBASE 2005).

One-dimensional wellbore stress models are used to constrain the relative magnitude of *S*_{hmin} and *S*_{Hmax} (in relation to *S*_{V}) with depth. One-dimensional wellbore stress models are based on Poisson's ratio (equation 2) and Young's modulus (equation 3), which are derived from dipole sonic and density wireline data (dynamic data). A dynamic to static conversion of Poisson's ratio and Young's modulus was derived regionally from rock mechanics laboratory measurements (equations 4, 5, 6, 7). Minimum and maximum horizontal stresses were then calculated using poroelastic stress equations (Eaton 1968, 1972, 1975; Thiercelin & Plumb 1994; equations 8 & 9), which incorporate the static Poisson's ratio, vertical stress, pore pressure, the static Young's modulus and Biot's coefficient, as well as tectonic strain in the minimum (ε_{min}) and maximum (ε_{max}) horizontal stress directions. Nominal tectonic strain values of ε_{max}=0.0009 and ε_{min}=0.0003 were used for the initial calculations before calibration (Brooke-Barnett *et al.* 2015). Pore pressure was calculated based on a freshwater hydrostatic gradient, which is commonly observed in undepleted coal reservoirs. Biot's coefficient was not independently constrained and was set as 1 to ensure consistency across the basin, thus true variation in poroelastic strain constants can be assessed.

**Dynamic Poisson's ratio (ν _{dyn}) (no units):**
(2)
where

*V*

_{p}is the compressional sonic velocity in m s

^{−1}and

*V*

_{s}is shear sonic velocity in m s

^{−1}.

**Dynamic Young's modulus ( E_{dyn}) (Pa):**
(3)
where ρ is the density in kg m

^{−3},

*V*

_{p}is the compressional sonic velocity in m s

^{−1}and

*V*

_{s}is the shear sonic velocity in m s

^{−1}.

**Sandstone static Poisson's ratio (ν _{stat}) (no units):**
(4)
where

*v*

_{dyn}is the dynamic Poisson's ratio.

**Siltstone static Poisson's ratio (ν _{stat}) (no units):**
(5)
where

*v*

_{dyn}is the dynamic Poisson's ratio.

**Sandstone static Young's modulus ( E_{stat}) (Pa):**
(6)
where

*E*

_{dyn}is the dynamic Young's modulus.

**Siltstone static Young's modulus ( E_{stat}) (Pa):**
(7)
where

*E*

_{dyn}is the dynamic Young's modulus.

**Strain-derived S_{Hmax} (Pa):**
(8)

**Strain-derived S_{hmin} (Pa):**
(9)
where σ

_{Hmax}is the maximum effective horizontal stress, σ

_{hmin}is the minimum effective horizontal stress,

*v*

_{stat}is the static Poisson's ratio,

*S*

_{v}is the vertical stress, α is Biot's coefficient,

*P*

_{p}is the formation pressure,

*E*

_{stat}is the static Young's modulus, ε

_{max}is the strain in the maximum horizontal stress direction and ε

_{min}is the strain in the minimum horizontal stress direction.

The initial calibration of the stress profiles was undertaken using the stress polygon method (Moos & Zoback 1990; Zoback 2007). This method uses the incidence of borehole failure (breakout and drilling-induced tensile failure) to estimate the stress conditions required to induce failure within the rock, using the frictional limit (defined by the friction angle) and compressional and tensional strength of the rock at the point of failure occurrence. The friction angle was calculated using the method defined by Lal (1999; equation 10). The unconfined compressive rock strength was defined based on empirical relationships listed in Chang *et al.* (2006). The equation defined by McNally (1987), based on data from the Bowen basin, was used for sandstones (equation 8). Where available, fracture closure pressures derived from pressure data such as leak-off tests (LOTs), diagnostic fracture injection tests and modular formation dynamic tester minifracs and pre-injection minifrac tests were used to constrain the minimum principal stress (Barree *et al.* 2007, 2009). Table 1 gives the wells and the specific calibration method(s) used.

**Friction angle (°) (Lal 1999):**
(10)
where *V*_{p} is compressional sonic velocity in m s^{−1}.

**Sandstone compressive rock strength (MPa) (McNally 1987):**
(11)
where Δ*t* is the compressional sonic slowness in μs/ft and *e* is the base of natural logarithm.

## Stress orientation

*In situ* stress data covering an area of 350×210 km (>70 000 km^{2}; for comparison, an area more than half the size of England) show significant variation in the orientation of *S*_{Hmax}. Based on the results, six distinct domains can be identified: the north Bowen, south Bowen and Burunga Anticline regions in the Bowen basin and the Roma Shelf and Taroom Trough regions in the Surat basin (Fig. 3).

Regional maps of the *S*_{Hmax} orientation and *S*_{hmin} magnitude have been constructed using all the available data from open file wells to December 2015. The *S*_{Hmax} orientation has been determined from observations of breakout or drilling-induced fractures on image logs or from breakout observed on four- or six-arm caliper logs. The *S*_{Hmax} orientation represented on the map in Figure 3 is derived from breakout data. The DITF data give the same result, but there are fewer data points and they are not discussed further herein. The mean *S*_{Hmax} azimuth is represented by the straight lines given in Figure 3. To avoid ambiguity, all data presented are from depths >450 m and from wells with <30° wellbore inclination.

This study builds on the findings of Brooke-Barnett *et al.* (2015) by applying the statistical methodology outlined by Hillis & Reynolds (2000, 2003) over both the Surat and Bowen basins. Consequently, the Rayleigh test was applied to the stress orientation data to determine the confidence of stress orientations over the study area (Mardia 1972; Table 2). Wells were also grouped into regions based on the underlying SEEBASE topography and *S*_{Hmax} orientation (Fig. 3) and the Rayleigh test was applied separately to each of these regions. The regions were then classified into six types using the following criteria: a type 1 region can reject the null hypothesis that stress orientations are random at the 99.9% confidence interval; a type 2 region can reject the null hypothesis at the 99% confidence interval; a type 3 region can reject the null hypothesis at the 97.5% interval; a type 4 region can reject the null hypothesis at the 95% interval; a type 5 can reject the null hypothesis at the 90% interval; and a type 6 region suggests that the null hypothesis cannot be rejected at the 90% interval (Hillis & Reynolds 2000, 2003). As per the methodology of Hillis & Reynolds (2000, 2003), Table 2 shows the results of the Rayleigh test as applied to the mean *S*_{Hmax} orientations from A to C quality borehole breakouts and DITF measurements. However, the mean statistics were also calculated using all borehole breakouts and DITF measurements (A to E quality) as well as the average *S*_{Hmax} for each well as per the methodology of Brooke-Barnett *et al.* (2015). Over the entire study area, the mean *S*_{Hmax} orientation is *c.* 42° N with a standard deviation of *c.* 36° (Table 2), giving the area a type 1 stress ranking. However, the orientation and quality of the *in situ* stresses varies significantly between the six domains that make up the whole area.

The northern Bowen domain is dominated by a NNE-trending

*S*_{Hmax}orientation; the dominant S_{Hmax}orientation here is*c.*22° N with a standard deviation of*c.*19° (Table 2). There are individual diversions from the dominant orientation in the very west of the study area; in the far north individual easterly trending as well as one southeasterly trending outlier are recorded. This region is designated as a type 2 stress region.The south-central Bowen domain of the Bowen basin forms a transitional corridor of variable

*S*_{Hmax}orientations; both NE and NW trends as well as easterly trends occur, which gradually changes to an east–west to WNW–ESE trend of*S*_{Hmax}further south. Overall, this region exhibits a mean*S*_{Hmax}orientation of*c.*70° N with a standard deviation of*c.*34° (Table 2). Despite the higher spread in orientation in this region, the sheer amount of reliable stress indicators (21 A–C type measurements) enable this region to have a type 1 stress ranking.The southern Bowen domain (immediately north of the Roma Shelf on Fig. 3) exhibits a very consistent east–west

*S*_{Hmax}orientation with a standard deviation of*c.*4° (Table 2). Similar to the north Bowen domain, the consistency of orientation between wells, despite the small sample size, means this domain also has a type 2 ranking.The Roma Shelf domain, in which most stress measurements are from wells in the Jurassic Walloons sequence, is again dominated by NNE- (

*c.*18°) trending*S*_{Hmax}orientations. Note that although the orientations between wells are relatively consistent (standard deviation of*c.*14°) the quality of stress indicators in this region is low, meaning there are insufficient data to assign a ranking to this area (Table 2).The Burunga Anticline on the eastern margin of the Taroom Trough displays a consistent NE trend of

*S*_{Hmax}(*c.*51°) with a standard deviation of*c.*14° (Table 2). The low standard deviation and high quality of stress indicators give this region a type 1 stress ranking.The Taroom Trough domain of the Surat basin (which overlies the depocentre of the Permian Bowen basin; Figs 2 & 3) to the east of the Roma Shelf shows a complex

*S*_{Hmax}orientation. In the basin centre the stress azimuths are dominated by broadly easterly trends (both ENE and ESE), but show a number of significant deviations from a clear overall trend. At the flanks of the depocentre the*S*_{Hmax}orientation swings into parallelism with the depocentre boundaries (a fault system in the east and a ramp in west, Brooke-Barnett*et al.*2015). Other datasets (Brooke-Barnett*et al.*2015) show a NE trend to the east, where the Walloons depocentre is underlain by a basement high (New England Orogen Region; Table 2). Despite the inclusion of additional data, this region retains the type 6 designation from Brooke-Barnett*et al.*(2015).

## One-dimensional wellbore stress models (mechanical Earth models)

A one-dimensional wellbore stress model (also called a mechanical Earth model) is a numerical representation of the geomechanical state of the subsurface over a given interval and combines known pore pressures, the stress state (vertical, minimum and maximum horizontal) and rock mechanical properties (uniaxial compressive strength, Young's modulus and Poisson's ratio). The one-dimensional wellbore stress models herein have been generated using RokDoc623 software.

The one-dimensional wellbore stress models have been compiled for four example wells within the Bowen basin (circled well locations, Fig. 3). The one-dimensional wellbore stress models are created using shear and p-wave sonic velocities and elastic models to produce estimates of stress and rock properties. The strainless one-dimensional wellbore stress models are validated and refined with known data, including leak-off data, DFIT or mini-frac closure pressure data, which gives an estimate of the minimum horizontal stress at a particular depth. Alternatively, laboratory-based rock strength testing or frictional limits based on stress indicators from image log analyses were used to calibrate the horizontal stress magnitudes (for method applied, see figure captions). One-dimensional wellbore stress models show systematic variations of Andersonian stress geometry with depth in the Bowen and Surat basins (see Flottmann *et al.* 2013; Brooke-Barnett *et al.* 2015).

At depths shallower than around 500–600 m TVD (total vertical depth, i.e. the depth below the surface measured from the drill rig floor), the stress state is characterized by a reverse stress regime (*S*_{V}<*S*_{hmin}<*S*_{Hmax}, where *S*_{V}=vertical stress, *S*_{hmin}=minimum horizontal stress and *S*_{Hmax}=maximum horizontal stress). Image log data in reverse stress regimes are dominated by borehole breakouts (DITFs are largely absent) and breakout orientations indicate significant variability in the azimuth of *S*_{Hmax} (Fig. 4). The one-dimensional wellbore stress models presented in Figures 5, 6, 7, 8 indicate low horizontal differential stresses (i.e. the difference between *S*_{Hmax} and *S*_{hmin}), in particular at depths where reverse stress regimes are dominant. The low differential stress is a likely cause of the scatter in the breakout orientations at shallow depths. Below 500–600 m TVD the stress geometry is typically of a strike-slip stress regime (*S*_{hmin}<*S*_{V}<*S*_{Hmax}). Image log data in strike-slip stress regimes show both borehole breakouts and DITFs (Fig. 4a), both of which occur dominantly in shale/siltstone units.

Figure 4b and c show the wellbore stress conditions at 400 and 800 m, respectively, using the stress and rock strength data given in Figure 6 for those depths. The stress/rock strength conditions at 400 m allow for broad compressive failure where the maximum horizontal stress exceeds the compressive failure. This condition is represented by the occurrence of scattered borehole breakouts. The stress conditions do not reach tensile failure, resulting in the sparse development of DITFs at this depth (Fig. 4a). At 800 m depth both the differential stresses and the uniaxial compressive strength are higher than at 400 m (Fig. 6). This results in the preferential development of DITFs (as the stress conditions exceed the tensile rock strength). Both DITFs and borehole breakouts occur in a well-defined narrow band at this depth. The data compilation in Figure 4a shows the dominance of DITFs at depths >1000 m; the dominance of DITFs at greater depths appears to be related to increasing differential stress with depth in this part of the Bowen basin.

The transition between reverse stress regimes and strike-slip stress regimes is also well documented in tiltmeter data acquired during hydraulic stimulations in the Jurassic Surat basin (Flottmann *et al.* 2013). The dataset presented here shows a similar stress regime transition at the same depth range, but in the Permian Bowen basin. The co-occurrence of the transition from a reverse to a strike-slip stress state at a similar depth range in two different basins suggests that the transition in stress state is controlled by the present day depth rather being controlled by geological or stratigraphic parameters.

A second transition from a strike-slip to a normal stress regime (*S*_{hmin}<*S*_{Hmax}<*S*_{V}) has been documented at *c.* 650–800 m in some areas in the Surat basin. This transition does not occur in the Bowen basin. The magnitude of differential stresses in the Surat basin also show significant variability. This appears to be attributed to both the nature of the basement and/or the thickness of the sedimentary section underlying the Surat basin (Brooke-Barnett *et al.* 2015).

## Mechanical stratigraphy

Rock properties (Young's modulus, the uniaxial compressive strength and Poisson's ratio) are constrained by rock strength testing from offset wells in both the interburden and the coals. *S*_{hmin} has been constrained using DFIT, minifrac and leak-off data. Results from one-dimensional wellbore stress models display some key contrasts in mechanical stratigraphy. In principle, (non-coal) interburden rocks have a higher rock strength than coals, which display a consistently low rock strength based on a high Poisson's ratio and low Young's modulus. Coals are dominated by a normal stress regime, regardless of whether the surrounding rocks are in a reverse, strike-slip or normal overall stress regime. Importantly, in reverse and strike-slip stress regimes coals exhibit generally lower overall stresses than the surrounding country rock (Fig. 9a, b). This has been established by numerous systematic DFIT tests in numerous wells in both the Bowen and Surat basins (Fig. 9a, b; Flottmann *et al.* 2013). The same result is achieved by establishing frictional limits theory based on image log analyses.

The relationships given in equations (8) and (9) suggest that rocks with a high Poisson's ratio (e.g. coals) are more susceptible to accommodating high stress in tectonic scenarios dominated by vertical ‘loading’ (overburden); conversely, horizontal ‘loading’ (i.e. the tectonic component) is dominantly accommodated in rocks with a high Young's modulus. Based on the one-dimensional wellbore stress models, the essential elements impacting the mechanical stratigraphy in the Bowen and Surat basins can be reduced to three key components: (1) coals with a comparatively high Poisson's ratio and low Young's modulus; (2) interburden rocks with a comparatively low Poisson's ratio and a high Young's modulus; and (3) a stress regime dominated by horizontal (tectonic) components.

In the following discussion we explore the oilfield impacts of the key elements of the mechanical stratigraphy in the Bowen and Surat basins, which is relevant to basins with similar conditions worldwide.

## Discussion

The data presented in this paper show variability in the plan view stress azimuths and Andersonian stress geometry with depth. The data and interpretations from the Permian Bowen basin presented here, in combination with similar data from the Jurassic Surat basin (Brooke-Barnett *et al.* 2015), show similar depths for the transition from a reverse to a strike-slip regime (400–600 m). This suggests that the first-order influence for the occurrence of this transition in stress regime is depth (TVD) rather than local conditions, such as the basin-specific stratigraphy. The transition from a reverse to a strike-slip stress state at *c.* 400–600 m depth thus appears to be typical for eastern Queensland. This is in contrast with interior parts of Australia (e.g. the Cooper basin), where a transition from a strike-slip stress regime at shallow depths to a reverse stress regime is reported to occur at depths in excess of *c.* 2.5 km (e.g. Reynolds *et al.* 2006).

Previous datasets presented in the World Stress Map (Heidbach *et al.* 2008, 2010) suggest an overall NNE–NE orientation of *S*_{Hmax} in the Bowen basin. Our data support this observation over the entire region, including both the Bowen and Surat basins (Table 2). However, our data show significant local deviations from the inferred regional trend. In particular, the south Bowen domain of the Bowen basin and the Taroom Trough domain of the Surat basin both display variable *S*_{Hmax} orientations and the Denison Trough domain displays an east–west *S*_{Hmax} which, although consistent, deviates significantly from the regional *S*_{Hmax} orientation.

The reasons why the orientation of *S*_{Hmax} azimuths is subject to significant local variations remains, at this stage, a matter of speculation. Brooke-Barnett *et al.* (2015) showed that *S*_{Hmax} azimuths in the Surat basin are significantly influenced by the underlying basement structures. However, available SEEBASE data show no obvious major structural trends that could guide stress rotations of the severity seen in the southern Bowen basin. The easterly stress orientations in the south-central and southern domains may be guided by deep-seated easterly trending fault systems and volcanism related to the opening of the Coral Sea during the Tertiary (Cook & Jell 2013). The changing composition of the deeper basement could result in changes in the bulk rock strength, which, in turn, could contribute to the stress rotations observed – in fact, the stress rotations may provide a guide for future remote deep data acquisition. Regardless, the observations presented here suggest that stresses in intracontinental basins can vary significantly and without any visible first-order manifestation in complementary datasets, such as faults or basin structure.

However, from an oilfield perspective, it is important to take local variations of *S*_{Hmax} into consideration. For horizontal and/or deviated wells, for example, the *S*_{Hmax} azimuths can have significant implications with regard to optimizing wellbore stability in deviated/horizontal wells in the context of rock strength and mud-weight optimization to prevent wellbore collapse. Wellbore stability is highest where the differential stresses around wellbores are at a minimum. This is particularly true in deviated/horizontal wells (regardless of the overall stress magnitudes). In many strike-slip regimes, which are common around the world, this often entails deviating the wellbore into *S*_{Hmax} as the differential stresses between *S*_{hmin} and *S*_{V} are less than, for example, the differential stress between *S*_{V} and *S*_{Hmax}.

In the context of fracture stimulation completions, a well deviated into *S*_{Hmax} may not be optimum because hydraulic fractures propagate preferentially into the azimuth of *S*_{Hmax}, resulting in a longitudinal fracture stimulation. Generally, transverse stimulations (i.e. the fracture propagates perpendicular to the wellbore) are more advantageous, particularly in low-permeability formations, because they access greater reservoir rock volumes. Consequently, the wellbore has to be deviated into *S*_{hmin} and, during drilling, the mud-weight window has to be adjusted to deliver a stable wellbore under high differential stresses acting perpendicular to the wellbore axis. Regionally varying stress azimuths as documented here highlight the need to acquire a complete stress dataset, allowing full stress tensor and stress azimuth descriptions to optimize both drilling and completion options. The local variability of the *S*_{Hmax} azimuths in our dataset demonstrates that it may be insufficient to rely on regional datasets alone; local conditions have to be established to allow for optimum drilling and completion designs.

Stress state transitions with depth, and between lithologies, are of primary practical importance. Away from the wellbore-influenced hoop stresses, hydraulic fractures open against the regional minimum principal stress. Consequently, hydraulic fractures are expected to be vertical in both strike-slip and normal stress regimes; however, in a reverse stress regime fracture opening is expected to be horizontal and leads to a horizontal hydraulic fracture. The occurrence of horizontal fracturing at shallow depth is corroborated by tiltmeter data in the Surat basin, which show a predominance of horizontal components (up to 100%) at shallow depths. At depths >400–500 m (Flottmann *et al.* 2013), tiltmeter data show a predominance of vertical fracture components.

Three-dimensional stress characterization is a key requirement for the planning and implementation of drilling and completions such as hydraulic fracture stimulations. Hydraulic stimulations propagate perpendicular to the lowest principal stress in the plane defined by the intermediate and maximum principal stresses; hydraulic fractures tend to grow in the azimuth of *S*_{Hmax}. The mechanical stratigraphy and stress magnitudes are of particular importance for the vertical containment of fracture growth. Fractures initiate and grow preferentially in formations with low *S*_{hmin}. Formations with low *S*_{hmin} have a low mean stress [(*S*_{V}+*S*_{Hmax}+*S*_{hmin})/3]; conversely, formations with a high *S*_{hmin} and a high mean stress tend to act as barriers to the vertical growth of hydraulic stimulations, thereby containing the stimulation to the target zone.

However, different Andersonian stress regimes result in contrasting stress accommodation and stress intensity in rocks with different rheological properties. Lithologies with a high Poisson's ratio and a low Young's modulus (such as coals) tend to accommodate high stresses in a normal stress regime (*S*_{V}>*S*_{Hmax}>*S*_{hmin}), where the maximum principal stress is vertical and the vertical component of stress is governed by Poisson's ratio (equations 8 & 9; see also Herwanger *et al.* 2015). Conversely, in tectonic regimes where the maximum principal stress is horizontal (reverse and strike-slip tectonic regimes), the stress distribution is governed by rocks with a high Young's modulus and a low Poisson's ratio (i.e. the tectonic component in equations 8 & 9). Consequently, in reverse and strike-slip regimes the interburden rocks (high Young's modulus and low Poisson's ratio) will accommodate high stresses and the coals will be comparatively less stressed.

The interaction of the Andersonian stress state and rocks of contrasting rheological properties thus has significant implications for the propagation of hydraulic fractures. In a normal stress regime, for example, where coals are comparatively highly stressed, hydraulic fractures will initiate in coals, but will tend to grow into (interburden) formations dominated by a lower mean stress (Fig. 10a). Our interpretation of one-dimensional wellbore stress models shows that, in the Bowen basin, the actual Andersonian stress states are reverse and strike slip and the coals tend to be the least stressed members of the mechanical stratigraphy. Interburden rocks with a high Young's modulus accommodate the horizontal tectonic stress component (equations 8 & 9) and are the most highly stressed components of the mechanical stratigraphy in the Bowen basin (Fig. 10b). Consequently, hydraulic fracture completions targeting coals are well contained in the low stress coals (Fig. 10b). Similar observations are documented in the Surat basin, where tracer logs show containment in coal during hydraulic treatments (Kirk-Burnnand *et al.* 2015).

Different Andersonian stress regimes can result in potentially stark contrasts in rock-specific stress states (Fig. 10a, b). This highlights the need for the full three-dimensional characterization of stress parameters to achieve the appropriate conditioning of one-dimensional geomechanical models, which are key (software) inputs for planning fracture stimulations or well planning. The three-dimensional variation of stress states in the Bowen basin highlight the interdependency between the plate tectonics boundary conditions, the local structural geological setting and rheological parameters, all of which ultimately contribute to the geomechanical conditions at individual wellbores.

## Conclusions

This paper documents broad, but systematic, changes in the *in situ* stress orientations and *in situ* stress state with depth in the intracontinental Bowen and Surat basins of eastern Queensland, Australia. Both basins show a transition from a reverse stress regime at depths shallower than 400–600 m; at greater depths strike-slip stress geometries are dominant. The (Andersonian) stress regime transition is primarily depth-controlled, independent of the basin and/or lithology. The spatial variability of the *in situ* stress distribution results from the interaction of regional intra-plate stresses with basin-scale structures and basement rheology. Stress geometries and rock properties materially influence drilling and completion considerations. In particular, hydraulic fracture completions have to be designed to take into account the rock-specific geomechanical conditions established from log-derived one-dimensional wellbore stress models. The data and analyses presented suggest that first-order regional stresses increase in complexity at a local scale; ultimately, the resolved geomechanical state at the wellbore level reflects a scale-dependent interplay of plate tectonic forces that are geologically modulated by the local structure and the mechanical properties of individual rock packages.

## Acknowledgments

We thank Santos Ltd and Origin Energy for permission to publish this paper. We acknowledge the efforts of countless colleagues and field personnel who contributed to data acquisition and discussions leading to the results presented here. Particular thanks to the convenors of the GSL conference Geology of Geomechanics for the opportunity to present and encouragement to publish. Tony Addis and Joe English prepared thorough and insightful reviews, which greatly improved the manuscript.

- © 2017 The Author(s). Published by The Geological Society of London. All rights reserved