Image analysis of acoustic data and interpretation of rock stress orientations for geothermal exploration in Gothenburg borehole GE-1, SW Sweden

Supplementary material: Raw data tables from analyses (drilling disturbances, stress indicators, pre-existing structures) are available at https://doi.org/10.6084/m9.figshare.c.7082902

Geothermal energy production is sometimes regarded as the ‘sleeping giant’ of energy. So far, only a tiny proportion of the geothermal potential has been realized, and mostly in areas with favourable geological conditions. Developments in rapid pneumatic percussion drilling techniques and engineered geothermal systems (EGS) have opened up the potential for targeting less favourable regions and rock formations for geothermal exploration, including Scandinavia. The energy company St1 initiated the game-changing pilot project in Otaniemi, Finland, that aimed to feed direct heat from 6 to 7 km deep boreholes to a local district heating network. While the project successfully demonstrated deep and fast drilling techniques and controlled fluid injection for geothermal stimulation to over 6 km depth (e.g. Kwiatek et al. 2019), post-stimulation problems in maintaining hydraulic conductivity of the EGS reservoir have put a halt to the project and the wells have been released to research (St1 2023).

Inspired by the early progress of the Otaniemi project, and as part of the ambition to deliver 100% recycled and renewable energy by year 2025, the municipal energy company Göteborg Energi AB initiated a pre-study followed by a drilling project to assess the geothermal potential in Gothenburg, Sweden. The drilling targets were steeply dipping and permeable fracture zones within bedrock composed of radioactive (RA) granites. If the radioactive gamma-ray decay from heat producing elements (K, U, Th) elevates the temperature and thermal gradient in the rock mass, this would result in the target temperature for direct heat production (∼120°C) being reached at a shallower depth than in the surrounding rock mass, and thus incur lower drilling costs. The first borehole, GE-1, is a 1 km deep, vertical and fully cored borehole completed in April 2021, which subsequently was logged with high-resolution acoustic televiewer (HiRAT), natural gamma-ray, resistivity and temperature tools in June 2021. The main aims of the drilling project were to: (1) measure in-situ temperatures and derive the geothermal gradient, (2) measure thermal properties of the bedrock and (3) constrain hydrogeological and mechanical properties of the fracture zone. This study addresses the third aim of the drilling project and presents data on downhole in-situ horizontal stress orientations, derived from HiRAT data.

The state of in-situ stress in Earth's crust influences bedrock stability and fluid flow patterns (e.g. Heffer and Lean 1993; Barton et al. 1995; Hickman et al. 1997; Ito and Zoback 2000; Townend and Zoback 2000; Rogers 2003). Knowledge of the in-situ state of stress, especially at greater depths, is critical for a range of processes and engineering applications (e.g. Engelder 1993; Amadei and Stephansson 1997; Zoback 2007; Zang and Stephansson 2010). For geothermal exploration, detailed information on the state of stress is needed to optimize the design of the underground installations to maximize fluid flow and minimize risks of borehole instability and surface disturbances from drilling, EGS stimulations and long-term operation of production and injection wells. Rock stress measurements are challenging to conduct because stress is a fictitious term and cannot be directly measured, albeit a practical term in engineering as it defines the dimensioning load of the EGS reservoir and its downhole installations. This study is a first step towards the establishment of a rock stress model for a future EGS facility. The final rock stress model should be based on integrated results from more than one stress measurement method (Stephansson and Zang 2012). The primary objectives of this paper are to constrain the orientation of horizontal stresses from borehole stress indicators and to discuss implications for geothermal exploration.

HiRAT data allow mapping of fracture occurrence and their geometry, as well as investigating of passive stress indicators, borehole breakout (BO) and drilling-induced fractures (DIFs) that form as the result of stress-induced failure in the borehole wall. For a vertical borehole in an isotropic and intact rock formation, drilled parallel with a principal (vertical) stress, BOs and DIFs continuously reveal the orientation of minimum and maximum horizontal stress, respectively. It is generally agreed that high-resolution borehole imaging (e.g. HiRAT) produces high quality assessments of in-situ stress orientation, especially in comparison with four-arm caliper tools. While high-resolution images indeed allow analysis of smaller details in the wellbore, interpretation of in-situ stress involves subjectivity, especially for metamorphic crystalline formations, that is not always straightforward. The secondary objective of this study is to test different strategies for interactive stress analysis through visual inspection of acoustic images. These results will highlight the need to develop better guidelines for data interpretation. We have conducted three types of image analyses, following Ask and Ask (2018) and Pierdominici et al. (2020).

In the following, we first present the geological setting of the region and brief results from the GE-1 borehole. Second, we present methods for stress analysis, including theories behind the stress indicators and how data collection and analyses were conducted. This is followed by the presentation of results of the present-day stress field and mapping of pre-existing structures. The final sections consist of discussions and conclusions. Tables with raw data from analyses are included in the Supplementary Material.

The bedrock in southwestern Sweden forms part of the 1150–900 Ma Sveconorwegian Province, the remnant of a large hot orogen in southwestern Fennoscandia formed during the assembly of the supercontinent Rodinia (Bingen et al. 2021). It is composed of lithotectonic segments that are separated by roughly north–south trending ductile deformation zones. The Gothenburg region is located in the easternmost allochthonous lithotectonic segment of the Sveconorwegian Province, the Idefjorden Terrane, which is directly overlying the parautochthonous Eastern Segment, along a major west-dipping deformation zone system, the Mylonite Zone (Andersson et al. 2002; Viola et al. 2011; Möller and Andersson 2018). Based on differences in protolith ages and metamorphic overprint, the Idefjorden Terrane is further subdivided in two lithotectonic subunits termed the Median and Western Segments (Fig. 1a; Berthelsen 1980; Bergström et al. 2020). In the Gothenburg region, these sub-segments are separated by an anastomosing deformation zone system, the Göta Älv Shear Zone (Park et al. 1991) and the bedrock in the Gothenburg region includes both the Median and Western Segments. This bedrock was mainly formed and partly reworked during the 1660–1520 Ma Gothian accretionary orogeny. The Gothian bedrock is dominated by calc-alkaline granitoids with minor volcanic deposits, the Western Segment includes extensive metasedimentary belts and late Sveconorwegian c. 920 Ma granite batholiths (Bergström et al. 2020). The Gothian–Sveconorwegian interlude included a period of crustal extension resulting in substantial 1340–1300 Ma bimodal magmatism in the Gothenburg area (the Kungsbacka suite), including emplacement of the 1336 ± 10 Ma Askim and 1311 ± 8 Ma Kärra Granites (Hegardt et al. 2007). These intrusions were later partly deformed at upper amphibolite conditions during the Sveconorwegian orogeny, including development of gneissic foliations, partial melting and crustal scale ductile deformation in the Göta Älv Shear Zone system (Bergström et al. 2020).

The Askim Granite is greyish red to reddish grey in colour, undeformed to foliated in discrete ductile deformation zones, and typically porphyric with c. 1 cm large, elongated K-feldspar augens (Hegardt et al. 2007). In contrast, the Kärra Granite is penetratively gneissic and veined, and has a high red K-feldspar content that results in a greyish red to red colour of the rock. The Kärra Granite has a conspicuously high U and Th content and a gamma activity of up to 140 μR/h, the Askim Granite has a lower gamma activity of ∼20 μR/h. The Kärra Granite, together with granitic bodies to the north, has in the literature been referred to as radioactive granites (RA granites; Samuelsson 1985). The extent and radioactive composition of the Kärra and Askim Granites, however, remain to be investigated in detail.

The GE-1 borehole is located near the SE corner of an old quarry in Högsbo, southern Gothenburg (Fig. 1b). The near-vertical borehole was fully cored in NQ size (∼76 mm) to a total depth (T.D.) of 1 km, using water as drilling fluid (no drilling mud). Casing was installed to 18 m depth. Table 1 shows the coordinates and elevation of borehole GE-1 (Axel Sjöqvist, pers. comm. 27 January 2023). The lithostratigraphic column (Hogmalm et al. 2021) and natural gamma-ray (NGR) logging results are shown in Figure 2.

Granitic gneiss is mapped from 0 to 70 m depth, and tonalite gneiss is mapped from 70 to 173 m. The average thorium (Th) content to 173 m is 2.20 ppm and the average uranium (U) content is 0.95 ppm (Hogmalm et al. 2021). Remaining cores to T.D. are interpreted to mainly be composed of RA granites. Pointwise measurements with a handheld gamma-ray spectrometer suggest average Th and U values of 10.4 ppm and 2.95 ppm, respectively (Hogmalm et al. 2021). This means that the radioactive content in the RA granites is about five times greater than in the altered diabase sills that were mapped from 400 to 420 m, 580–600 m and 900–910 m, and these sections are also associated with higher fracture frequency. In addition, pegmatite was occasionally observed in the core (Fig. 2).

Over 700 fractures were observed in the cores, 49% of these follow the foliation and are dipping towards the west, the remaining 51% of fractures cut across the foliation and the majority dips towards the east (Hogmalm et al. 2021). The vast majority of fractures (77%) are open fractures, but, at least some of them could have been opened during drilling. Only 22% of observed fractures in the cores are completely sealed. The fracture frequency derived from core measurements suggests a generally low fracture frequency (<5 fractures m⁻¹). Less than 25% of the analysed cores had a fracture frequency over 5 fractures m⁻¹. The highest fracture frequency (>15 fractures m⁻¹) is observed in only eight cores (40 m in length) collected from depth intervals within 100–170 m, and near 400, 900 and 985 m depths, respectively (Hogmalm et al. 2021).

Stress-induced failure in the borehole wall will occur in sections of the borehole circumference where the σ_θθ magnitude exceeds the rock strength. For vertical boreholes, drilled parallel with a principal (vertical) stress, this generates continuous information on the orientation of the local stress field, in contrast to most, if not all other stress measurement methods that provide pointwise measures.

BOs represent compressional failure and are initiated on opposite sides of the borehole wall in sections where σ_θθ exceeds the compressional rock strength. BOs grow by propagation of intersecting conjugated shear fractures into the rock in the direction parallel with S_hmin (e.g. Bell and Gough 1979; Zoback et al. 1985, 2003). The borehole wall is usually rough inside and smooth outside of a BO, resulting in an enlarged borehole diameter within a BO, while the diameter remains close to bit size outside of a BO. In general, BOs vary in length from less than a metre to several tens of metres (e.g. Shamir and Zoback 1992).

DIFs represent tensional failure that forms when σ_θθ exceeds the tensional rock strength. Equation (2) shows that high mud and pore pressures and thermal stresses favour the formation of DIFs. For boreholes drilled along a principal stress direction, axial DIFs are formed on opposite sides of the borehole wall along the borehole axis, parallel with the orientation of S_Hmax (Brudy and Zoback 1993, 1999; Aadnoy and Bell 1998). Axial DIFs are narrow features, generally ranging from 0.1 to 2 m in length (Brudy and Zoback 1999). These types of axial tensile fractures also form the base of hydraulic fracturing theory (e.g. Haimson and Cornet 2003).

Petal centreline fractures (PCFs) represent a less well-known type of drilling-induced fracture that develop some metres below the drill bit, and subsequently become crossed by the borehole (e.g. Kulander et al. 1990; Davatzes and Hickman 2010; Hehn et al. 2016). As a result, PCFs are present in both cores and in the borehole wall. Studies on oriented drill cores have shown that they occur parallel with S_Hmax (Kulander et al. 1990; Li and Schmitt 1997, 1998). Supporting results are provided by observations in image logging data of consistent stress orientations from BOs, DIFs and PCFs (e.g. Davatzes and Hickman 2010; Hehn et al. 2016). PCFs tend to develop in normal and strike-slip regimes, and their formation is supported by high mud and pore pressure, and a rough shape of the bottom hole (Li and Schmitt 1998).

The HiRAT tool consists of a piezo-electric acoustic transducer which emits ultrasonic pulses that are reflected at the borehole wall and received by the same transducer (e.g. Zemanek et al. 1969; Pierdominici and Kück 2021). The tool collects two types of data:

• Acoustic two-way travel time (TT), the time between transmission of the sonic pulse and reception of the reflected pulse at the transducer. The TT image yields information on the borehole shape (acoustic caliper).

• Acoustic amplitude (AMPL), the acoustic reflectivity of the borehole wall resulting from the acoustic impedance contrast between borehole fluid and wall. The AMPL image depends on the roughness and shape of the borehole wall and its acoustic properties.

The tool must be centralized in the borehole, and the acoustic contrast between drilling fluid and borehole wall should be sufficient. Artefacts from drilling, logging and processing may disturb the logging quality (e.g. Lofts and Bourke 1999; Lofi et al. 2018; Pierdominici et al. 2020). The effects from some artefacts can be mitigated during processing of the data, other artefacts cannot be remedied.

The GE-1 borehole was logged by Lund University on 15 June 2021 with a Robertson Geo High Resolution Acoustic Televiewer (Serial No. HiRAT 11883). Data acquisition was conducted with a logging speed of 2.5 m min⁻¹ (41.7 mm s⁻¹), which combined with a rotation rate of 20 revs/s produces 360 pulses/rev. This results in an axial resolution of 2.0 mm and a 1° radial resolution that corresponds to 0.7 mm for a 76 mm diameter borehole. The GE-1 borehole was logged from 3 to 1001 m depth. The GE-1 borehole was drilled nearly vertical, which implies that orientation downhole data are based on magnetometer sensors relative to magnetic north (°MN). The lack of input from accelerometer sensors implies somewhat reduced precision of the orientation of mapped fractures and stress indicators.

The analyses of stress indicators and pre-existing structures were conducted with respect to magnetic north (°MN). The orientation of true north (TN) is obtained by adding the magnetic declination (D) to the value of magnetic north (TN = MN + D). The magnetic declination is the angle between TN and the horizontal trace of the local magnetic field. Because the magnetic north is not stationary, the angle of the magnetic declination also varies at a site over time. Input of logging date, borehole coordinates and elevation (Table 1) in model WMM-2020 results in a magnetic declination of 4.23° and an uncertainty of 0.43° (NOAA 2023). We obtain values of S_Hmax (°N) by adding 4° to S_Hmax (°MN).

We used the log-analysing program WellCAD by Advanced Logging Technology (www.alt.lu). In general, the image logs show very good quality. Data handling and processing of all data encompassed the following steps, described in Advanced Logging Technology handbooks. (1) Bad traces are removed. (2) The data are oriented with respect to MN. (3) Two borehole (tool) deviation parameters are calculated: azimuth and tilt (inclination of the borehole with respect to the horizontal). (4) Caliper values are calculated. (5) The image data are centralized. (6) A 3D plot is generated from the TT and AMPL logs. (7) The image is filtered with an average filter and a filter window of three points for width and height. (8) A static normalization mode is applied. WellCAD allows interactive picking of pre-existing structures and stress indicators (BOs, DIFs, PCFs) in structure and breakout logs, respectively. These logs are overlaid on the AMPL and TT image logs, allowing interpreters to pick structures or stress indicators interactively. In WellCAD, the term ‘azimuth’ is used in both structure and breakout logs. However, in the structure log, azimuth refers to the dip direction, whereas in the breakout log, it denotes the strike of the stress indicator. To enhance clarity, we use ‘azimuth’ when discussing stress indicators, and ‘dip direction’ when discussing pre-existing structures. WellCAD has a more stringent data format for the dip of pre-existing structures (i.e. the angle between the horizontal plane and the structure plane) and the tilt of the stress indicator (the angle between the borehole axis and the failure plane).

The interpretation of stress indicators is not always straightforward, especially for metamorphic rocks that have a long and often complex deformation history. The World Stress Map (WSM) provides widespread analysis guidelines for stress indicators in their database (Heidbach et al. 2016), including guidelines for interpreting image logs (Tingay et al. 2016). While Tingay et al. (2016) describe the general appearance of stress indicators in image logs, the guidelines do not provide particular detail on the individual BO/DIF selection, and do not include guidelines for analyses of PCFs for image logging data.

For continuous, homogeneous, isotropic, linear-elastic (CHILE) rock conditions, the maximum BO width (i.e. opening angle) shows the point where the tangential stress concentration generated by the borehole exceeds the compressive strength of the rock. On the other hand, pre-existing geological structures, metamorphic rock textures (e.g. foliation), as well as non-geological artefacts in metamorphic crystalline rock formations may influence the orientation of stress indicators, as well as asymmetric BOs and DIFs. Different strategies picking the width of BO traces include selecting their median width (e.g. Wang and Schmitt 2020) and selecting their maximum width (e.g. Wang et al. 2023).

To shed light on the impact of different approaches for stress analyses and interpreters, we have applied three methods (A–C) of analyses, presented below.

BOs are picked based on observations in TT, AMPL and cross-sectional plots following Pierdominici et al. (2020). BO data are picked on opposite sides of the borehole wall, and separated into four subgroups (Fig. 3a–c): (1) BOs with high confidence and well-defined extent in TT and AMPL logs are defined as major breakouts and assigned Q1, a good quality; (2) BOs with moderate depth, length and width extent in TT and AMPL logs are assigned Q2, moderate quality (minor BO); (3) BOs that are well-developed only one side and/or less on the other side in TT and AMPL logs are assigned Q3, partial breakouts; and (4) BOs, where individual grain-fall-outs are observed only in the AMPL log but where rocks have not spalled off from the borehole wall, are assigned Q4, initial or proto breakout.

DIFs data selection is only based on observations in the AMPL log by visually fitting a line to the estimated mean DIF orientation on opposite sides of the wellbore (Fig. 3d). There are some sections where only one single DIF trace is observed, and if the DIF orientation agreed with nearby two-trace DIF orientations, the single-trace DIF is added. Orientations of PCFs are also interpreted by fitting a line to the estimated mean PCF orientation (Fig. 3e).

The same type of analysis of BOs, DIFs and PCFs as Method A is applied, differing in the implementation of a forced separation of 180° (mirror) between BOs, DIFs and PCFs. The clearest trace of each pair of stress indicator is selected as the controlling trace during analyses.

Method C modifies a methodology used for interpretation of stress indicators and pre-existing fractures during hydraulic fracturing data analyses of metamorphic rocks by Ask and Ask (2018) using the AMPL log. Four measurements are made at each BO and DIF zone, two measurements per trace. This provides information on the quality of the interpretation. The maximum and minimum widths of each BO trace are measured, providing four values per BO. The values are then converted to maximum horizontal stress (S_Hmax) orientation within the 0–180°N range, by adding or subtracting 90°. The maximum width and the best visual median fit of each DIF trace are measured. This means that a polygon is interactively drawn to frame both the maximum and minimum widths of BO traces and the maximum width of DIF traces, and that a line is drawn to visually fit the median DIF trace.

The selected BOs and DIFs are categorized. Each stress indicator is assigned a type, to signal whether the BO or DIF has been formed immediately below, above or within a fracture zone (Type 0), or whether the stress indicator has been formed in intact rock (Type 1). Second, the data are assigned to maximum or minimum width of interpretation. Note that only the good and moderate subgroups of BOs are considered (Q1 and Q2, respectively). Furthermore, Method C does not include analyses of PCFs.

After completing the analysis of the BO and DIF data, statistics on directional data by Mardia (1972) are used to obtain a weighted mean S_Hmax orientation and standard deviation for the two stress indicators in each hole. Each individual BO, DIF and PCF zone analysed using Methods A and B is weighted by its length to enhance the dominant azimuth orientation. Data qualities of BOs and DIFs were assessed according to the WSM quality ranking system (Tingay et al. 2016), which considers the number of observations, the consistency of results and the reliability of the data as a tectonic indicator.

It is not possible to fully distinguish open natural fractures from those that are sealed (filled or closed) using HiRAT data alone, but with inspiration from Massiot et al. (2018), we classified pre-existing features in borehole GE-1. Massiot et al. (2018) conducted a two-step classification of pre-existing features, first based on their appearance (i.e. contrast in AMPL and TT logs) and then based on their morphology (i.e. continuous, discontinuous, partial, or truncated features). Due to the different scope of this study, our analyses are less detailed compared to those conducted by Massiot et al. (2018). We identified two categories of Type A pre-existing features (Fig. 4a, c): single fractures and damaged/weak zones. Type A single fractures are continuous sinusoids, with high contrast on the AMPL log and visibility on the TT log. Type A damaged/weak zones show fracture aperture (enlarged borehole diameter), continuous sinusoids and are clearly visible on both AMPL and TT logs. Type B fractures have moderate to high contrast (dark colours) on the AMPL log (intermediate colours), but are not visible on the TT log (Fig. 4a, b). We identified three categories of Type C fractures: single, partial and truncated fractures (Fig. 4a–d). Type C single fractures have low contrast on the AMPL log (bright colours) and are not visible on the TT log. Type C partial and truncated fractures may have variable contrast on the AMPL log. Foliation is interpreted from closely-spaced thin sinusoids only visible on the AMPL log (Fig. 4d), and is consequently interpreted as Type B or C, depending on how well developed foliation is (i.e. their contrast on the AMPL log).

In contrast to Massiot et al. (2018), we classify all partial and truncated fractures as Type C fractures, and we omit the detailed analysis to assess whether fractures are open or sealed. Here, we assume that all Type A fractures are open, and that all Type B and C pre-existing features are closed. These assumptions introduce potential biases in data because interpretation of only HiRAT data is inconclusive. For example, drilling may erode soft fracture filling, leading to a potential overestimation of Type A fractures (i.e. presence of open fractures). On the other hand, the classification of partial and truncated fractures may lead to a potential underestimation of Type A fractures. Furthermore, both fractures and folds were observed in the cores (Hogmalm et al. 2021; Ladefoged 2021). It is likely that our interpretation of Types B and C fractures includes both fractures and folds, especially those interpreted as partial and truncated fractures. This bias implies a probable overestimation of brittle deformation structures and underestimation of ductile deformation structures in the borehole.

The logging quality is in general excellent. We observe some sections with drilling- and logging-induced artefacts, i.e. hole spiralling and tool-sticking (Lofts and Bourke 1999). Hole spiralling is due to borehole wall morphology and prone to occur when too few or poorly fitted reamer shells are mounted between the core bit and the outer barrel of the diamond drilling system. Minor sections with tool-sticking were observed at depth in the borehole, and there were tendencies for woodgrains and decentralized tool, but these artefacts do not influence interpretations.

Spiralling is observed from 264 to 892 m depth, with seven sections of pronounced spiralling and nine sections of weak spiralling (Table S1). These 16 sections range from 0.9 to 14.7 m in length, in total 61 m of the borehole length (6% of logged length). Sections with pronounced spiralling are only observed from 273 to 320 m depth (Fig. 4c), which is within a borehole section with higher NGR counts (Fig. 2b). Borehole sections with spiralling should be avoided during packer experiments (e.g. hydraulic fracturing and hydraulic testing of pre-existing fractures) because the packers cannot fully seal hole spiralling and there is a risk of breaking the packers.

A total of 92 BOs with a cumulative length of 58.9 m are measured from 172 to 999 m depth (Fig. 5a; Table S2). Following Mardia (1972), the length-weighted mean S_Hmax orientation and standard deviation is 157 ± 7°N (153 ± 7°MN; Fig. 6a). The overall orientation of BOs ranges from 42 to 81°MN and 218–261°MN. A small number of BOs are detected near 173 m (the lithologic boundary between gneiss and RA granite), and at 283 m (Fig. 5a). The BO density increases with depth; over 15 BOs are mapped between 415 and 710 m depths, and the majority of BOs (70) are mapped from 822 to 999 m. The variation in BO orientation is slightly smaller from 283 to 920 m. The lowermost 80 m of the GE-1 borehole hosts the highest density of BOs, and they reveal a gradual and general decrease in BO orientation below 925 m depth to T.D., from ∼80–65°MN for one BO trace and ∼260–235°MN for the other BO trace. There are several open and sealed fractures, especially in the deeper section; some of these fractures offset the BO orientations.

Figure 3 shows examples of appearances of BOs in the different subgroups Q1–Q4 in the TT, AMPL, 3D log, including their resulting cross-section plots. Subgroups Q1 (good quality) and Q2 (moderate quality) each contains 12 and 10 BOs, respectively. BOs in Q1 have a cumulative length of 4.3 m and have been mapped from 480 to 996 m. Corresponding numbers for the Q2 BOs are 3.3 m and 282–997 m, respectively. Fourteen partial BOs (Q3) are mapped from 172 to 924 m, with a total length of 4.1 m. With 58 BOs and a cumulative length of 47.2 m, the proto BOs (Q4) are by far the most abundant stress indicator. Based on the BO orientation only, the four subgroups generate similar length-unweighted mean and median orientations of BOs (Table 2); however, the total range (max–min values) are smaller for Q1 BOs (30°, 27°) and Q2 BOs (23°), compared with Q3 BOs (37°, 38°) and Q4 BOs (35°, 30°). Furthermore, the difference between the two traces (0–180°, 180–360°) in mean and median values differs by up to 2–3° from the theoretical values for the Q1 and Q3 BOs (Table 2).

A total of 69 DIFs with a cumulative length of 72.4 m are mapped from 174 to 900 m depth (Fig. 5a; Table S3). In total, their length-weighted mean S_Hmax orientation and standard deviation is 158 ± 5°N (154 ± 5°MN; Fig. 6b). DIFs are only observed in the AMPL images (Fig. 3e). Apart from a DIF outlier at 173 m, the mapped DIFs are fairly evenly bimodally distributed from 259 to 900 m depth. Their orientations are either 141–176°MN or 321–351°MN. The length-unweighted mean and median orientations are similar (Table 2). The opposing DIF traces achieve the theoretical angular difference of 180° within 1°.

A total of 13 PCFs are mapped within the interval from 177 to 886 m; in general PCFs occur parallel with DIFs (Fig. 5a). In total, the length-weighted mean S_Hmax orientation and standard deviation of PCFs is 159 ± 7°N (155 ± 7°MN; Fig. 6c).

The mirror analyses generate similar bimodal distributions and almost identical BO and DIF orientations to the manual analyses (Tables S4 and S5). Differences in interpretation are in part due to the high-resolution of the acoustic images. It is possible to zoom in on the depth scale but the radial scale (angular resolution) is constant. The small discrepancies between interpretations also reflect decisions made by the interpreter.

A total of 94 mirror BOs is mapped from 172 to 999 m depth, with a cumulative BO length of 58.7 m (Fig. 5b). In total, their length-weighted mean S_Hmax orientation and standard deviation is 157 ± 6°N (153 ± 6°MN; Fig. 6d). Mapped BOs range in orientation from 42 to 83°MN and 222 to 263°MN. Table 3 summarizes length-unweighted statistics for the subgroups; the results are similar, but not identical values as for the interpretations using Method A (Table 2).

A total of 72 DIFs with a cumulative length of 72.3 m are mapped from 174 to 900 m depth (Fig. 5b). In total, the length-weighted mean S_Hmax orientation and standard deviation is 158 ± 5°N (154 ± 5°MN; Fig. 6e). Total range in orientation of mapped DIFs is 141–174°MN and 321–354°MN.

A total of 14 PCFs are mapped from 177 to 886 m with a cumulative length of 4.0 m (Fig. 5b). In total, the length-weighted mean S_Hmax orientation and standard deviation of mirror PCFs is 163 ± 11°N (159 ± 11°MN; Fig. 6f), thus a slightly higher mean S_Hmax orientation and standard deviation than for Method A.

Figure 7 shows the downhole distribution of BOs and DIFs azimuth, including how the stress indicator was picked, and if borehole failure occurred in intact rock or in a fractured interval (i.e. just above, below or within a fracture zone, see Method C in the Methods section; Tables S6 and S7).

In total, 58 maximum width BO traces and 58 minimum width BO traces are measured in fractured rocks, and 18 maximum width BO traces and 18 minimum width BO traces are measured in intact rocks from 173 to 997 m depth. The strategy to select the maximum or the minimum width has influence on the azimuth of the individual BO; we observed differences over 10° in some cases. However, no systematic pattern is observed (Table 4). The majority of BOs (76%) have been formed near or penetrate a pre-existing structure; they have a cumulative length of less than 7 m. In comparison, the cumulative length of BOs in intact rock is ∼3 m. Fracture zones near T.D. yielded higher variation in S_hmin orientations, indicating open fractures. The low overall number of observations and total length of BOs do not allow detailed observations on stress variation in fractured v. intact rock, but statistics (Table 4) suggest 6–13° higher mean and median orientation of S_hmin in intact rocks.

In total, 118 maximum width DIF traces and 118 linear fit DIF traces are measured in fractured rocks, and 44 maximum width DIF traces and 44 linear fit DIF traces are measured in intact rocks from 174 to 900 m depth. Interestingly, significant amounts of observed DIFs are formed in intact rock, 27% of all DIFs, which represent sections where the basic CHILE conditions of DIF formation are fulfilled. The cumulative length of DIFs formed in intact rock is 15 m, while it is 41 m for fractured rock. The two methods for selecting DIFs result in uniform S_Hmax orientations, with few exceptions (Fig. 7). Indeed, the mean and median S_Hmax orientations for the four groups of intact and fractured rocks vary within 2°MN from each other (Table 4). A direct comparison between the methods for individual DIFs generally reveals similar values (within 2–3°). However, for individual DIFs, the difference can be up to 5° for intact rock and 8° for fractured rock (Fig. 7). We note that DIFs are not formed below 900 m depth.

A total of 1525 structural measurements was mapped, with 657 measurements (43%) made on foliation and the remaining 868 measurements (57%) made on natural fractures (Figs 4 & 8; Table S8). As mentioned above, our results likely are biased, and may overestimate the number of open fractures and brittle deformation structures and underestimate the number of ductile deformation structures. Some of the interpreted fractures (brittle structures) may be folds (ductile structures). We have no access to the drill cores, so core-log correlations are unavailable.

Type A fractures are interpreted as open fractures because they are detected on both AMPL and TT images (Fig. 4a, b). Type A fractures constitute 22% of all natural fractures, and are subdivided into two subgroups, single and high contrast fractures and damage/weak zones, respectively. A total of 181 single fractures, and 11 damaged/weak zones are measured. Type A single fractures are fairly evenly distributed with depth to ∼700 m depth, below which fracture density is distinctly lower (Fig. 7). Their dip directions generally have a bimodal distribution, with dip directions towards the NE to east (45–90°MN) and SW to west (225–270°MN) (Figs 8 & 9). However, scattered values outside these ranges are commonly observed at shallow depths (<200 m), as well as occasionally near 300, 400, and 900 m depths. In total, the dips of single fractures vary from 20 to 86°. Most fractures either have steep to moderate dips, i.e. having a dip of more than 60° or between 30 and 60°, respectively (Fig. 9). Only nine fractures have shallow dips (<30°). Steeply dipping (>60°) fractures are observed to ∼900 m depth, but most are observed above 700 m depth. The dip direction of most steeply dipping fractures is towards the NNE to east (Figs 8 & 9). Moderately dipping fractures (30–60°) are measured over the entire borehole length, but are more common between 100 and 530 m depths, with SSW–NNW dip directions (Fig. 8). All but one of the shallow dipping fractures have SW to west dip directions, and are observed above 500 m depth. Type A damage/weak zones are observed from 108 to 497 m depth (9 fractures), and at 903 and 941 m depths (two fractures). Most dip directions are towards the WSW (222–258°MN), with all damage/weak zones having shallow to moderate dip values (12–54°; Fig. 9).

Type B fractures are interpreted as sealed and likely filled fractures, with moderate to high contrast on AMPL images, but they are not visible on TT images (Fig. 4a–c). A total of 191 Type B fractures are measured, which corresponds to 22% of all fractures. They are measured over most of the borehole length with highly variable dip directions. The majority of Type B fractures follows a similar, but less pronounced, bimodal trend as the Type A fractures, with many dip directions towards the NNE–ENE and SSW–WNW (Fig. 8). More scatter in dip directions is observed above 200 m depths, but also in several intervals at depth (Fig. 8). The majority of Type B fractures are moderately dipping (30–60°). Steeper dip values also are observed, and are more common than shallow dipping fractures (<30°).

Type C fractures are interpreted as sealed fractures and are characterized by low contrast in AMPL images (Fig. 4a–d). In total, 485 Type C fractures are observed, corresponding to 56% of all natural fractures. Type C fractures have been subdivided into three subgroups, namely 386 low AMPL contrast fractures, 72 partial fractures and 27 truncated fractures (Fig. 4a–d). Figure 8 shows that their combined downhole distribution is similar to that of Type A fractures. Regarding the dip of these fractures, two main trends are observed (Fig. 8). The majority of NE to east dipping fractures are steeply dipping (>60°), while fractures with SW–WNW dip directions more often are steeply to moderately dipping. The majority of partial and truncated fractures either are steeply dipping (>60°), or moderately dipping (37–60°); only a few low AMPL contrast fractures but no partial and truncated fractures have shallow dips (Fig. 8). We also observe several Type C fractures from the three subgroups with NW to north dip directions. The sealed fractures (Types B and C) together correspond to 78% of all natural fractures.

Foliation is measured from 67 to 995 m depth, with a total variation in dip direction and dip ranging from 15 to 323°MN and 18–73°, respectively (Figs 4d & 8). However, most measurements are concentrated in the uppermost sections with granitic gneiss and tonalite gneiss, above 173 m depth. Within this interval, dip direction of foliation generally ranges from 210 to 305°MN and foliation dips are generally shallow to moderate (20–50°). Below 173 m, foliation is mapped within more limited depth intervals, e.g. ∼300–450 m, 495–530 m, ∼900 m, and below 970 m. Most dip directions within these intervals are towards the SW to west, but the total range is 5–300°MN (Fig. 9). The variation in dip values is similar as for the uppermost section, but moderate to steep dips (50–73°), and steep dips (60–73°) are measured near ∼520 m, ∼620 m, and ∼720 m depths.

The first part of the discussion regards the results of different strategies applied for image analyses of stress indicators. This is followed by a discussion of stress orientation in the GE-1 borehole and some potential uncertainties related to drilling conditions. Finally, implications for geothermal exploration are discussed.

The vertical and radial resolution of the HiRAT data in the GE-1 borehole are 2 mm in the vertical direction, and 0.7 mm in the radial direction. This may result in overconfidence in the ability to produce reliable assessments of in-situ stress orientation, also supported by the data produced in WellCAD (up to six significant numbers). The interactive onscreen interpretation of AMPL and TT images introduces a source of uncertainty, which is largely overlooked in current guidelines (e.g. Tingay et al. 2016).

We attempt to assess the size of uncertainty introduced during BO and DIF selection in WellCAD by applying three methods (A–C) using the same pre-processed data (no changes in setting or additional processing). For Methods A and B, two measurements are made for each individual BO and DIF, with one measurement per trace at opposite sides of the wellbore. For Method C, four measurements are made for each individual BO and DIF, with two measurements per trace at opposite sides of the wellbore.

Method A, the manual procedure, involves analyses of TT, AMPL, 3D logs and rose diagrams, and allows the picking of individual BO and TT traces, where the two traces may have different lengths, widths and other trace separations than 180°. Method B, the mirror procedure, reproduces the same geometry (orientation, length and width) of a trace to its counterpart with 180° separation. The mirror, however, does not always reflect the true position (orientation) and shape (length and width) of the second trace. Regardless, Methods A and B result in similar values of mean stress orientation and standard deviation for this dataset (Fig. 6). Note also that similar mean and median values are obtained for the BO subgroups (Q1–Q4) that are in accordance with the results from DIF analyses (Tables 2 & 3).

A slightly different approach for BO and DIF interpretation is applied in Method C. First, interpretations are based only on the AMPL images, which are more sensitive than TT images for fracture detection (e.g. Deltombe and Schepers 2000). Second, analyses of BOs and DIFs in Method C are categorized, considering whether stress indicators are developed in fractured rocks (Type 0) or intact rocks (Type 1). Third, the collection of four measurements of each individual BO and DIF (two values per trace) provides a measure of the accuracy of data, especially the standard deviation of the mean value of the four traces signal if the fracture traces are separated by 180° or not, and if interpretation is influenced by local heterogeneities in the wellbore, which commonly occur in metamorphic crystalline rocks.

The combined effect of a different interpreter and a slightly different strategy has resulted in interesting results (Fig. 7; Table 4). Most stress indicators are developed in or immediately adjacent to already fractured rocks. Only 24% of the BO data and 27% of DIF data are developed in intact rock, i.e. fulfils the basic CHILE assumptions for stress-induced failure theories. In general, an open fracture, or a fracture with different strength influences the stress locally (e.g. Haimson and Cornet 2003). Fluid flow may also alter the fracture walls and reduce their rock strength. In the GE-1 borehole, results from DIF analyses from the three methods yield identical mean and median values for the S_Hmax orientations, irrespective of whether the data were collected in intact or fractured rock (Fig. 8, Table 4). In contrast, Method C demonstrates differences in length-unweighted mean and median S_Hmax orientations for BOs and DIFs: BO analyses yield higher values (156–159°MN) for intact rocks compared to fractured rocks (148–149°MN), whereas DIF analyses reveal more similar values for intact and fractured rocks (150–152° and 153°, respectively). When comparing BO azimuths picked with maximum and minimum widths of individual BO traces, some traces show significant differences in orientation (≤11°MN), while others are nearly identical. The corresponding comparison of DIF azimuths picked with maximum width and linear fit generally demonstrates good agreement (less than 3°). However, differences of up to 5° and 8° were observed for intact and fractured rock, respectively.

Earlier studies based on HiRAT data from several 0.5–2.5 km deep boreholes in metamorphic crystalline rocks across Fennoscandia typically reveal only a few short sections with BOs and/or DIFs, at most (e.g. Ask et al. 2015; Ask and Ask 2017; Wenning et al. 2017; Pierdominici and Ask 2021; Juhlin et al. 2022). In comparison, a high occurrence of stress indicators has been observed in the GE-1 borehole (Figs 5 & 7). Up to 180 individual BOs, DIFs and PCFs are measured in the GE-1 borehole, with total cumulative length of at least 135 m (Fig. 6). The different stress indicators reveal very similar values of the mean stress orientation: in total, length-weighted mean S_Hmax orientations vary from 153 to 159°MN, with all but mirror PCFs (Method B) showing an even narrower range in mean S_Hmax orientations (153–155°MN). Associated standard deviations range from 5 to 11°, with mirror PCFs revealing the highest standard deviation (Fig. 6). This implies that length-weighted mean S_hmin orientations vary from 63 to 69°MN.

The cumulative lengths of good and moderately good and partial BOs (Q1, Q2, Q3) are 4.3, 3.3 and 4.1 m, respectively, according to Method A, and are comparable within the ‘normal’ outcome of other BO studies in Fennoscandian rocks (e.g. Ask et al. 2015; Ask and Ask 2017; Wenning et al. 2017; Pierdominici and Ask 2021; Juhlin et al. 2022. Proto BOs (Q4) are by far the most abundant BO type occurring in the GE-1 borehole, with 58 BOs and a cumulative length of 47.2 m according to Method A. Our results suggest that the four subgroups reveal similar stress orientations (Fig. 5, Tables 3 & 4). The inclusion of proto BOs greatly enhances reliability of the analyses, from 1% to 6% of continuous information on BO orientation over the logged borehole length. While the number of observed BO zones by far exceeds the required number (≥10 distinct BO zones) and standard deviation (≤12°) of WSM A-quality data, the obtained cumulative length of measured BOs only fulfils the length criteria of WSM B quality (≥40 m) if proto BOs are considered (Tingay et al. 2016). In the absence of proto BOs, the results from BO analyses would result in WSM quality D. The cumulative length of DIFs is higher than the cumulative length of BOs, and provides continuous information on stress orientation over 7% of the logged borehole length. Because the cumulative length is less than 100 m, the DIF analyses result in WSM B quality. However, because the WSM quality criteria for BOs and DIFs are identical, we note that the combined results of BOs (including proto BOs) and DIFs would qualify as WSM A quality, although there are some sections with simultaneous occurrence of BOs and DIFs. Presently, no WSM quality criteria exist for PCFs.

In addition to a high number of stress indicators, both BOs (Q3 and Q4) and DIFs are first observed at extremely shallow depths (173 m) for metamorphic crystalline rocks (Tables S2 to S7). Generally, BOs are developed at greater depths (e.g. Wenning et al. 2017; Pierdominici and Ask 2021; Juhlin et al. 2022), but have also been observed at shallower depths in Forsmark, on the east coast of Sweden (Ask et al. 2006; Ask and Ask 2007). It should be noted that the majority of BOs, DIFs and PCFs are formed at greater depths in the GE-1 borehole (Figs 5 & 7). In fact, most BOs are observed below 800 m depth. While DIFs are fairly evenly distributed from ∼260 to 900 m depth, most consistent DIFs are measured from ∼880–900 m depth, which also coincides with the deepest observations of DIFs in the GE-1 borehole. Below 900 m, the orientations of BOs reveal a gentle and steady rotation towards north (Fig. 5). The lack of DIFs and the gently northward rotating BOs below 900 m may be indicative of a change in the stress field, possibly in response to a nearby fault zone, for example the west-dipping Göta Älv Shear Zone (Fig. 1).

The nature of proto BOs is not yet fully understood. Deltombe and Schepers (2000) recognize that AMPL images can detect potential BO areas, i.e. partly failed but not yet broken out before becoming visible on TT images. Proto BOs have previously been observed in Fennoscandian metamorphic crystalline rocks, in Forsmark, on the east coast of Sweden (Ask et al. 2006; Ask and Ask 2007). Investigation of drill rig data showed an unusual amount of drill bit wear, high feed forces yet low penetration rates, and detailed lithological descriptions suggested hard and quartz-rich formations. It was proposed that proto BO formation could have been favoured by insufficient cooling of the drill bit during drilling, which caused local heating of the rock and generation of induced thermal stresses (Ask et al. 2006; Ask and Ask 2007). Brudy and Zoback (1993), Zoback et al. (2003), and others argue that DIF formation may also be triggered by local drilling conditions, e.g. induced tensile stresses due to high temperature contrasts between the drill bit or drilling fluid and the formation, and lowered effective stresses at the wellbore due to high pressure and high-density drilling fluid. Both examples result in lower magnitude of circumferential stress at the wellbore, distributed equally around circumference, which is why DIFs are also considered to reveal the maximum horizontal stress in intact rock. For PCFs, high drilling fluid and pore pressure and the rough shape of the bottom hole may trigger the formation of PCFs (Li and Schmitt 1998).

Present knowledge about the drilling conditions in the GE-1 borehole is almost non-existent. We know that water was used as a drilling fluid, and that water is much less effective than drilling mud for stabilizing the wellbore. However, high water pressures during drilling may occur. The GE-1 borehole was cored using a wireline system, which tends to generate even-shaped boreholes and bottom holes. This is confirmed by the HiRAT data, which generally show good borehole quality, with only 6% of drilling artefacts (spiralling) observed over the logged length (Table S1). The overall good borehole quality may suggest that feed forces and drilling fluid pressures were not extremely high during drilling. Temperature logging over two months after the completion of drilling operations reveals a maximum borehole fluid temperature at T.D. of ∼22°C (Hogmalm et al. 2021). Available data do not allow in-depth analyses of drilling, nor do they provide evidence as to whether drilling influenced formation of stress indicators. Other types of measurements (e.g. in-situ stress magnitudes, borehole groundwater pressures, rock strengths, anisotropy of the wellbore) are needed to confirm our interpretation of stress orientations. The observation of sections with both BOs and DIFs suggests that the differential horizontal far-field stresses are large enough to induce compressional and tensional failure. Increased drilling pressure decreases the circumferential stress by an equal amount around the wellbore, resulting in a narrowing of the compression failure zone and formation of DIFs (e.g. Zoback et al. 2003). We suggest that the distribution of BOs, DIFs and PCFs reflects the orientation of the far-field in-situ horizontal stresses, and that potential drilling conditions likely affected the wellbore equally. Thus, potential drilling disturbances may have greater impact on assessment of stress magnitudes using BOs, DIFs and rock strength data (e.g. Moos and Zoback 1990; Zoback et al. 2003).

During geothermal exploration, critical data are collected for design and mitigation purposes during construction and long-term operation of EGS reservoirs. In this study, we have characterized the orientation of horizontal stresses to 1 km depth and mapped the depth distribution, dip direction and dip of natural fractures and foliation. This allows us to make initial analyses by assessing whether observed fracture orientations are oriented in optimal directions for fluid flow, reviewing the current knowledge of the in-situ stress state in the region, conducting a preliminary assessment of the stress regime, and providing recommendations on additional investigations needed for future EGS development in the area.

The mean orientation of horizontal stresses obtained in this study provides valuable information on the regional stress field, which in turn, can be integrated into various types of numerical analyses (e.g. Jin et al. 2022). For crystalline basement rocks, past experiences suggest that orientation of in-situ stress generally remains rather constant with depth (e.g. Evans et al. 1999). This may imply that the results from this study could be extrapolated to greater depths, at least away from larger pre-existing structures such as fault zones. It is well established that the stress field is influenced by faults and other structures with different rock strength and/or fluid pressure than the surrounding rock mass. Our results reveal a small shift in BO orientation across some fractures (Fig. 3a), whereas the BO orientation remains uniform across other fractures. The former observation indicates differences in rock strength and/or fluid pressure across the fracture. Figure 3a also reveals that foliation planes are (sub)parallel with fracture planes. Detailed studies of such variation of stress orientation across fractures provide important information for understanding stress heterogeneities, including fault-foliation interaction. Such information is important for detailed characterization of a future reservoir.

A crucial component of a successful EGS development in Gothenburg is the creation of a permeable reservoir at a target temperature of ∼120°C, which is expected to occur at 5–7 km depths (Hogmalm et al. 2021). The crystalline bedrock in the study area is anticipated to have low permeability, although fracture zones and other types of pre-existing structures may locally enhance permeability. Hydraulic stimulation techniques generally are applied to enhance EGS reservoir permeability (e.g. Kwiatek et al. 2019). Zhang (2019) discusses the impact of in-situ stresses on hydraulic fracturing during EGS stimulation. In intact rock, hydraulic fractures are formed approximately perpendicular to the orientation of the minimum principal stress (Hubbert and Willis 1957). In stress regimes where the minimum principal stress is horizontal (i.e. normal and strike-slip stress regimes), hydraulic fracturing would generate vertical fractures, while horizontal fractures would be generated in a reverse stress regime where the minimum principal stress is vertical (e.g. Zhang 2019). The observation that DIFs are formed parallel with the borehole axis of the near-vertical GE-1 borehole, suggests that the minimum and maximum horizontal stresses and the vertical stress are principal stresses. Our results suggest a roughly NNW–SSE mean orientation of S_Hmax and ENE–WSW mean orientations of S_hmin from 0.2–1.0 km depth. As a result, hydraulic stimulation would generate vertical fractures striking towards the NNW–SSE in normal and strike-slip stress regimes, and horizontal fractures in a reverse stress regime.

The geometry of natural fractures may interact and limit propagation of hydraulic fractures during stimulation (e.g. Evans et al. 1999). These authors suggest that the hydraulic contact between injection and production wells is best engineered by opening or improving flow paths within the pre-existing fracture network. It is beyond the scope of this study to conduct detailed geomechanic analyses of fracture stimulation and fluid flow. Here, we compare inferred in-situ horizontal stress orientations with the dip and dip directions of Type A fractures (Fig. 9). We note that our current interpretation, which builds on the classification scheme of Massiot et al. (2018), is preliminary, because HiRAT images alone cannot unequivocally distinguish open fractures. Additional measurements are needed to verify our interpretations (e.g. high-resolution formation micro imager and temperature logs, and in-situ flow tests), as well as a more in-depth structural analysis (e.g. Massiot et al. 2018). Both strike-slip and normal faulting stress regimes would favour enhanced permeability of pre-existing steeply dipping fractures with dip directions parallel with the mean orientation of S_hmin (e.g. Zhang 2019). Our results suggest that a minority of Type A fulfils this criterion, namely those that are steeply dipping (>60°) and with dip directions towards WSW (∼225–270°MN; Fig. 9). The horizontal stresses are major and minor principal stresses in the strike-slip faulting regime, whereas they are intermediate and minor principal stresses in a normal faulting regime. Fractures with shallow dip (<30°) are likely best oriented for enhanced permeability in a reverse faulting stress regime, where the horizontal stresses are the maximum and intermediate principal stresses. Figure 9 also includes the dip direction and dips of foliation. Our results and those of Hogmalm et al. (2021) suggest that existing fractures often are (sub-) parallel with foliation, and because foliation represents local anisotropy, hydraulic linkage could be enhanced by favourable oriented foliation planes. While Figure 3a demonstrates an example of this observation, our results indicate that the potential for improved hydraulic linkage through foliation likely is limited, because the vast majority of observed foliation planes have moderate dip values (Fig. 9).

Few stress measurements have been conducted below 300 m in the Gothenburg area; in fact, the GE-1 borehole is the deepest borehole in the region (Hogmalm et al. 2021). Based on overcoring and hydraulic stress measurements in Fjällbacka (∼120 km north of the GE-1 borehole), Wallroth (1990) constrained a strike-slip stress regime to about 450–500 m depth, with a strike of the maximum principal horizontal stress of NW–SE. Stress measurements in deeper boreholes in other regions of Fennoscandia suggest that the thrust regime often is the dominant stress regime near the surface, down to ∼1 km depth (e.g. Stephansson et al. 1991; Martin 2007; Ask et al. 2009; Ask and Ask 2018, 2019), but transitions to a strike-slip regime may occur at shallower depths than 1 km (e.g. Martin 2007; Ask et al. 2009; Ask and Ask 2018, 2019), and sometimes to a normal faulting regime at greater depths (Ask and Ask 2019). Based on available data on stress regimes in the Gothenburg area and Sweden as a whole, it seems that the observed occurrence of BOs, DIFs and PCFs in the GE-1 borehole could indicate that a strike-slip stress regime is more likely to prevail to 900 m depth, than a normal faulting or a thrust faulting regime.

As mentioned above, stress indicators derived from HiRAT analyses provide data on the stress orientation, but not on the stress magnitudes. Nevertheless, the occurrence of DIFs is indicative of high differential horizontal stresses (e.g. Zoback et al. 2003), and PCFs are prone to develop in strike-slip faulting or normal faulting stress regimes (e.g. Li and Schmitt 1998). Additionally, we observed some sections with an overlap in the formation of BOs and DIFs. To generate sufficient amplitude of the circumferential stress around the wellbore, exceeding both compressional and tensional rock strengths, we propose that it is more likely for the differential horizontal stresses to correspond to the maximum and minimal principal stresses, rather than involving the intermediate principal stress. This leads us to suggest that the distribution of observed stress indicators in the GE-1 borehole supports a strike-slip stress regime, rather than a normal faulting or a thrust faulting stress regime, a suggestion further supported by the studies cited above.

For the further development of a deep EGS system, the current embryo of a rock stress model needs to be further developed. In parallel, more in-depth structural analyses (e.g. Massiot et al. 2018) as well as more advanced hydraulic stimulation strategies (cf. review of Jin et al. 2022) should be developed. The observation that the orientation of horizontal stress in crystalline basement rocks is uniform with depth away from faults (Evans et al. 1999) needs to be verified. Similarly, if EGS development is likely to target larger fault zones in the area, it is important to assess how these faults react to the prevailing stress field (e.g. Evans et al. 1999) and hydraulic stimulation (e.g. Jin et al. 2022). Additional investigations can be conducted in the GE-1 borehole. The co-occurrence of BOs and DIFs and the existence of drill cores offer the opportunity to develop stress polygons that assess the stress regime (e.g. Moos and Zoback 1990; Zoback et al. 2003). However, such studies also require access to pore pressure data. We have discussed above that drilling conditions may have generated DIFs and PCFs at lower effective circumferential stresses that are lower than the in-situ stress field. To constrain the stress field in the GE-1 borehole, hydraulic stress measurements using hydraulic fracturing and hydraulic testing of pre-existing fractures (e.g. Haimson and Cornet 2003) are required to constrain the effective stress field to 1 km depth and beyond. A local seismic network could collect natural microseismicity and focal mechanism of those events. Integration of stress indicators, hydraulic measurements, and focal mechanisms are needed to develop the stress model in the region of borehole GE-1, in subsequently deeper boreholes to address the much greater anticipated depth of the envisioned EGS. Improved imaging of the subsurface (e.g. reflection seismic data), detailed structural analyses and numerical modelling studies will also be needed to develop the strategy for future EGS development.

The main objective of this study is to assess the orientation of horizontal stresses from borehole stress indicators, and to discuss the implications for geothermal exploration. We conducted detailed analyses of HiRAT images from the near-vertical, 1 km deep GE-1 borehole in Gothenburg, SW Sweden. The bedrock in Gothenburg has a long and complex deformation history, involving metamorphic recrystallization, and several phases of ductile and brittle deformation (e.g. Andersson et al. 2002; Viola et al. 2011; Bergström et al. 2020; Bingen et al. 2021). Thus, it is one of the more challenging rock types for stress measurements.

The results suggest a uniform NNW–SSE orientation of the maximum horizontal stress, based on an unusually high abundance of BOs, DIFs and PCFs observed from 0.2–1.0 km depth. The formation of observed stress indicators could have been influenced by drilling conditions, for example, induced thermal stresses and/or high drilling fluid pressures. Unfortunately, further investigations cannot be conducted due to the lack of drill rig data. We argue that induced thermal stresses and high drilling fluid pressure likely influence the effective circumferential stress uniformly around the wellbore. This would likely have little impact on the interpretation of stress orientations, but would adversely affect further analyses of stress magnitudes from BOs and DIFs. It is outside the scope of this study to conduct detailed analyses of the geometry of the natural fracture network and the inferred stress field.

Following the works of Hubbert and Willis (1957), Evans et al. (1999) and Zhang (2019), we propose that some dip direction and fracture dips have favourable orientations for EGS stimulation in different stress regimes. These include steeply dipping fractures with WSW dip directions for strike-slip and normal faulting stress regimes, and shallow dipping fractures for a reverse faulting regime.

A limited number of stress measurement studies targeting depths below 300 m have been published beyond the Gothenburg area. The results from relatively local study (Wallroth 1990) and other studies in the Nordic countries (e.g. Stephansson et al. 1991; Martin 2007; Ask et al. 2009; Ask and Ask 2018, 2019), together with the observed occurrence and distribution of stress indicators in the GE-1 borehole, allow us to propose that the stress regime in the GE-1 borehole likely is a strike-slip faulting regime. This interpretation needs validation through further tests, for example, hydraulic stress measurements and earthquake focal mechanisms. Our study is the first step of a future final rock stress model, but further studies in the GE-1 borehole and from greater depths are needed for the development of an EGS system in Gothenburg.

The second aim of this study is to test different strategies for interactive stress analysis through visual inspection of acoustic images, and to highlight the need to develop better guidelines for data interpretation. We have tested and compared three methods of data interpretation. Length-weighted mean orientation of maximum horizontal stress for Methods A and B for BOs are 157 ± 7°N and 157 ± 6°N, respectively (Fig. 6). DIF and PCF analyses yield similar values, 158 ± 5°N (DIFs) and 159 ± 7°N and 163 ± 11°N (PCFs). The interpretations of horizontal stress for Method C suggest that interpretation of BO orientations, especially in fractured formations, increases the variability of results. In contrast, DIF data appear to produce more similar results in intact and fractured rocks, regardless of analysis method.

This study is a contribution to a test project for geothermal exploration in Gothenburg, financed by the Swedish Energy Agency and Göteborg Energi AB. We thank Göteborg Energi AB for permission to use their data, Simon Rejkjær (Lund University) for obtaining HiRAT data, and Axel Sjöqvist (Axray Scientific AB) and Johan Hogmalm (University of Gothenburg) for providing data access and information about the project. We thank Jenny Andersson (Geological Survey of Sweden, Uppsala University) for information and review of the regional geological description. We thank two anonymous reviewers and the editor of the GSL Special Publication, Cécile Massiot, for constructive comments that have greatly improved the manuscript. Special thanks to Production Editor Rachael Kriefman and the production team for bringing the manuscript to print.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

MA: conceptualization (lead), formal analysis (supporting), investigation (equal), methodology (equal), writing – original draft (lead); SP: conceptualization (supporting), formal analysis (lead), investigation (equal), methodology (equal), visualization (lead), writing – original draft (supporting); J-ER: investigation (equal), resources (equal), writing – original draft (supporting).

Our work has been funded by the Department of Geosciences, Uppsala University (Ask), Geomechanics and Scientific Drilling, GFZ (Pierdominici), and Engineering Geology, Lund University (Rosberg).

The dataset generated and analysed during this study is included as Supplementary material. The original image log is owned by the municipal energy company Göteborg Energi AB, who can be approached for information.