Subsurface fracture network characterization for a CO₂ basalt mineralization target: new insight on fluid flow from wellbore logs for the Jizan Group Open Access

The JG is a thick sequence of late Oligocene–early Miocene volcanic and clastic sedimentary rocks deposited within the continental rift valley preceding the opening of the Red Sea (Schmidt et al. 1983) (Fig. 1). Fossil evidence and ⁴⁰Ar–³⁹Ar dating of volcanic and subvolcanic rocks indicate that the JG lithologies range in age from 30 to 21 Ma (Madden et al. 1983; Pallister 1987; Sebai et al. 1991).

The JG and its lateral equivalent, the Sita Formation (Pallister 1987), form a continuous belt comprising predominantly mafic volcanic and volcaniclastic rocks interbedded with subordinate clastic sedimentary rocks, which crop out along the southern Red Sea coastal plain between latitudes 15 and 21°N. These rocks were deposited into a continental rift possibly extending from northern Somalia to as far north as the Gulf of Suez (Mège et al. 2016). The volcanic rocks of the JG consist primarily of basaltic lava flows, tuffs and agglomerates, accompanied by subvolcanic rocks such as dykes, sills, gabbro lopoliths and granite stocks, and represent a bimodal basalt–rhyolite magmatism (Schmidt et al. 1983; Coleman and McGuire 1988; Coleman 1993; Basch et al. 2022). The volcanic rocks have undergone extensive hydrothermal alteration, including devitrification and mineral replacement that occurred at temperatures up to 300°C (Torres 2020; Oelkers et al. 2022). The petrographic analysis of the wellbore cuttings of the Jizan pilot site, presented by Omar et al. (2025), shows that the basalts at depth have fine- to medium-grained intergranular textures composed mainly of plagioclase (45–50 vol%), clinopyroxene (25–30 vol%), Fe–Ti oxides (c. 5 vol%) and variable amounts (5–10 vol%) of altered glass (e.g. chlorite and zeolites). X-ray fluorescence and X-ray diffraction data indicate that the basalts are slightly altered, with loss on ignition (LOI) of <3.4% and a Chemical/Plagioclase Index of Alteration (CIA/PIA) of c. 50% (Omar et al. 2025).

The JG unconformably overlies the Proterozoic basement and erosional remnants of pre-rift sedimentary rocks of Paleozoic and Mesozoic ages and is generally tilted ∼30° towards the Red Sea due to the presence of numerous steep and low-dipping antithetic normal faults (Voggenreiter et al. 1988; Bohannon 1989; Fedorik et al. 2023). This group is overlain in the Red Sea basin by Miocene and younger sedimentary rocks, which include conglomerates, sandstones, shales, anhydrite and the mid-Miocene salt (Gillman 1968; Hughes and Johnson 2005). According to the structural model proposed by Fedorik et al. (2023), the JG fills a set of half-graben bounded from the west by antithetic normal faults with fault activity migration towards the Red Sea. These half-graben are progressively tilted due to the accumulation of younger layers of the JG and the progressive migration of faulting towards the continent–ocean boundary (Delaunay et al. 2023).

A few studies have investigated the fault and fracture network of the different lithologies in the Jizan area (i.e. Arrofi and Abu-Mahfouz 2023; Fedorik et al. 2023). These show that the dominant fault and fracture set in the area strikes from 140° to 160°N, which is approximately sub-parallel to the Red Sea axis. The outcrop fracture analysis of Fedorik et al. (2023) (Fig. 1b) shows that this set is composed of normal faults and mode-I fractures approximately parallel to the maximum principal horizontal stress (S_Hmax). These are often accompanied by a conjugate fracture set, composed of NNW–SSE- and WNW–ESE-striking shear fractures. These fractures can be the result of the magmatic loading and migration of faults towards the Red Sea that bends the juxtaposed layers. An additional but minor joint set, striking orthogonal to the Red Sea, was observed only inside large dykes (Fedorik et al. 2023) and may be due to shrinkage cooling (columnar jointing).

As part of the pilot test, five vertical wells from 120 to 1001 m deep were drilled on a site located to the north of the Jizan Refinery and Power Plant (Fig. 2).

To preserve basalt permeability, the wells were drilled underbalanced with an air hammer rig and local freshwater conditioned with surfactants (soap) only. Results demonstrated that JG basalts are easy to drill through with an average penetration rate of 7.0 m h⁻¹. Conditions in the subsurface are benign, with no overpressure, nor corrosive or toxic fluids (e.g. H₂S) encountered. Jizan basalts proved to be competent lithologies capable of maintaining hole integrity in 700 m long open hole completions (Fig. 2). Injection testing of three deep wells on the site (i.e. W1, W3 and W4) demonstrated good injectivity in wells W3 and W4 with a maximum sustained injection rate of 1.1 kg s⁻¹ bar⁻¹ in W4. Consequently, wells W3 and W4 were selected to serve as an injection–production pair (doublet) for the CO₂ mineralization pilot test (Arkadakskiy et al. 2024). A variety of wireline logs were acquired before the pilot test to characterize subsurface discontinuity networks and their flow properties in the wells including optical imaging (OPTV, Geovista), acoustic imaging (BHTV, Robertson Geologging Ltd), bulk density (FDSB, Geovista), gamma ray (SGAM and NGRS, Geovista), temperature and conductivity (TCLI, Geovista), flow (impeller flowmeter, FLOW Geovista) and dual neutron porosity (DNNS, Robertson Geologging Ltd) (see Appendix A for specifications). To interpret, classify and categorize discontinuities, we analysed the optical and acoustic logs (i.e. OPTV and BHTV) using the software WellCAD (Advanced Logic Technology 2023). We subdivided the observed discontinuities as follows: layering (which includes all primary features such as stratification and flow banding), mechanically open natural fractures (MOF, also called natural fractures with a mechanical aperture), closed/filled natural fractures and damage zones (Fig. 3).

The layering is classified mostly using OPTV through inspection of the optical texture of the rocks. Individual layers are separated by surfaces that clearly represent contacts between rocks of different textures (e.g. lithology/facies/formation contacts) and generally include primary features such as stratification and flow banding. Fractures and damage zones are classified using both the OPTV and BHTV through an inspection of the rock optical texture and acoustic amplitude, respectively. Fractures are identified as surfaces that cut through texturally homogeneous rocks, contacts between rocks with different textures (i.e. layering) and/or other fracture surfaces. Whereas MOFs display clear evidence of an aperture both on the OPTV and BHTV (i.e. reduction of the acoustic amplitude or signal loss), at the logging resolution (OPTV ∼0.5 mm; BHTV ∼1 mm), closed/filled fractures do not. On the OPTV most fractures are associated with quartz–carbonate (calcite) mineralization, with the closed/filled fractures appearing as light-coloured (whitish) sinusoidal lines along the walls of the borehole. Damage zones can be described as zones of diffuse deformation that do not form discrete planes/surfaces, when looking at the OPTV. These display a sinusoidal zone where the acoustic signal is reduced.

During the wellbore image log interpretation, we also mapped stress-induced borehole breakouts (BOs) (see example in Fig. 2). Using the World Stress Map (WSM) guidelines (Heidbach et al. 2016), we then estimated the trend and ranked the quality of the present-day S_Hmax.

The wellbore image logs revealed 789 depth intervals with stress-induced borehole BOs. This is a statistically representative dataset and enables us to reliably derive the azimuth of the present-day S_Hmax in each well (Table 1). Following the WSM quality ranking criteria (Heidbach et al. 2016), we rank the resulting S_Hmax azimuth in three of the five wells with a B quality. The BOs in well W2 result in a D quality because the number of breakout zones and total length in this well are low (19 and 10.83 m, respectively).

The average S_Hmax azimuth is N138°E with a standard deviation of 23°. The image log from the shallowest well, W5, does not show any BOs. The image log from the second shallowest well, W2, shows only a limited number of BOs, implying a 2016 WSM quality rank of D. This observation is expected since at shallow depth stress magnitudes are not yet sufficient for the hoop stress to break the rock at the wellbore wall in compression. In the three remaining wells with a B quality S_Hmax azimuth, the present-day stress orientation is relatively uniform across the area where the wellbores are located. Furthermore, the S_Hmax azimuth is consistent with the regional tectonics (i.e. normal faulting in situ stress environment, where the overburden is presumed to be the maximum principal stress, S₁) since it is sub-parallel with the Red Sea axis (Delaunay et al. 2023) as well as parallel with the strike of regional, basin-bounding normal faults (Voggenreiter et al. 1988; Bohannon 1989; Fedorik et al. 2023).

From the wellbore image logs, we mapped 2449 discontinuity planes and derived their dip angle and dip directions. Plotting the pole of each discontinuity in a lower hemisphere projection, we recognized five distinct sets of discontinuities (J1–J5) (Fig. 4; Table 2). The mean dip and dip direction of these five sets is as follows: 057°N/59° (J1); 276°N/46° (J2); 215°N/57° (J3); 123°N/48° (J4); 354°N/48° (J5).

Sets J1 and J2 are dominant, representing 44.5 and 27.1% of the entire discontinuity population, respectively. In three wellbores (i.e. W3, W4 and W5) these can be subdivided into subsets (Table 2). Set J5 is the least frequent; while it is clearly identifiable in W1, it is less clear in W4.

Plotting separately the discontinuity orientation according to the discontinuity type (e.g. banding/layering/foliation (Fig. 5a), MOFs (Fig. 5b), filled fractures (Fig. 5c) and damage zones (Fig. 5d)) makes it possible to better understand which sets represent the different discontinuity classes.

The stereoplot of filled fractures (Fig. 5c) indicates the presence of all five discontinuity sets (i.e. J1, J2, J3, J4 and J5). The stereoplot of MOFs (Fig. 5b) suggests the presence of the J1, J2 and J3 sets only. The stereoplot of banding/layering/foliation (Fig. 5a) suggests the presence of J1, J2, J4 and J5. The stereoplot of damage zones (Fig. 5d) suggests only the presence of J1 and J2. Therefore, while sets J1 and J2 are statistically representative for all discontinuity types, set J3 is statistically representative only for the filled fractures and MOFs. Sets J4 and J5 are statistically identified only for banding/layering/foliation and filled fractures.

Since W5 was not investigated by the flowmeter sonde, any discontinuity/fracture intensity and MOF analyses for this well could not be included in this work. Figure 6 and Table 2 highlight the results from the linear fracture intensity (P10) analysis of the wells logged with the flowmeter sonde. These clearly reveal that discontinuity sets J1 and J2 have the highest intensity in terms of respective mean and maximum values. In contrast, sets J4 and J5 are the least intense. Figure 6 shows that filled/closed fractures are the most intense discontinuity type, while banding/layering/foliation is the second most intense type. In Table 3 we report the statistics for each discontinuity type per wellbore.

Both optical (i.e. OPTV) and acoustic (i.e. BHTV) image logs reveal some MOFs with a wide apparent mechanical aperture (>5 cm). Two examples are shown in Figure 7.

The optical as well as the acoustic image tools from W1 reveal three wide MOFs that, from top to bottom, are oriented 039°N/54°, 067°N/51° and 042°N/55° (Fig. 7b). W2 has at least two closely spaced, wide MOFs with a mean orientation of 054°N/81° (Fig. 7c). In W3 we interpret one wide MOF oriented 048°N/53° (Fig. 7d), and W4 has four wide MOFs oriented, from top to bottom, 054°N/61°, 215°N/77° (Fig. 5e), 031°N/85° and 054°N/69°.

A well-developed natural fracture network characterizes the subsurface JG lithologies, from the shallowest (W5) to the deepest investigated portion (W1) (Fig. 4; Table 2). We distinguished four different classes (Fig. 3), namely, layering, MOFs, closed fractures and deformation zone(s). No direct evidence of faulting was found, since no displacement or stress-related breakout rotations (e.g. Zoback 2007) were identified. In addition, five discontinuity sets were also detected (Fig. 4). The two most important discontinuity sets inferred by their intensities are J1, dipping 60° towards 57°N, and J2, dipping 46° towards 276°N (Table 1). The most important discontinuity sets are extensional faults and fractures that dip c. 60° towards the NE and the layering of the JG gently dipping towards the W. Moreover, we identified three other discontinuity sets, namely J3, J4 and J5 striking WNW–ESE (J3) and WSW–ESE (J4 and J5). These findings and observations are in agreement with the structural geological model proposed by Fedorik et al. (2023) from outcrop investigations in the area.

According to our classification, discontinuity sets represent clusters characterized by specific orientations, whereas discontinuity types represent clusters characterized by specific geological features. Therefore, we are not associating individual discontinuity sets with a specific discontinuity type, or vice versa. Some discontinuity types tend to display orientations coherent with some discontinuity sets; however, this correlation is not explicit. For example, orientations of both closed fractures and MOFs are compatible with J1 (Fig. 5b, c). Despite that, closed fractures reveal orientations parallel to all interpreted discontinuity sets, whereas natural fractures that exhibit a clear apparent mechanical aperture only occur in sets J1, J2 and J3. Layering (Fig. 5a) is clearly present in sets J1 and J2 and it is far less obvious in sets J4 and J5. In contrast to the natural fractures (Fig. 5b, c), which are the dominant type in set J1, layering is dominant in set J2 (276°N/46°). These observations agree with the structural model proposed by Fedorik et al. (2023), and the general dip of the JG towards the SW. J1 represents a discontinuity set resulting from the original layering of the JG deposit setting in the half-graben. These are progressively tilted (to the SW) due to the accumulation of younger layers and the progressive migration of faulting towards the continent–ocean boundary (Delaunay et al. 2023). J2 probably represents discontinuities related to the emplacement of bedding-orthogonal dykes, which are successively tilted/rotated together with the bedding. The presence of J2 in both the layering and fracture types probably indicates that the mechanism generating this set (e.g. SW–NE extension related to the opening of the Red Sea) influences both dyke intrusion and faulting/fracturing in the area.

Finally, the damage zone class (Fig. 3) is clearly visible in sets J1 and J2, with the latter being the dominant set. This suggests that damage zones are mostly related to the layering of the JG (e.g. bedding and bedding-orthogonal dykes; Fedorik et al. 2023). Conversely, the orientation of natural fractures suggests a genesis related to the main normal faults in the area, which dip c. 60° towards the NE (Fedorik et al. 2023). During the wellbore log analysis and interpretation, we did not find any evidence of faults or faulting, such as displacement or borehole breakout rotation, but we cannot exclude their presence. In particular, most of the natural fractures with a wide mechanical aperture show an orientation comparable with the faults reported in the structural model of Fedorik et al. (2023); therefore, considering their anomalous aperture, we cannot exclude that they could represent partially mineralized extensional faults.

The wellbore image logs enable a quantitative characterization of discontinuity networks, defining the orientation and intensity values in the subsurface and correlating these to the natural fracture networks identified in outcrops. The wellbores are vertical (maximum deviation from <1 to 5°, see Table 1); therefore, according to Terzaghi's bias (Terzaghi 1965), subvertical discontinuities are missed or strongly under sampled. This is evident from the discontinuity stereoplots (Fig. 4), which are void of and/or underrepresented in features with dip angles above 80°. We presume that this could be a reason for the difference in fracture intensity values derived in this study (maximum intensity ∼6 m⁻¹; Table 2; Fig. 6) versus those reported by Fedorik et al. (2023) (maximum intensity ∼54 m⁻¹; Fig. 1).

By combining interpreted discontinuities and BOs with other wireline logs – in particular flowmeter, dual neutron porosity (DNP), formation density (FD) and gamma ray logs (Fig. 8) – we made correlations with subsurface fluid flow patterns. Notably, a close look at the wireline logs shows that the FD does not depict the true bulk density of the JG rocks. The four plotted FD logs indicate the mean compensated bulk densities range between 2.1 and 2.25 g cm⁻³, respectively. Geological knowledge of the area, drill cutting laboratory analyses (i.e. X-ray fluorescence, X-ray diffraction and thin-sections; Omar et al. 2025) and lithological interpretation of the OPTV image logs indicate that slightly altered basaltic massive lavas, dykes and sills represent a large fraction of the drilled rock volume. The density and porosity ranges for such massive basalts are between 2.8 and 3.0 g cm⁻², and 4 and <8%, respectively (e.g. Busch et al. 1992; Sigurdsson and Stefansson 2002; Gilbert and Salisbury 2011; Lee et al. 2021; Brett-Adams et al. 2023). Usually, errors in the compensated bulk density values measured by the FD log may be due to the tool's shallow depth of investigation (i.e. ≤10 cm) (Rider 1991), which makes it susceptible to the hole condition (e.g. a caved or rough hole can decrease the logged FD). Notwithstanding, this does not seem to be the case here, since no relationship between stress-induced borehole BOs and density is apparent. Therefore, we assumed that the FD logs were affected by instrumental and/or calibration errors.

In contrast, the DNP logs reveal porosity values in agreement with those reported in the literature (Brett-Adams et al. 2023). Mean compensated porosities are 6.2 and 7.6% (Limestone Porosity Units, LPU) in W3 and W4, respectively. The DNP tool does not directly measure formation porosity; instead, it is a hydrogen indicator (Rider 1991; Ellis and Singer 2007). In sedimentary rocks (e.g. limestone, dolostone, sandstone) hydrogen is mostly pore volume water-bound. In basaltic rocks, hydrogen might be present in mineral-bound water (e.g. illite, kaolinite, chlorite, smectite) (Rider 1991; Ellis and Singer 2007). Thus, DNP logs in basaltic rocks do not only reflect the presence of free water but also the composition of original mineral and/or alteration products. For these reasons, in this study, the DNP and FD logs are not considered useful to investigate the subsurface fluid flow of the JG basalts.

The flowmeter logs, i.e. fluid volume (FV), fluid temperature (FT) and fluid conductivity (FC; here, conductivity refers to electrical conductivity), are critical for the identification of flow zones (e.g. Hess 1986; Tsang et al. 1990; Rider 1991; Barton et al. 1995; Le Borgne et al. 2007). FV values abruptly change at discrete depths in wells W3 and W4, as do the FT logs for W2, W3 and W4 and the FC logs for W1, W2, W3 and W4. In general, these abrupt changes are correlated across all three log types and imply the inflow or outflow of fluids.

Looking at the spinner flowmeter logs (FV; Fig. 8), we identify one principal flow zone in well W3 at 320 m and one in well W4 between 379 and 390 m. The observed changes in FV do not appear to be related to the overall discontinuity/fracture intensity (P10) or breakout occurrence. Instead, they seem to coincide with the occurrences of fractures with wide mechanical apertures (≥5 cm; see Fig. 7). These fractures strike around NW–SE and dip towards the NE and SW at c. 60° and 80°, respectively (Fig. 8). Similarly, most of the FT and FC abrupt changes appear to be related to fractures with wide mechanical apertures that strike NW–SE (Fig. 8). The FT and FC variations possibly represent changes of the fluid flow (i.e. FV) below the resolution of the spinner flowmeter measurements (Drury 1989; Barton et al. 1995; Paillet 1998; Wacker and Cunningham 2008; Chatelier et al. 2011; Dausse et al. 2019).

The stress-induced wellbore failure analysis in wells W1, W2, W3 and W4 revealed 789 borehole BOs with a total length of ∼268 m or nearly 17% of the cumulative imaged borehole length (which is 1598 m; see Table 1). The image log from well W5 does not have any stress-induced BOs, likely because this well is the shallowest (Fig. 2c) and stress concentrations are not sufficient to cause compressive wellbore wall failures to occur. This hypothesis is further supported by the small number of BOs detected in the image log of W2, which is the second shallowest well on the site. No shallow stress-induced borehole BOs are expected because stress magnitudes are not yet sufficiently high to generate a stress concentration (i.e. hoop stress) at the wellbore walls for the rock to fail in compression.

The mean BO orientation for each of the four analysed wellbores ranges from 36.6°N to 52.3°N (Fig. 1), with an overall mean orientation for all four wells together (length-weighted average) of 48.2°N. This represents a mean present-day S_Hmax of 138–318°N (±23°) (Fig. 9).

According to Heidbach et al. (2010), the WSM quality rank of the estimated S_Hmax is B, meaning that more than five distinct BOs with a combined length higher than 40 m were mapped in four out of five well logs.

This S_Hmax orientation is consistent with regional present-day tectonics and associated stress patterns. In other words, S_Hmax is sub-parallel with the Red Sea mid-oceanic ridge axis trend (Delaunay et al. 2023). Thus, the corresponding orientation for the present-day minimum principal horizontal stress is NE–SW. These results are also consistent with the normal faulting stress regime in the region (Basch et al. 2022; Delaunay et al. 2023; Fedorik et al. 2023) including the main fracture sets, J1, J2 and J3, interpreted both from surface outcrops (Fedorik et al. 2023; Fig. 1a) and subsurface image logs (Fig. 3). Correlating present-day S_Hmax with MOF strike, we find that most natural fractures with a noticeable mechanical aperture strike (sub-)parallel to S_Hmax (i.e. J1, 325°N; J2, 183°N; J3, 122°N), with a maximum variation of c. 40°.

Studies based on subsurface data (e.g. Barton et al. 1995; Finkbeiner et al. 1997; Zoback 2007; Sathar et al. 2012) conclude that fracture hydraulic conductivity is mainly controlled by the presence of so-called critically stressed natural fractures. These can be described as fractures that are optimally oriented for compressive shear failure. According to this theory, fractures that are mechanically active are hydraulically conductive and fractures that are mechanically inactive are presumed hydraulically dead (Zoback 2007). In normal faulting stress regimes, critically stressed fractures should be oriented sub-parallel to S_Hmax but will form an angle of 30 to 60° with the main principal stress, the overburden (S_v). Our results fit this hypothesis, showing that fractures associated with fluid flow (Figs 7, 8) trend sub-parallel to S_Hmax and mostly dip c. 60°. However, an appreciable fraction of natural fractures with similar orientations (see yellow dots on stereoplots of Fig. 8) and mechanical aperture appear not to be hydraulically conductive (based on the FT, FC and flowmeter logs). Therefore, we can presume that fluid flow is not only controlled by the in situ stress field but also by other factors, such as fracture partial mineralization that can form cement bridges, allowing fractures to be mechanically open at depth and, therefore, creating a pathway for fluids (Laubach et al. 2004).

The results from this study provide useful lessons about subsurface fracture networks and in situ flow behaviour that might inform development plans for upscaling the Jizan CO₂ mineralization pilot project. The main flow zones in the near wellbore region appear to focus along natural fractures with a wide mechanical aperture. These strike NW–SE and dip NE, parallel to major regional structures (i.e. faults and dykes as mapped by Fedorik et al. 2023). Although we did not notice any direct relationship between these hydraulically conductive fractures with wide mechanical apertures and major regional structures, we can hypothesize that these structures play a critical role for fluid flow. Therefore, any future well locations should target such hydraulically conductive zones that follow a NW–SE trend and can be identified locally (in terms of their depths) along wellbore logs (i.e. image and flowmeter logs) and regionally over the Jizan area (e.g. faults and dykes as mapped by Fedorik et al. 2023).

In addition, the natural fracture population interpreted as being ‘closed’ (i.e. without any noticeable mechanical aperture) might still contribute to flow if we consider that fractures can still be hydraulically conductive along sub-millimetre aperture zones (i.e. below the image tools' resolution limits). Furthermore, in the near wellbore zone, the acidic fluid injected can dissolve some of the minerals in the ‘closed’ fractures thus enhancing hydraulic conductivity (Bacci et al. 2011). A variety of publications (e.g. Gale 1982; Olson et al. 2007; Bisdom et al. 2016) highlight the importance of such small-scale flow zones in the context of the overall hydraulic plumbing and flow behaviour of a reservoir formation. Thus, fluids may distribute quickly away from wellbores through fractures with wide mechanical apertures and then feed into small-scale diffuse fracture sets in the far field. These latter sets may then effectively distribute CO₂-rich fluids over large reservoir volumes providing appreciable surface area to react and mineralize.

In this conceptual model the presence of MOFs is a necessary but not sufficient requirement since not all MOFs appear to show significant flow. Thus, for upscaling purposes it is important to continue acquiring image and flowmeter logs in order to verify and validate flow locally within a well's observed and mapped fracture networks.

We have characterized the subsurface natural fracture network as well as potential flow zones in five wells drilled as a part of the pilot testing of the CO₂ mineralization potential of the JG basaltic rocks in the SW region of Saudi Arabia. To do this, we utilized wellbore image logs as well as other log data. The main findings are:

• The subsurface discontinuity network, composed of layering, fractures with a mechanical aperture, filled/closed fractures and damage zones, is characterized by the presence of five sets of discontinuities, i.e. J1 (057°N/59°), J2 (276°N/46°), J3 (215°N/57°), J4 (123°N/48°) and J5 (354°N/48°), where J1 and J2 are the most intense.

• The classification of the different discontinuity types allowed us to determine that closed fractures display an orientation compatible with all the detected discontinuity sets (J1 is the most frequent), the banding/layering/foliation surfaces are compatible with sets J1 and J2 (J2 is the most frequent), and the fractures with a mechanical aperture are compatible with J1, J2 and J3 (J1 is the most frequent).

• There is a consistent correlation of the subsurface fracture networks with those previously identified on the surface, which agrees with the regional structural model of Fedorik et al. (2023). The primary layering we describe confirms that JG volcanics have a general tilt towards the SW. These also exhibit the presence of extensive NE-dipping fractures with mechanical apertures (e.g. J1) that could possibly represent faults (although wellbore logs do not provide direct evidence of that), and NW–SE-striking fractures (e.g. J1, J2, J3). These fractures display evidence of fluid flow and are interconnected by fractures that strike in different directions (e.g. J4 and J5).

• Flowmeter logs indicate that subsurface fluid flow is mainly governed by NW–SE-striking fractures with wide apparent apertures (≥5 cm).

• An abundance of high-quality wellbore BOs provides a robust estimate of the present-day stress with depth; we find a maximum principal horizontal stress (S_Hmax) azimuth of 138–318°N, which is consistent with the present-day regional tectonics and (sub-)parallel to the MOFs.

The structural characterization of the subsurface fracture network and wellbore BOs enables us to expand and extrapolate the geological understanding developed from outcrop studies into the subsurface. While the primary porosity of the JG basalts is low, their permeability is greatly enhanced by the presence of an intense fracture network. This network is characterized by the presence of fractures with wide mechanical apertures that are nearly parallel to the present-day maximum horizontal stress. A few of these fractures are associated with changes in the flowmeter, temperature and electrical conductivity logs, which indicates fluid flow localization along these. Why some of the MOFs are hydraulically conductive and others are not is not understood and requires further investigation and data acquisition. For example, additional flowmeter logs with a higher resolution (lower flow measure range) such as the heat pulse flowmeter (measure velocity range: 0.1 to 3 m/min) would help to better understand the subsurface plumbing and corresponding fluid flow patterns in the JG.

Our findings are consistent with the positive results from the pilot test and demonstrate that the subsurface at the site has sufficient permeability to transmit injected H₂O–CO₂ fluid along preferential flow paths, represented by the fractures with a wide mechanical aperture. This indicates that the JG is a good target for in situ CO₂ mineralization in basalt owing to the extensive fracture network and the presence of highly permeable fluid flow zones, which allows carbonated waters to flow and react with a large volume of rock. Future research consisting of relogging the same wells after the CO₂ injection with a flowmeter and optical/borehole televiewer could be useful to understand how fluid flow pathways have evolved and changed in response to the in situ CO₂ injection and mineralization processes.

We would like to acknowledge Saudi Aramco, which provided the wellbore log data and gave permission to publish the results as part of their broad initiative to assess the potential for subsurface carbon mineralization in Saudi Arabia.

NM: conceptualization (equal), data curation (equal), formal analysis (equal), funding acquisition (equal), investigation (equal), methodology (equal), project administration (equal), resources (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – original draft (equal), writing – review & editing (equal); JF: conceptualization (equal), funding acquisition (equal), project administration (equal), resources (equal), supervision (equal), validation (equal), writing – original draft (equal), writing – review & editing (equal); YP: conceptualization (equal), validation (equal), writing – review & editing (equal); DB: conceptualization (equal), validation (equal), writing – review & editing (equal); MA: conceptualization (equal), funding acquisition (equal), validation (equal), writing – original draft (equal); AO: conceptualization (equal), validation (equal), writing – review & editing (equal); EO: conceptualization (equal), funding acquisition (equal), project administration (equal), validation (equal), writing – review & editing (equal); AA: conceptualization (equal), funding acquisition (equal), project administration (equal), resources (equal), validation (equal), writing – review & editing (equal); HH: conceptualization (equal), funding acquisition (equal), project administration (equal), resources (equal), supervision (equal), validation (equal), writing – review & editing (equal); SA: conceptualization (equal), funding acquisition (equal), project administration (equal), resources (equal), supervision (equal), validation (equal), writing – review & editing (equal); NK: conceptualization (equal), funding acquisition (equal), project administration (equal), resources (equal), supervision (equal), validation (equal), writing – review & editing (equal); ZA: conceptualization (equal), funding acquisition (equal), project administration (equal), resources (equal), validation (equal), writing – review & editing (equal); SRG: conceptualization (equal), funding acquisition (equal), project administration (equal), supervision (equal), validation (equal), writing – review & editing (equal); GB: conceptualization (equal), supervision (equal), validation (equal), writing – review & editing (equal); TF: conceptualization (equal), funding acquisition (equal), methodology (equal), project administration (equal), resources (equal), supervision (equal), validation (equal), writing – review & editing (equal).

This research received funding from Saudi Aramco and King Abdullah University of Science and Technology (KAUST) Research Funding Office (Award No. 4357).

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.

The wellbore logs used for the subsurface fracture network characterization in the study are not publicly available due to restrictions made by Saudi Aramco (data owner). In this study the data were used under the permission of Saudi Aramco.

In this appendix we provide details for the wireline logging tools used in this research (i.e. OPTV, BHTV, FDSB, SGAM, NGRS, TCLI, FLOW, DNNS) (Tables A1–A8). The specifications were retrieved from the website of Geovista (https://www.geovista.co.uk/witeline-logging) and Robertson Geologging Ltd (https://www.robertson-geo.com/borehole-structure-logging and https://www.robertson-geo.com/specification-sheets/) and were accessed on 3 January 2025.