The architecture of basalt reservoirs in the North Atlantic Igneous Province with implications for basalt carbon sequestration Open Access

Supplementary material: Processing parameters for the Viðoy overview and Viðareiði close-up photogrammetry models are available at https://doi.org/10.6084/m9.figshare.c.6949132

As global greenhouse gas emissions continue to rise, the requirement for rapid development of carbon capture and storage technologies is critical for addressing climate change (Benson et al. 2005). While classic CO₂ storage utilizes sedimentary reservoirs for sequestration, a new and exciting option is emerging where the greenhouse gas is stored in mafic or ultramafic rocks such as basalts (e.g. McGrail et al. 2006, 2011; Matter et al. 2009). Through this process, CO₂ reacts with the mafic minerals in basalt to form carbonates, leading to storage in solid mineral form (Lackner et al. 1995; McGrail et al. 2006; Gíslason and Oelkers 2014). The advantages of storing CO₂ in basalt include: (1) rapid mineralization and subsequent permanent storage: studies indicate near complete mineralization of CO₂ dissolved in water within two years after injection into a basalt reservoir (Matter et al. 2016); (2) large reservoir volumes are available worldwide in the form of Large Igneous Provinces (LIPs) and mid-oceanic ridges (McGrail et al. 2006; Snæbjörnsdóttir et al. 2014); and (3) the period needed for monitoring may be significantly reduced as the CO₂ quickly mineralizes and becomes immobilized (Snæbjörnsdóttir et al. 2020).

For the last decade, the CarbFix group in Iceland has injected CO₂ into onshore basalt reservoirs, proving the feasibility of this CO₂ storage method (Matter et al. 2016; Snæbjörnsdóttir et al. 2017; Gíslason et al. 2018). Nearly 65% of the CO₂ is mineralized within two months after injection and there are no signs of reduced permeability due to the mineralization (Clark et al. 2020). However, onshore injection in Iceland is so far on a small scale, injecting c. 12 000 tons of CO₂ annually (Snæbjörnsdóttir et al. 2020). In the future, however, storage at the gigaton (Gt) scale will be needed to counteract the current global emission rates (IPCC 2018). There is an urgent need to appraise the potential of offshore storage in the North Atlantic Igneous Province (NAIP) and other LIPs where there are large volume (often >100 000 km³) volcanic sequences that could potentially provide Gt scale storage sites if appropriate reservoir rocks can be identified and utilized (Planke et al. 2021).

The Faroe Islands, located in the NE Atlantic Ocean (Fig. 1a), expose a c. 3.2 km thick sequence of subaerially erupted basalt lava flows and volcaniclastic rocks called the Faroe Islands Basalt Group (FIBG) (Ellis et al. 2002; Passey and Bell 2007; Passey and Jolley 2008). The FIBG is a part of the much larger NAIP (Fig. 1a) created during the Paleocene to early Eocene (c. 62–54.5 Ma; Wilkinson et al. 2017) break-up between Greenland and Eurasia that led to extensive volcanism (Saunders et al. 1997; Passey and Bell 2007). The uplifted basalt sequences on the Faroe Islands can, therefore, be used as an analogue for offshore volcanic reservoirs of the NAIP along the coast of Norway, the UK and Greenland.

The basalt formations on the Faroe Islands are dominated by two types of lava geometric flow facies: simple (tabular) and compound lava flows (Rasmussen and Noe-Nygaard 1970; Waagstein 1988; Passey and Bell 2007). The simple lava flows comprise one singular lava sheet lobe, while the compound lava flows consist of several smaller anastomosing flow lobes (Walker 1972; Passey and Bell 2007). The internal architecture of each flow lobe can usually be divided into: (1) an upper and lower crust (flow top and flow base) that can be either brecciated/rubbly or vesicular; (2) a massive and more homogeneous flow core; and (3) transition zones between the flow top and flow cores (Wilmoth and Walker 1993; Self et al. 1996; Waagstein 1998; Passey and Bell 2007). Depending on whether the crust of the lava flow is brecciated or vesicular, the lobes can be categorized as either ‘a’ā, pāhoehoe or transitional between the two (Dutton 1884; Macdonald 1953; Wentworth and Macdonald 1953). Further subclassifications of the transitional lava facies include spiny pāhoehoe (Rowland and Walker 1987), slab pāhoehoe (Wentworth and Macdonald 1953) and rubbly pāhoehoe (Keszthelyi 2000; Keszthelyi and Thordarson 2000), while the classic pāhoehoe lava can be divided into S- and P-type pāhoehoe (Walker 1989), dense blue glassy pāhoehoe (Hon et al. 1994) and shelly pāhoehoe (Swanson 1973). The best reservoir properties of the FIBG are more likely to be present in brecciated flow crusts than in vesicular flow crusts as the pore network is usually more connected (Couves 2015; Rosenqvist et al. 2023). Flow cores are likely to display caprock properties, as are the clay-rich volcaniclastic rocks (Rosenqvist et al. 2023). However, the suitability of a layer as either a reservoir or seal for CO₂ storage depends on the layer morphologies, associated facies and internal architecture (Anthonsen et al. 2014).

The reservoir properties and reactivity of the Malinstindur and Enni formations have previously been characterized in Rosenqvist et al. (2023). These formations represent the stratigraphic top of the FIBG and are therefore expected to display less alteration and to be the best onshore analogue to the top of the NAIP basalts where offshore CO₂ storage may be attempted in the future (Neuhoff et al. 1999; Walton and Schiffman 2003; Planke et al. 2021). The top of the NAIP was drilled on the Vøring and Møre volcanic margins in the NE Atlantic during IODP Expedition 396 in the autumn of 2021 (Planke et al. 2023), and understanding the 3D morphologies of these basalt reservoirs will be important when assessing the potential for offshore CO₂ storage in the area.

Previous studies have investigated the geometry of the Faroe Islands lava successions, focusing on lava flow morphologies and geometries, and inter-lava sediment reservoir architectures (Vosgerau et al. 2016; Jolley et al. 2022), while photogrammetry studies from volcanic outcrop examples are increasingly being utilized to better understand subsurface volcanic sequences (Greenfield et al. 2020; Famelli et al. 2021; Buckley et al. 2022). This paper builds on these previous works, focusing on a multiscale characterization of the internal architecture of the FIBG sequences that is used to create a conceptual basalt reservoir model of volcanic margin prospects in the NAIP relevant for reservoir modelling. Reservoir models for injection scenarios in lava sequences have been published previously (Pollyea and Fairley 2012; Jayne et al. 2019), but have usually been restricted to layer cake stratigraphy typical of relatively simple lava flow sequences typical of only certain portions of LIP sequences. This study aims at highlighting the importance of understanding the internal architecture of volcanic reservoirs at multiple scales for modelling, monitoring and prospecting.

There are seven different geological formations making up the >6.6 km thick FIBG (Fig. 1b) where four are dominantly composed of basalt lava flows and three formations represent local hiatuses or slowdowns in lava emplacement and are sedimentary in nature (Waagstein 1988; Passey and Jolley 2008; Jolley et al. 2012, 2022). The lower c. 3.4 km of the FIBG is only accessed through the Lopra-1/1A borehole, and the deepest part of the borehole reaches the Lopra Formation (c. 1.1 km of drilled thickness; Fig. 1b; Hald and Waagstein 1984; Ellis et al. 2002; Chalmers and Waagstein 2006; Passey and Jolley 2008). This formation is mainly composed of hyaloclastites, volcaniclastic sandstones and sills (Hald and Waagstein 1984; Ellis et al. 2002; Chalmers and Waagstein 2006). Above the Lopra Formation, the 3.25 km thick Beinisvørð Formation (Fig. 1b) comprises thick (average c. 20 m) simple lava flow facies with thin red volcaniclastic rocks (red beds) between (Hald and Waagstein 1984; Passey and Bell 2007). However, recent studies have shown the presence of compound flow fields in the Beinisvørð Formation, later covered by distally erupted simple flows (Jolley et al. 2022). Resting above the Beinisvørð Formation is the <15 m thick Prestfjall Formation (Fig. 1b) consisting of coal-bearing lacustrine sediments and pyroclastic deposits and above this the <50 m thick Hvannhagi Formation (Fig. 1b) composed primarily of debris and lahar deposits associated with low-angle shield volcanoes (Rasmussen and Noe-Nygaard 1970; Passey 2004; Jolley et al. 2022).

The focus of this study, however, is on the uppermost basalt formations: the Malinstindur and Enni formations separated by the Sneis Formation, where the lava flow reservoir properties have previously been investigated (Fig. 1b; Couves 2015; Rosenqvist et al. 2023), while the reservoir properties of the inter-lava volcaniclastic rocks have also been investigated by Ólavsdóttir et al. (2015). The Malinstindur Formation overlies the Hvannhagi Formation and is dominated by compound-braided lava flows interbedded by thin red beds (Rasmussen and Noe-Nygaard 1970; Waagstein 1988; Passey and Bell 2007; Passey and Jolley 2008). However, recent studies show how the deposition of the Prestfjall and Hvannhagi formations is complex and controlled by the volcanic landscape, and to the north, compound lava flows similar in character to the Malinstindur Formation can be found in topographic lows below the Hvannhagi Formation (Jolley et al. 2022). A regional unconformity marks the boundary between the Malinstindur Formation and the overlying Sneis Formation (Passey and Jolley 2008). The <30 m thick Sneis Formation (Fig. 1b) comprises volcaniclastic sandstone through to conglomerate (Passey and Bell 2007). The >900 m thick Enni Formation overlies the Sneis Formation (Fig. 1b; Noe-Nygaard and Rasmussen 1968; Rasmussen and Noe-Nygaard 1970) and comprises overlapping lava fields of compound-braided and simple lava flows interlayered with thin volcaniclastic red beds (Passey and Bell 2007; Passey and Jolley 2008).

From studies of secondary zeolite minerals precipitated from hydrothermal fluids circulating through the volcanic sequences, it is estimated that c. 1 km of overburden has been eroded from above the exposed Enni Formation (Waagstein et al. 2002; Jørgensen 2006). Uplift and doming in the late Eocene to early Oligocene lead to general tilting of the FIBG strata towards the SE (Boldreel and Andersen 1993, 1994). This means that the stratigraphically highest formations are generally found on the central eastern to northeastern islands (Passey and Jolley 2008). In addition, the magmatism led to the emplacement of intrusions (i.e. large saucer-shaped sills), and several periods of syn- and post-magmatic deformation led to the formation and reactivation of large faults and joints (Geoffroy et al. 1994; Hansen et al. 2011; Walker et al. 2011a, b, 2012).

The landscape on the Faroe Islands is characterized by a topography consisting of steep mountainsides and sea cliffs. This produces ideal locations for collecting large-scale photogrammetry and tracing lava flows over kilometre-long distances. In addition, the drilling of several tunnels in the archipelago has led to a large number of borehole core intervals being recovered. This provides a unique combination of data that can be used for the multiscale characterization of volcanic margin basalt sequences of the NAIP.

Fieldwork was conducted on the Faroe Islands during the summer of 2020 and 2022, to collect photos, field logs and field observations. The photogrammetry data in this study consists of six models: four overview models and two close-up models (Fig. 1b; Table 1). The overview models from Svínoy, Viðoy and Sandoy (Fig. 1b; Table 1) were used to trace geological units and interpret stratigraphy, lava flow morphologies, facies associations and bedding orientations. The models from Svínoy and Sandoy (Fig. 1b) were provided by the Geological Survey of Denmark and Greenland (GEUS) and were collected by helicopter in 2010 (Vosgerau et al. 2016). The photos for the overview photogrammetry model from Viðoy were collected using a combination of digital camera (ground-based) and drone photos. A higher-resolution photogrammetry model was created for the Viðareiði outcrop on Viðoy (Fig. 1b). In addition, a higher-resolution model from the SE tip of Svínoy was provided by GEUS, complimenting the Svínoy overview model (Fig. 1b). These two close-up models were used to interpret the internal architectures and lobe morphologies of simple and compound-braided lava flow facies. A week was also spent at the Jarðfeingi core store making detailed lithological logs of three drill cores: the Viðareiði tunnel core (ViTu-1) from Viðoy, the Sandoy tunnel core (ST-7) from Sandoy and the Dalur tunnel core (Dalur-2) from Sandoy (Fig. 1b; Table 1). The core logs were used to study the internal lobe architectures in the lava sequences and compare this to the photogrammetry models.

The photos collected in the field were processed in AgiSoft Metashape Professional (AgiSoft 2023) to create the photogrammetry models. Details about the software versions and exact processing parameters can be found in the Supplementary material. The drone images contain Global Positioning System (GPS) coordinates, while the ground-based images used in the Viðoy overview model were referenced using ground control points (GCPs). The coordinates for the GCPs were taken from orthophotos (The Faroese Environment Agency 2009) at sea-level. After aligning the photos, the ‘reduced overlap’ function in AgiSoft was used to remove redundant images. When the photos were correctly aligned, a dense point cloud was created and used to build a 3D mesh and texture was added. Before the models were exported to the interpretation software, the background and redundant bits were cropped out. Position control was also performed by matching the models with the background map (satellite pictures) in AgiSoft.

The models from Svínoy and Sandoy, provided by GEUS, were generated from overlapping images acquired from a helicopter in 2010 using a Canon EOS-1Ds Mark III digital camera with a 36 × 24 mm CMOS sensor with 21 megapixels and a 35 mm Canon lens. See Vosgerau et al. (2016) for data acquisition details. The camera system (camera + lens) was calibrated before the survey and was kept locked with the focus set at infinity using duct tape. The GEUS models were prepared following the outline of Sørensen and Dueholm (2018). The relative connectivity between the images was established using the AgiSoft software Metashape (AgiSoft 2023) for tie-point generation. Subsequent triangulation or bundle adjustment of the images was then undertaken in the Anchor Lab software 3D Stereo Blend (Anchor Lab 2023) using a combination of the tie-points, GPS data acquired with the images, sea-level levering points and control points taken from orthophotos (The Faroese Environment Agency 2009). The orientation of the images (results of the triangulations) was exported from 3D Stereo Blend and used to generate 3D mesh models of the Svínoy and Sandoy cliffs using nFrames software SURE (nFrames 2023).

LIME Virtual Outcrop Geology (Buckley et al. 2019) was used to interpret the models. When interpreting facies and lithologies in the models, a simplified classification method based on macroscopic features, described by Passey and Jolley (2008), was used. Only parts of the Sandoy large-scale models and the Svínoy close-up models were interpreted, while the full models were interpreted for the Svínoy and Viðoy large-scale models and the Viðareiði close-up model. However, reservoir characterization based solely on photogrammetry models has limitations and many fine-scale and textural/mineralogical details cannot be fully constrained by visual models because of resolution. These details may be accessed in field outcrops, but accessible exposures can often suffer from extensive surface weathering. An alternative is to utilize cores collected from the subsurface, which are unaffected by recent surface weathering.

Three onshore cores, drilled through the upper stratigraphy (Malinstindur and Enni formations), were identified as relevant examples for investigating the subsurface character of the photogrammetry models used in this study. Graphic facies logs were generated for the chosen cores including analyses of: (1) facies of the individual units, classifying them as either sedimentary or volcanic with volcanic subclassifications of (a) simple or compound-braided flow lobes, and (b) pāhoehoe, ‘a’ā or transitional lava facies; (2) intra-facies of the lava flow lobes, including four possible classifications: flow top, transition zone, flow core and flow base with differentiation of brecciated or vesicular type crusts; (3) textural features, including grain size, phenocrysts, colour, vesicle distribution, fracture distribution and fracture fill; (4) visible alteration of the groundmass of the basalt, classified from largely unaltered (1), slightly altered (2), moderately altered (3), to deeply altered (4); and (5) primary (original) vesicle porosity (the percentage of vesicles) in the basalts estimated based on visual inspection. Thin-sections were collected for nine samples to aid the textural and facies interpretations of the cores.

After geological interpretation of the photogrammetry models and the core logging, flow morphology parameters such as lobe thickness were measured. To determine thickness distributions, eight logs were generated from the four large-scale photogrammetry models. Only the units fully exposed in cliffs (logs A, B, C, G and H) were used to measure the unit thicknesses and distributions in the photogrammetry models. Because of model resolution and outcrop weathering conditions, only the Viðoy and Svínoy overview models were used to measure unit extents. Compound lobe thicknesses were measured for one log in the Svínoy close-up model and in the three logged cores. The lateral extents of the compound flow lobes in the outcrops were measured for the Svínoy close-up model. The layer orientations (bedding) were measured in all four photogrammetry models. Orientations of major lineaments were also measured for the large-scale photogrammetry models. The thicknesses of flow tops, flow cores and flow bases were measured from the core logs. All measurements included in the statistical analysis are listed in Table 2.

Around 150 m of the upper part of the Malinstindur Formation and the lower c. 450 m of the Enni Formation are exposed within the photogrammetry models (Figs 2 & 3), while the studied cores cover intervals within the upper c. 120 m of the Malinstindur Formation and the lower c. 800 m of the Enni Formation (Fig. 4). The part of the Malinstindur Formation encountered in the Viðoy overview model is mainly covered in scree and vegetation. Field observations confirmed that the formation here contained compound-braided lava flow facies separated by thin volcaniclastic sandstones (red beds). The same can mostly be seen for the log from the Viðoy core (ViTu-1; Fig. 4). However, some tabular simple lava flows up to c. 10 m thick are identified in the Malinstindur Formation in the ViTu-1 core (Figs 3 & 4). Distinguishing thin simple lava flows (typical of sheet lobe flows) from thick compound flow lobes (typical of hummocky pāhoehoe) can be challenging in core and partially covered outcrops as they form a continuum and the lateral extent is unknown (Self et al. 2021). In general, a compound designation was given for the cores to packages of multiple flow lobes with individual flow lobes <5 m and without a clearly defined inflated flow core or asymmetrical distribution of vesicles typical of simple flows (Self et al. 2021).

The Enni Formation comprises both compound-braided and simple lava flows (Fig. 2). The simple flows appear both as individual flows in between compound-braided lava flows and in packages of up to c. 70 m in thickness that extend throughout the Viðoy and Svínoy overview models (>6.6 km; Fig. 2a, b). Volcaniclastic sandstones are also present between the lava flows. These are not observed in the Viðoy overview model owing to model resolution and vegetation cover. In the Enni Formation intervals in the photogrammetry models and core logs, c. 45% of the thickness volume comprises simple lava flows, while c. 53% is classified as compound-braided lava flows and c. 2% is interpreted as volcaniclastic rocks. In addition, the presence of one ponded flow and a small eroded lava channel are observed in the Enni Formation in the Viðoy and southern Sandoy models, respectively (Fig. 2b, d). The Sneis Formation is present only as thin (<3 m) reddened volcaniclastic rocks in the ViTu-1 and ST-7 cores (Figs 3 & 4).

The 41 simple flows that lie completely within the range of the Viðoy and Svínoy overview models (Fig. 2a, b), meaning that both natural flow terminations can be observed in the models, were used to measure the lateral extent of these lava flow facies. The measured simple flows are all in the Enni Formation and have an average observed lateral extent in the outcrops of c. 1000 m (Fig. 5a; Table 3) with measurements ranging from c. 130 to c. 4600 m (Fig. 5c; Table 3). The extent of the compound-braided lava flows in the outcrops is difficult to assess as all of these flows continue out of the photogrammetry models, and the compound nature of the flows make them difficult to separate from each other in eroded hillsides. They were therefore not possible to measure. However, together they make up compound flow packages (consisting of several compound flows) >100 m thick that continue for >6.6 km, remaining remarkably consistent in thickness across this distance (Fig. 2a, b). The volcaniclastic interbeds continue throughout the Svínoy, Sandoy North and Sandoy South models (>5.7 km; Figs 2a & 3). In these models the sedimentary units are also seen to drape onto each other, merging several thin beds into thicker packages as is also illustrated in Vosgerau et al. (2016). They may not be visible locally owing to a combination of vegetation cover or low photogrammetric resolution. Alternatively, they may be absent owing to local palaeotopography or loading/erosion from the overlying lava flow.

Lava flows and volcaniclastic rocks show a generally uniform thickness distribution with a pinching out in thickness towards their termination (Fig. 2). For the volcaniclastic beds, local thickness variations tend to be on the tens of centimetres scale while metre-scale variations can occur for the lava flows (Fig. 2c, d). The simple lava flow thicknesses, determined from the photogrammetry logs (A–H; Fig. 3) and the core logs, show a similar average thickness of c. 12 m (Table 3; n = 40 for the photogrammetry and n = 19 for the cores). The minimum and maximum thicknesses of the simple lava flows from the photogrammetry logs and cores range between c. 5 and c. 40 m (Fig. 5b; Table 3). The thicknesses of the compound-braided flows, determined from logs A, B, C, G and H (Fig. 3) in the photogrammetry models, is on average c. 25 m (Table 3; n = 20). The thicknesses range between c. 7 and c. 55 m (Fig. 5b; Table 3). In the core logs (ViTu-1 and ST-7) the compound flows have an average thickness of 15 m (n = 6) and the thicknesses range between 6 and 30 m (Fig. 5b; Table 3). The compound flow packages contain c. 5 lobes up to c. 27 lobes in the studied cores. The volcaniclastic interbeds in the photogrammetry logs A, B, C, G and H (Fig. 3) have an average thickness of c. 2 m (n = 19) and the thicknesses range between 0.5 and 5 m (Fig. 5d; Table 3). Thinner volcaniclastic beds can be observed in the cores (ViTu-1 and ST-7), with an average thickness of c. 0.4 m with a standard deviation of c. 0.3 m (n = 11; Table 3). The thicknesses range between c. 0.1 and 1 m (Fig. 5b; Table 3).

The internal architecture of the compound-braided lava flows can be studied in the close-up photogrammetry models from Svínoy and from Viðareiði. The Svínoy outcrop model provides a 2D vertical cross-section through lava flows where the individual flow lobes of two compound-braided lava flows were traced (Fig. 6a). The c. 150 m long interpreted segment contains larger lobes with clusters of smaller lobes on the sides (Fig. 6a). The average lateral extent of the lobes is 26 m (n = 95), with thicknesses ranging between c. 1.9 and c. 190 m (Fig. 6c). The larger lobes often pinch and swell in thickness both where they are filling in underlying topography and where the flows seem to inflate locally creating an uneven top surface (Fig. 6a). One log (Log I) was constructed through the two compound flows in the Svínoy close-up model to measure the flow lobe thicknesses and compare to the flow lobe thicknesses measured in all the core logs (Fig. 6a). The average lobe thicknesses are 2.2 and 1.1 m with standard deviations of 1.5 and 1.0 m for the photogrammetry log (n = 18) and core logs (n = 97), respectively. While the lobe thicknesses measured in Log I range between 0.3 and 4.6 m, the thicknesses in the core logs range between <0.1 and 5 m (Fig. 6d).

The photogrammetry model from Viðareiði on Viðoy exposes a sub-horizontal transect through a compound-braided lava flow. In the horizontal direction the lobes seem to have no observable preferred direction of elongation (Fig. 6b). However, the lobes are not complete as uneven erosion may have removed parts of them. The preserved parts have mostly rounded forms with different lobe segments sticking out. The exposed parts of the lobe segments vary in extent from c. 5 cm up to 7 m (Fig. 6b).

Distinguishing between erosional surfaces and topographically undulating uneven surfaces created at the top of lava flows during emplacement can be challenging. Except for local topography, the flows and beds appear mostly sub-horizontal, and the average (true) dip measured for all the photogrammetry models is c. 4° with a mean dip direction towards the ESE of 117° with a 95% confidence interval of ±22° (Fig. 7a). Large-scale lineaments (faults or eroded dykes) can be seen cutting through the full thickness of all the photogrammetry models. A total of 20 subvertical lineaments with little to no visible displacement are observed and their orientations measured. The lineaments have an average dip of c. 80° and two main orientations, NE–SW and NW–SE (Fig. 7b). The two orientations appear in some places to form a conjugate set, e.g. in the Viðoy photogrammetry model (Fig. 2b). A reverse fault with a general NW–SE orientation and maximum displacement of 2 m can be observed in the Svínoy overview model (Fig. 2a). The fault has a dip of c. 21° and dip direction of 235° towards the SW (Fig. 7b).

The core logs from the three studied boreholes (ViTu-1, ST-7 and Dalur-2) display the complexity of the facies variations within the volcanic sequences of the Malinstindur and Enni formations (Fig. 4). The simple flows within the two formations have a classic internal architecture containing an upper and lower crust (flow top and flow base) and a massive flow interior (flow core; Fig. 4). Transition zones are sometimes present between the flow top and flow core in the thicker simple flows (Fig. 4). Within the ViTu-1 and ST-7 boreholes, simple flows with a rubble zone in both the lower and upper flow crusts can be observed (Fig. 4). The crusts show signs of autobrecciation as they contain angular clasts with similar phenocryst sizes and distributions as the flow core and ingested clasts in the flow cores and transition zones. These lava flows may be interpreted as thin crusted ‘a’ā lava facies; however, the general absence of ‘a’ā lava flows in the FIBG as described by Passey and Bell (2007) suggests that they are transitional. A significant amount of the studied simple lava flows is also seen to have a well-developed breccia in the upper crust and a vesicular crust with minor to no clasts at the base (Fig. 4). These flows are interpreted as transitional lava flows that began as pāhoehoe prior to upper surface disruption during flow. Some rubble can also be found within examples of the compound flow lobes, indicating transitional flow emplacement was not restricted to the simple lava flows.

Simple flows and compound flow lobes with an upper and lower (often pipe-vesicle-bearing) vesicular crust, separated by a low vesicularity (massive) flow interior/core, are also present in the three drilled cores (Fig. 4). These are interpreted as pāhoehoe lava facies. In addition, S-type (spongy) pāhoehoe flow lobes are present within the compound-braided flows (Fig. 4). These lobes are often thin and vesicular throughout without a clear internal zoning except for a slightly higher vesicularity at the centre of the lobe (Fig. 4). Finally, the sedimentary interbeds are mostly fine- to medium-grained sandstones with a high clay proportion. In places, ranging clasts from underlying lava crusts are integrated in the sandstones (Fig. 4).

The thickness of the flow tops (including the transition zones), flow cores and flow bases of the simple lava flows (>5 m thick) in the three boreholes were measured, as were the thicknesses of the internal zones in the flow lobes that were <5 m thick (generally thought to be compound) that also contained a clear flow crust and flow core. The thinnest flow lobe observed with this internal division was 0.85 m thick. The thickness of the simple flow tops ranges between 1.3 and 11.4 m with an average thickness of 3.3 m (n = 18). The flow tops of the <5 m thick lobes range in thickness between 0.1 and 2.3 m with an average of 1.1 m (n = 20). The thickness of the simple flow cores ranges between 0.4 and 25.8 m with an average thickness of 6.9 m (n = 18). The flow cores of the <5 m thick lobes range in thickness between 0.4 and 2.8 m with an average of 1.0 m (n = 20). The thicknesses of the flow bases of the simple flows range between 0.1 and 1.5 m with an average thickness of 0.6 m (n = 21). The flow bases of the <5 m thick lobes range between 0.1 and 0.9 m with an average thickness of 0.4 m (n = 20). The S-type pāhoehoe flow lobes and thinnest compound flow lobes are not included in these calculations as they do not contain the three-part division in internal architecture. In the drill cores, however, they compose components of packages that are more than 20 m thick (Fig. 4). The thickest rubbly flow intervals are found in the flow tops of the simple flows and reach up to c. 12 m in thickness (Figs 4 & 8).

Most of the vesicles and fractures in the basalts in the cores are filled with secondary minerals; however, some open porosity does occur. The secondary mineral infill comprises mostly clay and zeolites whereas the amount of carbonates is minor. Within the individual drill cores and within individual flows or lobes the secondary infill differs, with some zones containing mostly zeolites, some mostly clay and others open vesicles with clay or zeolite coatings. A clear alteration/weathering profile is seen in the groundmass of most of the lava flows with the highest degree of alteration occurring at the lobe margins and diminishing towards the lobe interior (Fig. 4). The flow crusts generally show the highest degree of alteration, varying between being slightly altered (2) to moderately altered (3) and deeply altered (4). The flow cores mostly vary between being largely unaltered to slightly altered in their groundmass, with the centre of the thickest flow cores, with the lowest original porosity, often being the least altered (Fig. 4).

The primary vesicle porosity estimated from the drill cores generally ranges from c. 0% to c. 50%, with the lowest estimates being in the massive flow interiors and the highest estimates being for the flow crusts and spongey S-type flow lobes (Fig. 4). The vesicle primary porosity of the flow crusts (both vesicular and rubbly) generally ranges between c. 5% and up to 50%. Some of the most densely vesicular spongy flow lobes may also have primary porosities above 50%. The flow cores generally show no primary vesicle porosity; however, vesicular zones or bands locally increase the original porosity to c. 5–7%, often connected to the presence of segregation veins created by residual fluids during cooling (Fig. 4; Kuno 1965). The most vesicular zones, flow crusts and spongey lobes, are generally where the smallest vesicles are present, while larger vuggy vesicles are sparsely distributed in the flow cores (Fig. 4).

In addition to the vesicles, primary fractures coated or filled with secondary minerals are present in the drill cores (Fig. 4). The fractures are roughly evenly distributed throughout the cores, with more fractures occurring in or near the brecciated zones or in the flow interior of the thickest flow lobes (simple flows). The flow cores of the simple flows contain more of the large subvertical fractures; however, zones of heavy horizontal fracturing are also present (Fig. 4). Thin, clay-filled fractures are common in the transitional lava flows. The presence of fractures may contribute to a higher total primary porosity.

Studying the morphologies/geometries, compartmentalization and connectivity of the reservoir and sealing units in onshore outcrops is essential to understanding how and where CO₂ may migrate and react in an offshore basalt reservoir (e.g. as done for inter-lava clastic beds by Vosgerau et al. 2016). Most significantly, the simple and compound-braided lava flows display differences in geometries and internal architecture, presenting as either barriers or migration paths that control fluid flow in a reservoir setting. The individual simple flows in the photogrammetry models generally range in lateral extent from c. 130 up to c. 3000 m, with some extending up to 4600 m. In comparison, low-viscosity lava flow lengths measured from the offshore Rosebank field (Faroe–Shetland Basin) in the NAIP are seen to range between c. 1 and c. 22 km but are thought to easily extend several tens of kilometres in unrestricted flow paths (Millett et al. 2021a). Walker (1973) presented similar numbers in his measurements of 479 Quaternary to recent eruptions, indicating lava flow lengths ranging from <1 to <16 km with a median flow length of 4.1 km. This is several times longer than the outcrop measurements of simple flows from the Enni and Malinstindur formations indicate. However, it is important to consider that the measurements of lateral extent through arbitrary cross-sections (determined by the outcrop orientations) do not necessarily represent the maximal extent of the lava flows. The 3D morphology of the simple and compound-braided lava flows is likely shaped by: (1) the pre-existing topography of the lava field; (2) the lateral extent and cooling of the lava, causing breakouts and fingering of the lava flow; and (3) local drainage patterns and tectonic features (Millett et al. 2017). All of this can give the lava flows a complex 3D morphology with several different branches diverging from the main flow, as also seen for present-day lava flows, e.g. from Hawaii (e.g. Abrams et al. 1991; Hon et al. 1994), Iceland (e.g. Pedersen et al. 2022) and Italy (e.g. Tarquini and Favalli 2011; Corradino et al. 2019). This means that the individual flow segments observed and measured in the 2D photogrammetry models may be of the same lava flow event and connect in 3D. The maximal extent of the simple and compound lava flows in the studied Enni and Malinstindur formations may indeed reach tens of kilometres.

The thickness of the simple flows, measured from the photogrammetry models and drill cores, ranges between c. 5 and 40 m. The measurements are similar for both type of logs, and the large-scale photogrammetry models are good for tracing and measuring simple lava flows because of their thick tabular shape. The thicknesses of the compound lava flows are similar to that of the simple flows, from c. 5 to 50 m. However, most of the compound flows are seen to be more than c. 10 m thick. The measured flow thicknesses are similar to those measured by Hald and Waagstein (1984) and Boldreel (2006) for simple flows of the Beinisvørð Formation in the Lopra-1/1A borehole (c. 5 to 50 m thick). Furthermore, the onshore measurements from the Enni and Malinstindur formations have similar thicknesses to the offshore lava flows in the Rosebank field, with an average thickness of 11.2 m and simple flow thicknesses ranging between ≤5 and up to c. 50 m (Millett et al. 2021a). In comparison, lava flows recorded in Ocean Drilling Program (ODP) boreholes from Site 642 on the Vøring Margin, offshore Norway, show a thickness range between 0.6 and 18 m, likely comprising a combination of compound lobes and simple lava flows similar to what is seen in the studied Enni and Malinstindur formations (Planke 1994).

The compound lobes range between c. 2 and c. 190 m in lateral extent in the photogrammetry models. However, as with the extent of the lava flows, these measurements are likely not representative of the maximal extent of the compound lobes. Smaller toe lobes, seen in the drill cores, may also not be detected in the photogrammetry models because the resolution of the models is too low. In general, the compound lobe thicknesses measured in the cores documented thinner lobes than for the photogrammetry Log I. Either thin lobes are not present at the location of Log I or the resolution of the photogrammetry model is too low to recognize these. The compound lobes are generally seen to range in thickness between c. 5 cm and 5 m. These thickness measurements are similar to measurements from modern-day Hawaii where the pāhoehoe lobes are described to have initial thicknesses of c. 10 to 50 cm (Hon et al. 1994; Keszthelyi and Denlinger 1996). Later, inflation of the flow lobes will cause the thicknesses to increase to c. 1 to 5 m (Hon et al. 1994).

The thickest and widest lobes measured in the photogrammetry models approach the simple flow facies in geometry. Mattox et al. (1993) divided the pāhoehoe lava flow into two different categories: large primary flows and smaller breakouts. It is likely that the largest (widest and thickest) flow lobes observed in the compound flows in the studied Enni and Malinstindur formations represent primary lava flow paths that have experienced significant inflation while the smaller lobes represent breakouts from the primary flows (Mattox et al. 1993). Looking at the compound lava flows as a system of primary flows and secondary breakouts, it is likely that at least some of the lobes are connected in 3D. The photogrammetry model from Viðareiði shows that there is no visible preferred orientation of the lobes. However, the lobes observed in this model are <7 m in extent and may only represent breakouts from larger flows or lava tubes. Furthermore, the topography of this lava field may be too variable to determine the shape and orientation of the lobes.

The volcaniclastic interbeds are present between several of the lava flows in the volcanic sequences but make up less than c. 2% of the volume in the Malinstindur and Enni formations. Even though they are common and laterally extensive (>4500 m), they are thin (<5 m thick), often on a tens of centimetre scale, as also supported by the studies of Vosgerau et al. (2016). The sediment layers may also be thin or absent owing to local topography at the time of deposition or owing to loading or erosion by succeeding lava flows, implying that at the present day, they are not necessarily fully continuous. These thin volcaniclastic rocks are also present in the offshore areas of the NAIP such as the Vøring Margin and the Rosebank field in the NE Atlantic (Planke 1994; Millett et al. 2021a; Planke et al. 2023).

The Faroese lava fields of the Malinstindur and lower Enni formations consist of a combination of P-type and S-type pāhoehoe lava, with local transitional rubbly pāhoehoe lava flows, moving towards ‘a’ā lava flow facies. These flow variations have distinct distribution and original porosity variations that will affect the character and horizontal continuity of a reservoir system. The transition of pāhoehoe to the ‘rubbly pāhoehoe’ seen in the cores is likely due to a combination of an increase in lava supply rates, a local increase in palaeoslope, rapid cooling and a local increase in strain rates (Peterson and Tilling 1980; Cashman et al. 1999; Hon et al. 2003; Duraiswami et al. 2014). This means that the presence of rubbly v. vesicular flow crusts may be very local, controlled by vent proximity and flow field distribution as characterized in the 2014 Holuhraun flow field (Voigt et al. 2021), but possibly also affected by, for example, the palaeotopography of the lava field (Duraiswami et al. 2014). Although rubbly flow crusts were observed within the drill cores in this study, field observations from the Enni and Malinstindur formations done by Passey and Bell (2007) show that the rubbly crusts are usually not extensive in outcrop.

The flow cores of the simple flows are the thickest zone encountered within the lava flows, and these flow cores reach up to c. 25 m in thickness. Plotting the flow core proportion v. total flow thickness for the simple flows reveals that the thickness of the flow core and the proportion of the flow that is massive flow interior increases with the total flow thickness (Fig. 8a). The proportion of flow core in the simple lava flows generally plots in the same area as measurements for flow core proportions from simple flows of the Rosebank field (Millett et al. 2021a) and from the Beinisvørð Formation on the Faroe Islands (Fig. 8a; Nelson et al. 2009). They also plot similarly to flow core proportions for flow lobes in the Enni and Malinstindur formations measured in the Glyvursnes-1 and Vestmanna-1 boreholes by Nelson et al. (2009) (Fig. 8a).

The flow tops of the simple flows reach up to c. 10 m, while the thicknesses of the flow bases reach up to c. 1.5 m. There is a clear trend of increasing flow top thickness with increasing flow thickness for the simple flows (Fig. 8b). However, there is no clear relationship between the flow base thickness and lobe thickness for the simple flows, meaning that the flow bases do not seem to increase in thickness together with the total flow thickness. The upper crust will grow during inflation of the lobe as new lava filled with gas bubbles is injected, and the bubbles will rise to the surface and add to the vesicular upper crust thickness, which will reflect the period of active inflation (Self et al. 1998). According to Self et al. (1998), the lower crust grows five to ten times slower than the upper crust, meaning that the basal crust is expected to remain thin during lobe inflation. As the flow core proportion is expected to increase with increasing flow thickness while the flow bases remain thin, it is likely correct to assume that the flow top thickness will grow less and less with increasing total flow thickness. An approximate, expected relationship is indicated in Figure 8b. This is supported by the estimates of Hon et al. (1994), who has quantified an exponential relationship between the top crust thickness and the time needed for it to grow.

The spongey S-type compound flow lobes are not included in these statistics as they are vesicular throughout. These spongey lobes occur in thick (>20 m) intervals suggesting that they are indeed S-type lobes emplaced by a proximal frothy lava sequence similar to present-day Hawaiian pāhoehoe flows (Wilmoth and Walker 1993). The S-type pāhoehoe lobes remain vesicular throughout and display no apparent movement of bubbles during emplacement, because the porosity reaches above a threshold (30%), constituting a high-yield strength foam that resists deformation (Wilmoth and Walker 1993).

The large variations in lobe sizes make the compound-braided flows very heterogeneous in both vertical and horizontal directions. Compartmentalization of the reservoir will present itself as several ‘isolated’ flow core lenses surrounded by flow crusts where the crusts and flow cores will display different reservoir properties. By moving, for example, 100 m in a horizontal direction in a compound-braided flow, several flow cores and flow crusts may be encountered. The connectivity of multiple flow crusts could, however, be better for the compound flows than simple flows, as several flow crusts may be in contact with each other. In addition, crust-to-core ratios are generally higher for thinner lobes, suggesting greater net-to-gross ratios that could lead to a larger primary storage potential of this facies (Nelson et al. 2009; Yi et al. 2016). For simple lava flows, the same flow core and crust is likely to continue for up to several kilometres. Vertical compartmentalization will be present in the flow crusts v. flow cores, while horizontal variations in the crust's transmissivity could occur with transitions between the predominantly rubbly and vesicular crusts. Connectivity of the flow crusts in simple flows may occur through adjacent crusts or through fracture networks.

Large heterogeneities in siliciclastic CO₂ reservoirs have previously been considered to negatively impact injectivity and storage capacity (Anthonsen et al. 2014). However, larger heterogeneities in the reservoir rock, present as thin clay layers between the reservoir sand, have been shown to spread the CO₂ plume more efficiently after injection (Sundal et al. 2013, 2015). CO₂ sequestration in basalts may, however, present new conditions. Pollyea and Fairley (2012) did, however, show in their simulations of CO₂ injection into basalt reservoirs that the spatial distribution of heterogeneous permeability structures will affect the formation injectivity significantly. They concluded that the successful field implementation of CO₂ sequestration in fractured, low volume basalt will be largely dependent on the connectivity of the high-permeability zones (e.g. rubble zones) therein (Pollyea and Fairley 2012).

The flow crusts of the lava flows and the spongey lobes are expected to hold the largest reservoir potential as the primary porosities are estimated to be highest here (supported by Rosenqvist et al. 2023). The crusts often contain primary fractures and primary vesicle porosities above c. 15%, used as a porosity threshold for basalt CO₂ reservoirs in, for example, McGrail et al. (2006) and Reidel et al. (2002). The rubbly crusts will likely be more fractured and have a more connected pore network than the vesicular crusts, and therefore present the most originally porous and permeable units (Rosenqvist et al. 2023). In contrast, the estimates of primary vesicle porosity in the flow cores suggest a largely impermeable caprock lithology (supported by Rosenqvist et al. 2023), and flow cores are known to act as aquitards in areas such as the Columbia River Basalt Province, USA (McGrail et al. 2011). The flow crusts of the lava flows are generally seen to be more weathered than the flow cores, which can both affect porosity by pore clogging and may also affect the reactivity of the rock in contact with dissolved CO₂ (Gudbrandsson et al. 2011).

McGrail et al. (2006) and Reidel et al. (2002) used a thickness of 10 m with a porosity above 15% as a threshold for potential reservoirs in basalt sequences. However, vesicular crusts are generally expected to reach above c. 30% porosity for the rock to be permeable, without accounting for fracturing (Rintoul and Torquato 1997). The thickest vesicular crusts are expected for the flow tops of the thickest simple flows. Brecciated flow tops of simple flows are seen to reach above 10 m in thickness and 15% in primary porosity, but they are rare, and the high occurrence of transitional lava flows makes their lateral continuity uncertain. However, connectivity of the flow crust of individual simple lava flows, similar to what was seen for the Wallula CO₂ reservoir in the Columbia River Basalt Group, may lead to a larger reservoir volume where the 10 m thickness threshold is no longer relevant (McGrail et al. 2011).

The compound flows contain very thin flow crusts; however, the high connectivity of the crusts of different lobes could lead to a large combined reservoir thickness. The compound-braided lava flows are observed to make up packages up to >100 m in thickness in the studied formations. In addition, the intervals of spongey pāhoehoe flow lobes make up c. 20 m thick packages of generally high primary porosity. The primary porosity of the vesicular lobes does, however, rarely reach above 30%, and it is likely that these pāhoehoe lava reservoirs would have to be fractured to be permeable, as seen in the CarbFix reservoirs on Iceland (Ratouis et al. 2022). The large-scale fractures seen in the photogrammetry models and the smaller scale fractures (likely from cooling and local strain; e.g. Self et al. 1996) may contribute to this fluid flow (Fisher 1998). However, an in-depth study of the fracture network on these volcanic sequences would have to be performed to assess their full impact on the reservoir system.

For siliciclastic CO₂ reservoirs, a caprock should generally have a thickness of >50 m (Anthonsen et al. 2014). Both flow cores and the volcaniclastic sandstones of the studied Malinstindur and Enni formations were determined by Rosenqvist et al. (2023) to hold caprock properties. The volcaniclastic interbeds have a large horizontal extent, but their low and variable thickness (<5 m) makes them poor seals. They may, however, baffle the fluid migration locally. The flow cores of the simple lava flows typically only reach up to 25 m in thickness. A package of several thick simple flows, seen to be present in the studied sequences, could, however, work as a stacked caprock–reservoir sequence with primary and secondary seals and reservoirs as seen in, for example, the Wallula injection site (McGrail et al. 2011). For an offshore reservoir, a sedimentary seal overlying the volcanic packages may be the best alternative (Planke et al. 2021). However, the CarbFix method of injecting CO₂ dissolved in water removes the need for a caprock as the fluid is no longer buoyant and represents an alternative storage method used at the CarbFix injection sites (Snæbjörnsdóttir et al. 2014; Sigfusson et al. 2015; Gunnarsson et al. 2018).

The results of this study of the Faroe Islands basalts are relevant for modelling CO₂ injection scenarios for basaltic lava flow dominated reservoirs. Most models of CO₂ sequestration in basalt reservoirs focus on fractured largely homogeneous reservoirs consisting of layers with a high lateral continuity and with a large grid size (Aradóttir et al. 2012; Jayne et al. 2019; Ratouis et al. 2022). However, Pollyea and Fairley (2012) and Pollyea et al. (2014) have demonstrated the importance of heterogeneity in fractured, low volume basalt reservoirs on parameters such as injectivity and confinement. In addition, Jayne et al. (2019) showed that permeability uncertainty on a reservoir-scale considerably affects the accumulation and distribution of injected CO₂. Therefore, combining the site-specific conceptual reservoir models with knowledge about the internal architecture variations and uncertainties will be important in optimizing reservoir modelling of subsurface injection sites in the future.

A conceptual reservoir model is presented here (Fig. 9) to summarize the observations and measurements of the internal geometry of basalt reservoirs of the NAIP. The model is based on the studied Enni Formation in the FIBG but applied to an idealized offshore storage scenario where the pores have not been as severely clogged by secondary zeolite and clay minerals. In the Enni Formation, the volume ratio between simple and compound-braided lava flows seems to be close to one, with both facies appearing in alternating sub-horizontal packages c. 70–80 m thick. For modelling scenarios, this will be important. Each package will display different reservoir properties because of the different internal geometry. The alternation in facies will promote vertical reservoir compartmentalization, and fluids will likely flow and spread differently in the different layers. The important fluid pathways will be through connected, permeable flow crusts and through open fracture systems. Thick flow cores may work as seals for CO₂ migration, while thin flow cores and volcaniclastic rocks are likely to work as baffles for CO₂ migration.

This study highlights the importance of using several types of data to assess the internal architecture of volcanic reservoirs. While the drill cores give vital information on internal flow architectures and reservoir properties in 1D, the photogrammetry models are important for assessing layer geometries, facies associations, layer orientations and identifying large-scale fractures in 2D (to 3D). In addition, the use of modern-day analogues of active lava fields are important to determine the 3D morphology of the reservoir system of these extremely heterogeneous reservoirs. By using these different data types and characterizing these reservoirs in high detail it also becomes evident what will be below seismic resolution in an offshore reservoir or what may be the 3D geometry of the layers encountered in an offshore borehole. The results of this study may be used to predict the lateral extent of different lava reservoir types encountered in an offshore well based on information about lava types and thicknesses. This is important for identifying potential reservoirs and seals, but also for monitoring injected fluids and estimating storage capacity in a CO₂ storage scenario. For modelling of fluid flow in volcanic reservoirs, this study will also be important as it provides an understanding of which features will be below the grid-resolution of the model.

The studied Malinstindur and Enni formations are examples of subaerially erupted lava flows of the NAIP and are analogous to offshore sequences characterized as ‘landward flows’ (Ellis et al. 2002; Horni et al. 2017). These formations consist of a combination of compound-braided lava flows and simple flows with various thicknesses. Other endmember formations do, however, occur within the FIBG and may be expected to occur in other places within the NAIP. One example is the Beinisvørð Formation that comprises thick (average 20 m thick) simple lava flows or the Prestfjall and Hvannhagi formations that make up a >50 m thick sedimentary package within the FIBG (Rasmussen and Noe-Nygaard 1970; Hald and Waagstein 1984; Jolley et al. 2022). The studied Malinstindur and Enni formations do, however, resemble the lava facies encountered in the landward flows on the Vøring Margin offshore Norway, a place that is currently being evaluated for its CO₂ storage possibilities (Planke 1994; Planke et al. 2023). The volcanic flows observed in the 642E borehole on the Vøring Margin are 0.6–18.5 m thick, containing vesicular or brecciated flow crusts, with thin (<1 m) intravolcanic volcaniclastic sandstones (Planke 1994). Furthermore, while the volcanic flows on the Faroe Islands contained mainly vesicles clogged by secondary minerals, the upper c. 1000 m of Hole 642E displayed mostly open porosity that could potentially constitute porous and permeable reservoirs (Planke 1994).

Sequestration in siliciclastic reservoirs below or within the volcanic sequences of the NAIP may also present a possibility for CO₂ storage. The discovery of hydrocarbon reservoirs in intra- and sub-basalt sediments of the Rosebank field in the Faroe–Shetland Basin shows that such reservoirs might be present in the NAIP (Duncan et al. 2009; Schofield and Jolley 2013; Hardman et al. 2019). The facies assemblage in the Rosebank field is similar to that observed in the studied upper part of the FIBG (Millett et al. 2021a) but has not yet been appraised from a CO₂ injection perspective. Lava flows are known to represent effective reservoirs in many scenarios globally (Schutter 2003), and recent studies on lava-hosted petroleum reservoirs have highlighted the strong facies control on reservoir potential (Millett et al. 2021b; Marins et al. 2022). To date, critical details on reservoir properties from the NAIP offshore lava sequences are lacking. Indirect evidence for flow potential within the lavas exists in the form of a permeability estimate from Hole 642E based on temperature probe measurements (Harris and Higgins 2008), and massive drilling fluid losses in 217/15-1z linked to evidence for simple lava flow tops and drill cuttings indicating open vesicular porosity (Millett et al. 2016). Clearly, improved constraints on the nature, porosity and permeability of rocks from the subsurface volcanic rocks of the province, such as those collected in the recent drilling of IODP Expedition 396 (Planke et al. 2023), are required in order to better constrain injection scenarios offshore and will form the focus of future work. In addition to offshore potential, Eidesgaard (2021) also suggested that the FIBG may present possible CO₂-storage sites in Faroese areas. Finally, understanding the internal architecture of these basalt sequences is not only important to identify potential reservoirs for CO₂ storage, but also for understanding groundwater flow and petroleum migration in volcanic margin basalts.

In this study, the internal morphologies and architecture of the Malinstindur and Enni formations on the Faroe Islands have been characterized based on field observations, interpretations of photogrammetry models and logging of onshore drill cores. The results were used to characterize potential reservoirs and seals for CO₂ sequestration in volcanic margin basalt sequences of the NAIP.

• The Enni and Malinstindur formations of the Faroe Islands comprise a mixture of subaerially erupted simple and compound-braided lava flows with thin (<5 m thick) volcaniclastic sandstones appearing between the layers. The studied Malinstindur Formation consists mainly of compound-braided flows with some isolated simple flows occurring, while simple flows become more prominent in the Enni Formation, making up c. 50% of the studied formation thickness. The same flow facies often appear in stacks of more than 50 m in thickness.

• The simple lava flows of the Enni Formation extend from c. 130 to >4600 m in the outcrops, with thicknesses of the flows ranging between c. 5 and 40 m. Based on the low-viscosity basaltic compositions of the lava flows, their consistent thicknesses over these distances and literature examples, it appears likely that the maximum extent of the lava flows may reach several tens of kilometres with unrestricted flow paths. The low-porosity, low-permeability flow cores of the simple flows may reach up to c. 25 m in thickness, which is too thin to present a potential reservoir seal on its own. However, the presence of stacked simple flow cores can represent primary and secondary seals, cumulatively creating effective composite seals.

• The originally high-porosity and potentially high-permeability flow crusts of lava flows may present potential reservoirs for CO₂ storage. The thickest flow crusts are found for the flow tops of simple lava flows and reach above 10 m in thickness. However, connectivity between flow crusts in simple flow packages can make up stacked reservoirs with larger combined thicknesses.

• The thickness of the compound flows ranges between 5 and 50 m. The internal architecture of the compound flows is more heterogeneous, consisting of several connected flow lobes. The compound flow lobes in the outcrops range in lateral extent between 2 and c. 180 m, with measured thicknesses from the photogrammetry models and drill cores ranging between c. 5 cm and 5 m.

• The flow crusts of the individual compound lobes are generally seen to be less than 2 m thick. However, the connectivity of the flow crusts in unaltered compound-braided flows is likely high and they may make up thick (>50 m) combined reservoirs. In addition, compound flows consisting of S-type spongey pāhoehoe lava may contain >20 m thick vesicular intervals that can represent potential reservoirs. The vesicular reservoirs may, however, have to be fractured for the rocks to be adequately permeable.

• The low-permeability volcaniclastic beds are generally seen to be too thin (<5 m thick) to be used as effective seals by themselves but may work as baffles for CO₂ migration, promoting plume spreading, and/or contribute to composite sealing where thicker. In addition, the presence of sedimentary inter- or sub-basalt reservoirs offshore should be appraised as potential CO₂ storage sites in the future.

• The use of onshore analogues for multiscale and multidisciplinary analyses of the internal architecture of volcanic margin basalt sequences is essential for the understanding and 3D control of offshore volcanic reservoirs. In the future, further research on the multiscale fluid pathways in basalts will be essential for the advancement of the basalt carbon sequestration technology.

We would like to thank the Faroese Geological Survey (Jarðfeingi) for their help and guidance in planning the fieldwork and with preparing the cores and samples. In particular, the authors kindly thank Jana Ólavsdóttir, Óluva R. Eidesgaard and Turid Hátún Madsen for their insight and assistance. In addition, we would like to thank Max Meakins for his help in acquiring images for photogrammetry models, and Hans Jørgen Kjøll, Olivier Galland and François Renard for their help with the fieldwork. Thanks to the reviewers Aimee Whatling and David Jolley for improving this work with their feedback and suggestions, and thank you to the corresponding editor Ben Kilhams and the rest of the GSL team for organizing this special issue.

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.

MPR: data curation (equal), formal analysis (lead), funding acquisition (supporting), investigation (lead), methodology (equal), project administration (lead), validation (equal), visualization (lead), writing – original draft (lead); JMM: conceptualization (equal), investigation (supporting), methodology (supporting), supervision (supporting), validation (supporting), writing – review & editing (equal); SP: conceptualization (equal), methodology (supporting), supervision (equal), validation (supporting), writing – review & editing (supporting); RMJ: investigation (supporting), methodology (supporting), writing – review & editing (supporting); SRP: data curation (supporting), funding acquisition (supporting), validation (supporting), writing – review & editing (supporting); EVS: data curation (lead), funding acquisition (supporting), writing – review & editing (supporting); HV: data curation (supporting), funding acquisition (supporting), writing – original draft (supporting); BJ: funding acquisition (supporting), project administration (supporting), supervision (equal), writing – review & editing (supporting).

This project has received funding from the Faculty of Mathematics and Natural Sciences at the University of Oslo though the project CO2Basalt and CASP through the Andrew Whitham CASP Fieldwork Awards 2020. We acknowledge funding from the companies of the SINDRI Group for the collection of digital photographs for photogrammetry. S. Planke acknowledges support from the Norwegian Research Council through Centre of Excellence funding by CEED and the PALMAR project (projects no. 223272 and 336293). R. Johannesen acknowledges funding from the Faroese Research Council and Betri Stuðulsgrunnur.

The data that support the findings of this study are available from the Geological Survey of Denmark and Greenland (GEUS), but restrictions apply to the availability of these data, which were used under licence for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of GEUS.