In this work, we identified all the structures that characterise the Fremrinamar Fissure Swarm through the analysis of original photogrammetry-derived orthomosaics and DSMs, created from different sets of historical aerial photos, reaching a pixel size of 0.53–0.65 and 2.14–2.59 m/pix, respectively (Figures 2C, D), as well as by field surveys. The consistency of DSMs derived from aerial photos has been already confirmed by Gomez et al. (2015), Ishiguro et al. (2016), and Grottoli et al. (2020).

A total of 404 aerial photographs were acquired from the National Land Survey of Iceland (https://www.lmi.is/). Images covering an interval of 8 years (1983, 1990, and 1991) were captured using a camera with a focal length of 151.78 mm, at a constant flight altitude of 5,486 m a.s.l., and with an overlap of 60% along the flight path (https://www.lmi.is/is). Images were processed with Agisoft Metashape v.1.7.1 (https://www.agisoft.com/), a Structure from Motion (SfM) software that allows to obtain spatial data from digital images. The SfM technique is based on the recognition of similar features from different images, acquired with a high degree of overlap (Westoby et al., 2012; Schonberger and Frahm, 2016; Bonali et al., 2019a), generally 60%–80%. Photo alignment was performed with medium accuracy settings, while the re-elaboration of the dense point clouds was performed with medium quality and mild depth filtering (further details are in Agisoft (2023)). These attributes were chosen, after different attempts, to reach the best compromise between good-quality results and relatively short-processing times. Due to the extent of the study area (130 × 8–9 km), we created two different point clouds with different sets of photos, taken in different years. To georeference the point clouds, 545 total recognisable features, or Ground Control Points (GCPs), of known coordinates were selected within the point clouds. X,Y,Z data are from the ISV50 database, from the National Land Survey of Iceland (https://www.lmi.is/is/landupplysingar/gagnagrunnar/is-50v). We obtained two orthomosaics, with resolutions of 0.53 and 0.65 m/pixel, and two DSMs, with resolutions of 2.14 and 2.59 m/pixel, both referenced to the UTM28N-WGS84 datum. The resulting resolutions are comparable with the one of existing DSMs, such as the ArcticDEM (https://www.pgc.umn.edu/data/arcticdem/) which is derived from high-resolution satellite images, reaching a resolution of 2 m/pix, within the study area. The perspective of being able to merge the analysis of DSMs and the associated orthomosaics, led us to select them as the best tool to evaluate and classify the features that compose the Fremrinamar Fissure Swarm.

The Fremrinamar Fissure Swarm was then mapped at a 1:50,000 scale in a GIS environment (QGis v. 3.28.4), to achieve a better characterisation if compared to the current state of the art, which is a depiction at a scale of 1:250,000 up to 1:100,000 (Sæmundsson, 1977; Sæmundsson et al., 2012; Sigurgeirsson et al., 2015). The scale of the analysis was also selected following equation 13.1 from Gudmundsson (2011), which points out that when the ratio between half the strike dimension of a fracture segment and half the distance between the midpoints of collinear segments gets closer to the value 1, the segments act as a single fracture (for more details, see Sneddon and Lowengrub, 1969; Gray, 1992). This approach has been applied on key areas, where structures were identified and mapped. We observed that a classification at a scale of 1:50,000 gives similar results as the one performed with the above-mentioned criterion.

Structures composing the Fremrinamar Fissure Swarm were classified into eruptive fissures, dry fissures (here referred to as extension fractures), and normal faults. The latter were recognised through the identification of better-lit or shaded slopes on orthomosaics and shaded DSMs, also classifying them into E− and W-dipping scarps (Figures 2C, D). Eruptive fissures were identified by their proximity to volcanic deposits or through the morphometric analysis of eruptive centres (for details on the method see Tibaldi, 1995; Tibaldi and Bonali, 2017).

We proceeded with a quantitative analysis of our dataset, determining length and azimuth values for each structure, as well as the X and Y coordinates of the structures’ mean points. The previous parameters were related through the realisation of scatterplots to evaluate their variation along the entire Fremrinamar Fissure Swarm. A total of 761 normal faults were identified and characterised, also measuring their vertical offsets (Figure 3A), through the implementation of topographic profiles on photogrammetry-derived DSMs. Here, the vertical offset is given by the difference in elevation between the tangent to the top and the bottom of each scarp (e.g., Hjartardóttir and Einarsson, 2021). Fracture density and spacing were evaluated through 263 N106°E-striking transects, spaced 500 m from each other, both considering the entire dataset and one subgroup at a time (eruptive fissures, extension fractures, and normal faults). For normal faults with a length >2 km, slip profiles (or displacement-length profiles) were collected to evaluate the overall along-axis rift propagation of the Fremrinamar Fissure Swarm (details of the method in Manighetti et al. (2001); Tibaldi et al. (2020a)). These profiles are representative of the fault offset variation compared to its length. Similar analyses, performed worldwide, show that although slip profiles have great variability in their shape, they often display similar first-order asymmetrical geometry, as shown by Manighetti et al. (2001).

Measuring points were placed at a distance of 200 m from each other, along the normal fault scarps, and offset values were collected through the obtained DSMs. Slip profiles were then classified in five classes, based on a quality parameter that evaluates the consistency and linearity of a fault scarp, along its strike dimension, and the affected substrate (PRE- or POST-LGM). Results were analysed without considering fault scarps that are too eroded and thus do not allow to clearly infer a propagation direction, regardless of whether they cut through PRE- or POST-LGM units (class E). Afterwards, faults showing local erosion of their scarp, and affecting PRE- and POST-LGM units, were excluded (classes C and D, respectively) and the analysis focused only on the normal faults that do not show any interruption or clear local erosion of their scarps, both in PRE- and POST-LGM units (classes A and B, respectively).

Field surveys were then performed to: i) validate the classification of structures based on remote sensing analysis; ii) validate fracture length and azimuth, especially focusing on specific locations; iii) integrate the dataset with structures under the resolution level of our DSMs. In the field, additional scarps with a continuous vertical offset ≥1.0 m were identified (Figure 3B).

The analysis of the Fremrinamar Fissure Swarm at a scale of 1: 50,000 allowed us to identify 2,035 structures (Figures 4A–C), classified in 146 eruptive fissures (7.2%), 1,128 extension fractures (55.4%), and 761 normal faults (37.4%) of which 421 dip towards the east and 340 towards the west. The overall group of structures strike N-S to NNE-SSW predominantly, with a peak between N0°E and N10°E, associated with minor peaks striking NNW-SSE (Figures 4D, 5A). The average strike value is N6.8°E with a standard deviation (SD) of 15.7°. Structure length ranges from 7.7 to 14,200 m, with a mean value of 437.2 m and a SD of 799.5 m (Figure 6A). Eruptive fissures show a general NNE-SSW strike, with a major peak between N10°E and N20°E. The average strike value is N14.4°E with a SD of 13.9° (Figures 4E, 5B). They reach a maximum length of 2,485.2 m, with a mean value of 286.2 m and a SD of 298 m (Figure 6B). Extension fractures display a broad N-S to NNE-SSW strike, with a major N0E° to N10°E peak, and minor NNW-SSE peaks. The mean strike value is N7.4°E, with a SD of 16.3° (Figures 4F, 5C). Extension fractures show a maximum length of 2,508 m, with an average value of 165.5 m and a SD of 178.3 m (Figure 6C). Normal faults broadly show a N-S strike, with a major peak between N0°E to N10°E, and minor NNW-SSE to NNE-SSW peaks (Figures 4G, 5D). Both E-dipping and W-dipping normal faults show the same general trend and major peaks (Figures 4H, I). Normal faults exhibit a mean strike of N4.5°E and a SD of 14.5°. E-dipping normal faults display an average strike value of N4.2°E and a SD of 13.6°. W-dipping normal faults, however, show greater variability, with an average strike of N4.9°E and a SD of 15.6°. Normal faults generally display higher length values, if compared to the other structures of the fissure swarm. The highest value, 14,200 m, is reached by a W-dipping normal fault. The average length value for the entire normal fault set is 869 m, with a SD of 1,160 m (Figure 6D). E− and W-dipping normal faults display mean length values of 800 and 954 m, and a SD of 989 and 1,338 m, respectively (Figures 6E, F).

Both strike and length values were related to the UTM X-Y coordinates (UTM zone 28N - WGS84) to evaluate the spatial variation of our dataset. Specifically, the Fremrinamar Fissure Swarm covers an area of 130 × 8–9 km (Figures 4A–C). Normal faults seem to have the highest spatial distribution, covering the entire fissure swarm from south to north. It is possible to notice three gaps in their spatial covering, near the Y coordinates 7,260,000, 7,280,000, and 7,350,000 (Figures 5D, 6D). This could be due to the presence of the more recent Ketildyngja, Búrfell and Sveinar lava flows (4.0, 3.2, and 11 ka, respectively) (Sæmundsson et al., 2012) covering the area and concealing these structures. Extension fractures show a more limited spatial distribution than normal faults (Figures 5C, 6C), while eruptive fissures are mainly located in the southern area of the system, where the central volcano is located (Figures 5B, 6B). It is possible to observe that the fissure swarm shows the highest variability in strike values in its southern portion. Moreover, strike values show a clear anticlockwise rotation starting from the Y coordinate 7,340,000, from a prevailing N-S to NNE-SSW strike in the southern area to predominant NNW-SSE values in the northern one (Figure 5A). This rotation is clearly displayed by the entire dataset, and becomes indisputable when analysing extension fractures (Figure 5C).

Structure length distribution is not constant along the entire system but shows its highest values in the central portion of the fissure swarm (Figure 6A). This is particularly clear for the normal faults subgroup (Figures 6D–F).

The spatial relation between our dataset and the X coordinate (UTM zone 28N - WGS84) was evaluated but did not provide any innovative results (for details see the Supplementary Materials).

For the entire dataset, the longest structures display a N-S strike, specifically between N0°E and N10°E (Figure 7A). The greatest lengths are reached by normal faults (Figures 7D–F), while eruptive fissures and extension fractures, although showing the same strike distribution, hardly exceed 2000 m in length (Figures 7B, C).

The analysis of fracture density was performed through 263 transects, with a spacing of 500 m and striking N106°E (Figures 8A–C), hence parallel to the regional direction of plate spreading in the Northern Volcanic Zone (Hjartardóttir et al., 2012). For each transect, we recorded the number of intersecting structures (i.e., fracture density), first considering the entire dataset (Figure 8D), and then distinguishing between eruptive fissures (Figure 9A), extension fractures (Figure 10A), and normal faults (Figure 11A), also considering their dip direction (Figures 11E, G). For the entire dataset we also recorded the rift width, evaluating the distance between the westernmost and the easternmost structures along each transect (Figure 8E), and the average spacing (Figure 8F). Also, we reported the distance between the westernmost and easternmost structures only considering similar features (referred to as “width” in the graphs) and their average spacing through the ratio width/N. of a single type of feature (Figures 9B, C, 10B, C, 11B, C, F, H). Along each transect we also considered the dilation (Figure 11D) starting from the extrapolation of fault heave, considering a fault dip of 75° (which corresponds to the mean dip observed for Pleistocene to Holocene normal faults within lava successions in Iceland) (Forslund and Gudmundsson, 1991).

For the entire dataset of structures, the highest density (21) is reached along transect 161 (Figure 8D). Fracture density is not consistent along the entire rift. Starting from the northernmost portion of the fissure swarm, it is possible to identify a gradual increase in structure density up to transect 69 (near the intersection with the Grímsey Lineament), where the first peak is located (19 structures). This is followed by an area of decreasing density of fractures up to transect 91, where there is a gradual increase in the number of structures all the way up to the previously mentioned maximum peak, at transect 161. In the southern portion of the rift, the fracture pattern is mostly irregular, with two areas characterised by a paucity of structures, between transects 188 and 205, and near the central volcano. These areas are marked by the presence of more recent lava flows, concealing older structures (as already mentioned in paragraph 4.1). Rift width varies from 0 to 8.8 km, with the highest value reached at transect 71 and a tendency to decrease northward (Figure 8E). The southern and northern parts of the fissure swarm are generally characterised by the lowest width values, with a mean value of ≃3 km. An irregular trend in rift width characterises the southern portion. The average spacing of the fissure swarm reaches its maximum value (≃3 km) at transect 197 (Figure 8F). The highest values are widely located in the central portion of the rift, while deformation becomes more localised elsewhere. The average spacing in the northern and southern areas is rather steady with two peak areas, near transects 28 and 240.

The number of eruptive fissures equals zero up to transect 125 (Figure 9A), where it is possible to observe a gentle increase in their number towards the south, as a maximum peak of 3 eruptive fissures is reached at transects 190, 206, 216, and 246. The width of the fractured area equals zero up to transect 187 (Figure 9B), with a gradual increase up to transect 197, where it reaches its maximum (5.7 km). In the southern portion of the rift, the width of the fractured area is uneven and generally confined within the range of 1–3 km. The average spacing of eruptive fissures equals zero in the northern part of the rift and reaches a peak of 2.85 km at transect 197 (Figure 9C). Values are generally higher in the central portion of the rift, including values within the range of 2–3 km, than in the southern portion, where values are within the range of 1–2 km.

The northernmost portion of the Fremrinamar Fissure Swarm is characterised by the absence of extension fractures for the first 4 km. Up to transect 48 it is possible to observe a gentle increase in their number, followed by a subsequent decrease. Fracture density is mainly concentrated in the central and southern portions of the rift (Figure 10A). Two main peaks of 11 extension fractures are reached at transects 204 and 211. Rift width presents a clear, gradual increase in its values up to transect 177, where it reaches the peak of 6.63 km (Figure 10B); this is followed by a decrease in its southernmost portion, becoming null from transect 253. Average spacing values show a similar configuration, with the highest values located in the central and southern portions of the rift (Figure 10C). Values range from 0 to 1.4 km in the northern area, and from 0 to 1.7 km in the central one, while in the southern area the spacing ranges from 0 up to the maximum, 1.95 km, at transect 208.

Normal faults cover the entire fissure swarm from north to south, but they are mainly located in the northern and central portions of the fissure swarms (Figure 11A). The main peak of 16 normal faults is reached at transects 69 and 70 (within the intersection with the Grímsey Lineament), while it is possible to identify a slightly smaller peak (14 faults) at transect 161. The number of normal faults equals zero between transects 189 and 198 and clearly decreases near the central volcano. This could be due to the presence of younger lava flows covering the area (as already noticed in Figures 5D, 6D). E-and-W-dipping normal faults show a similar pattern, with the highest structure density located in the northern and central areas of the rift, and a subsequent decrease near the central volcano (Figures 11E, G). The main peak of E-dipping normal faults is located at transect 64, with a total of 10 scarps, and a second one (8 scarps) at transect 177 (Figure 11E). W-dipping normal faults reach a major peak of 9 scarps at transect 173, and show a minor one (8 scarps) between transects 71 and 69 (Figure 11G).

The area covered by normal fault scarps (“width” in the graphs) is quite heterogeneous along the fissure swarm (Figure 11B). From transect 0 to 28, values range from 0 to 4.86 km, while they show a decrease up to transect 62. Successively, the rift is characterised by higher width values, reaching a peak of 8.86 km at transect 71 (Figure 11B). From this point and up to transect 125 there is a slight decrease in width values, followed by an increment up to the second peak (7.46 km) at transect 176. The width of the faulted area is less evolved in the southern part of the rift and shows two main areas where values are equal to zero (possibly linked to the presence of recent lava flows, as explained before). The average spacing of normal faults is moderately consistent along the rift (Figure 11C), although it is possible to notice slightly higher spacing in its northernmost area. However, the maximum value of 3.63 km is however reached at transect 188. In the southern part of the fissure swarm, spacing values show a small increase up to the area near the central volcano, followed by a decrease up to the southernmost section. The average spacing of E− and W-dipping normal faults displays an analogous arrangement, with the higher values located in the northern part of the rift, and a slight increase of spacing values towards transects 187 and 184 respectively (Figures 11F, H). The southern area is characterised by a lack of faults, followed by a modest increase as far as the central volcano, accompanied by a decrease towards the southernmost sections.

The analysis of the dilation highlights that this value clearly decreases from the southern sector of the Fremrinamar Fissure Swarm toward the northern area (Figure 11D); this becomes more evident if the southernmost portion of the fissure swarm, where more recent lava flows cover the structure, is omitted. This distribution, suggesting a higher deformation in the southern Fremrinamar Fissure Swarm, could be an indicator of the direction towards which the deformation is developing. Dilation values also display a strong local increase in the area where the Fremrinamar Fissure Swarm intersects the GRL.

Regarding the relationship between eruptive fissures and normal faults, we noticed that most fissures are located within the first 600 m from a normal fault scarp (Figure 12A) and very few eruptive fissures are located at their tip or along normal faults (yellow column in Figure 12A). Considering eruptive fissures that are located within a graben, we observed that they are usually shifted towards one of the border faults, and only in few cases they are situated at the centre of the graben (Figure 12B).

Our dataset includes 76 normal faults (with a total length > 2 km), among which 3 (≃4%) show a symmetrical profile (Figure 13A and Supplementary Figure S3), 58 (76.3%) an asymmetrical one (Figures 13C, D), and 15 (19.7%) were not classified (class E) due to extensive erosion of the scarps or offsets beneath the resolution limit of our DSM (Figure 13B and Supplementary Figure S4). The analysis of slip profiles (of classes from A to D) allowed us to classify the direction towards which the deformation is developing for the asymmetrical ones, revealing 31 (40.8%) profiles propagating towards the north (Figure 13C and Supplementary Figure S5), and 27 (35.5%) profiles defined by a southward propagation (Figure 13D and Supplementary Figure S6). In Figure 14 we show the faults of the Fremrinamar Fissure Swarm with colours corresponding to the four types of slip profiles. It is possible to observe a slight predominance of north-tapering slip profiles, within the Fremrinamar Fissure Swarm, considering scarps belonging to classes from A to D.

The difference becomes more evident considering only normal fault scarps belonging to classes A and B, which are the most reliable, where we obtained 14 profiles tapering towards the north, with a cumulative length of 55.5 km, and 10 profiles tapering towards the south, with a cumulative length of 32.3 km.

For the entire set of normal faults, we collected maximum throw values (as reported in section 3.2). Generally, the highest throw values are located in the southern and central portion of the Fremrinamar Fissure Swarm (Figure 15). It is possible to observe a slight increase in throw values for normal faults located within the intersection with the Grímsey Lineament, and a clear decrease in these values north of it.

The general architecture of the Fremrinamar Fissure Swarm is characterised by the change of the overall trend from N-S/NNE-SSW in the central-southern part to NNW-SSE in the northern part (Figure 5). Considering that the azimuth of the regional tectonic spreading for the Northern Volcanic Zone is around N104-106°E, according to geomagnetic data (as far back in time as 2.6 Ma) (DeMets et al., 1994; DeMets et al., 2010; Hjartardóttir et al., 2012), the overall azimuth of the Fremrinamar Fissure Swarm structures is mostly orthogonal to the regional least principal stress (σ₃), but only in the southern and central part. This is consistent also if we consider other spreading vector values based on GPS data from the last 25 years (Metzger et al., 2011; Metzger et al., 2013; Drouin et al., 2017; Sigmundsson et al., 2020). This change of the overall fracture trend, occurs in correspondence of the intersection with the Grímsey Lineament (Figure 4A). This area of intersection is commonly reported in previous papers (e.g., Einarsson, 1991; Hjartardóttir et al., 2016b). The Grímsey Lineament is composed of a series of 5–20 km wide NNW-trending offshore grabens, with dominant normal faulting along NNW- to N-S-striking planes and subordinate strike-slip faulting (Rögnvaldsson et al., 1998; Lund and Townend, 2007). We would like to underscore that our field observations did not suggest any clear strike-slip movements in the Grímsey Lineament-Fremrinamar Fissure Swarm overlapping area, nor did we recognise any strike-slip offset landforms on the high-resolution mosaics. Since the entire northern part of the Fremrinamar Fissure Swarm shows an anticlockwise rotation of the whole set of fractures, it seems that the Fremrinamar Fissure Swarm here rotates around a “pivot” point assuming an orientation parallel to the grabens of the Grímsey Lineament. Such observations support the hypothesis explained in the model of Gudmundsson (2007), where the Grímsey Lineament is related to the propagation of the Fremrinamar Fissure Swarm to the north, towards the Kolbeinsey Ridge. The stress field generated by their interaction results in the northernmost part of the Fremrinamar Fissure Swarm bending towards the Kolbeinsey Ridge, as shown by our data (Figure 5). The en-échelon system of grabens and horsts which constitutes the currently active Grímsey Lineament (Gudmundsson, 1993) forms a part of the bent and overlapping extension of the Fremrinamar Fissure Swarm. Additionally, Rögnvaldsson et al. (1998) showed that most earthquakes associated with the Grímsey Lineament occur on N-striking left-lateral strike-slip faults, and that no through-going WNW-striking faults parallel to the Grímsey Lineament are present. With regard to this, we would like to point out that the rift zones belonging to the Northern Volcanic Zone are arranged with an overall left-stepping geometry, consistent with the interaction with the Kolbeinsey Ridge.

This overall observation results in an anticlockwise rotation of maximum horizontal stress (σ_hmax) and minimum horizontal stress (σ_hmin) near the ends of the rift zones of the Northern Volcanic Zone, which could be recognised at the level of surface structures. Differently from Pasquaré Mariotto et al. (2015), who studied the intersection between the Húsavík-Flatey Fault and the Theistareykir Fissure Swarm, in our case it was not possible to define the local orientation of σ₃ due to a lack of opening vectors along the extension fractures. However, we can assume the orientation of the structures as representative for the local σ_hmax (Gudmundsson et al., 1993). This interpretation can be made at least for the post-LGM, since normal faults in the north are not eroded by glaciers, as we observed during the field survey. A rotation of the σ_hmax and σ_hmin north of the Grímsey Lineament can be extended as far back as 1 Ma, according to the model proposed by Gudmundsson (2007).

We also noticed that the longest structures are not perpendicular to the plate spreading direction (Figure 7), either north or south of the intersection with the Grímsey Lineament. The possible change in direction north of it has been already explained above and can follow the scheme from Gudmundsson (2007). South of it, in the case of direct control of plate tectonics, we could expect that the structures would become longer, and thus more developed, as they gradually assume a strike perpendicular to the regional far-field tectonic extensional stress. The regional spreading values of N106°E (Hjartardóttir et al., 2012) and N112°E (Drouin et al., 2017) for the whole Northern Volcanic Zone, increase to N109°E and N115°E for the northern part of the Northern Volcanic Zone (Metzger et al., 2011; Metzger et al., 2013). Our data indicate that extensional deformation occurs mostly along N90°-100°E values, thus suggesting a slight anti-clockwise rotation of the spreading direction along the Fremrinamar Fissure Swarm. This behaviour was already observed by Corti et al. (2021), who found an overall average spreading direction of N99.5°E along the nearby Krafla Fissure Swarm. To explain it, Corti et al. (2021) suggested two hypotheses, which might be complementary: i) the development of the structures has been influenced by the propagation of magma from the magma chamber below the central volcano; ii) the regional stress field has a local slight reorientation in the southern part of the rift, where σ₃ attains a more E-W trend. We would like to highlight the fact that the spreading directions collected at the Fremrinamar Fissure Swarm and Krafla Fissure Swarm are based on geological data that cover the Holocene time span. Therefore, also the time range difference might explain the more marked rotation of spreading compared to GPS data, which are related only to the last 25 years.

Finally, we found that normal faults along the Fremrinamar Fissure Swarm have an asymmetric development, with the west-dipping faults being longer, having a wider range of strike, and being slightly less abundant than the east-dipping faults (Supplementary Table S1). This suggested that the west-dipping faults are more capable of accommodating the extensional deformation and are more mature. With regard to the latter point, Drouin et al. (2017) suggest that, based on GPS data, the axial zone of the Northern Volcanic Zone corresponds to the Krafla Fissure Swarm, as indicated also by the data from Sigmundsson et al. (2020). This implies that the Fremrinamar Fissure Swarm could represent the eastern part of the Mid-Atlantic ridge, where a predominance of west-dipping faults is expected. Differently, the Theistareykir Fissure Swarm is located to the west of the Krafla Fissure Swarm, and shows a series of well-developed east-dipping normal faults (Tibaldi et al., 2020a). This overall scheme is in agreement with the architecture that is expected for slow spreading oceanic ridges, where normal faults predominantly face the axial zone, as highlighted for the Reykjanes Ridge (Searle and Laughton, 1981). Additionally, we observed that longer west-dipping faults are located south of the intersection between the Fremrinamar Fissure Swarm and the Grímsey Lineament, and decrease north of it, together with the rift width. This decrease is noticed especially if we consider faults longer than 2 km (Supplementary Table S2). This can be interpreted as due to a greater extension south of the intersection, where most strain is expected, considering available GPS data (Metzger et al., 2011; Drouin et al., 2017; Sigmundsson et al., 2020).

It is interesting to compare the above observations with similar settings in the conterminous rift zones. In the Theistareykir Fissure Swarm, although the overall trend of rift structures follows the regional stress field (Bonali et al., 2019b), an important change in the direction of structures has been observed along the intersection between the Húsavík-Flatey Fault and Theistareykir Fissure Swarm by Pasquaré Mariotto et al. (2015) and Tibaldi et al. (2016). Particularly, these authors observed local reorientation of the fractures with the presence of en-echelon, NNW-SSE-striking normal faults and tension fractures, as well as the transition of some faults into tension fractures north of the prolongation of the Húsavík-Flatey Fault, and the decrease in the cumulated offset of all the faults north of the Húsavík-Flatey Fault prolongation. They also recognised an example of a laterally offset post-LGM lava flow along a WNW-ESE-striking right-lateral strike-slip fault. In fact, the Húsavík-Flatey Fault is composed of a swarm of WNW-ESE-striking right-lateral strike-slip faults, which locally assume transtensional kinematics (Sæmundsson, 1974; Gudmundsson, 1993; Gudmundsson, 2007; Tibaldi et al., 2016). Thus, differently from the intersection Grímsey Lineament - Fremrinamar Fissure Swarm, at the Húsavík-Flatey Fault-Theistareykir Fissure Swarm intersection, a well-defined buffer zone of interaction is visible, as shown also by analogue experiments (Bonali et al., 2018).

Most of the asymmetric fault slip profiles of the Fremrinamar Fissure Swarm, especially those belonging to classes A and B, show a tendency to taper northward (Figure 14). This can be interpreted considering that single faults develop in the direction towards which their slip profile tapers out (Manighetti et al., 2001), indicating a tendency of the faults to propagate towards the north. This is confirmed both by the number of faults and their total length. This configuration is consistent with a general northward propagation of the Theistareykir Fissure Swarm (Tibaldi et al., 2019; Tibaldi et al., 2020a) and the Krafla Fissure Swarm (Corti et al., 2021). Sæmundsson (1974) suggested, so far, the hypothesis of a general overall northward propagation of the entire Northern Volcanic Zone. This setup is also corroborated by a series of other clues: i) the total number of normal faults decreases from the central portion of the rift towards the north, as also observed in the Theistareykir Fissure Swarm by Tibaldi et al. (2020b) and within the Krafla Fissure Swarm by Corti et al. (2021); ii) fault offsets are higher in the southern and central parts of the fissure swarm and are less developed in its northern part, as noticed by Tibaldi et al. (2016) in the Theistareykir Fissure Swarm; iii) dilation values show a clear decrease towards the north, as observed by Tibaldi et al. (2020b) and Corti et al. (2021) in the Theistareykir Fissure Swarm and Krafla Fissure Swarm respectively; iv) rift width tends to decrease northward, in agreement with the propagation model proposed by Karson (2017) in which a propagating rift can be represented as a wedge-shape area where the spreading decreases towards its apex (i.e., in the direction toward which the rift is developing).

At the Theistareykir Fissure Swarm, Tibaldi et al. (2019), Tibaldi et al. (2020a) documented in great detail that the majority of normal faults propagated outward from a central volcano, and that this might relate to the outward propagation of dykes from the magma chamber. At the Theistareykir Fissure Swarm, the central volcano is located in the middle part of the rift, and thus the faults propagated northward, north of the volcano, and in the opposite direction, south of the volcano. This double outward propagation of faults has been observed also around the Ado’Ale volcano in the Ethiopian Rift (Dumont et al., 2017). With regard to the relationship between lateral dyke propagation and surface deformation, there is a general lack of data in the literature especially in terms of fault tapering. However, we would like to remark two case studies that proposed important observations about this topic. Tibaldi et al. (2020c) studied an example of surface deformation induced by dyking, during the Krafla Fires (1975–1984). Here, the authors recognised a narrow graben that formed due to a N-propagating dyke, characterised by faults that are clearly visible to the south and fade into an extensional fractures field to the north, suggesting a tapering of the graben faults towards the north. During the Bardarbunga rifting episode in 2014, the formation of a graben was documented (e.g., Ruch et al., 2016). Values of subsidence along the graben decreased from the southern and central part of it (4.5–6 m) to the northern part (1–2 m) where the main eruptive site is located (Hjartardóttir et al., 2016a; Müller et al., 2017), consistently with the propagation of the dyke from the SW to the NE (Sigmundsson et al., 2015).

At the Fremrinamar Fissure Swarm, the central volcano is located in the southernmost part of the rift (Figure 14), therefore the northward propagation of dykes from the magma chamber can explain the propagation of the faults in the same direction, along the whole rift. This could also explain the concentration of the eruptive fissures in the southern segment of the rift, around the central volcano, where shallow dykes, propagating laterally outward from the magma chamber, intercepted the surface (Figure 9).

Another observation that supports the role of magma in dictating the rift structures, is given by the increase of the rift width north of the Fremrinamar central volcano, between transects 70 and 180 (Figure 8E). This widening of the rift may correspond to a tendency of dykes to deepen outward from the central volcano, as observed by Hollingsworth et al. (2013), who suggested a general tendency to rift widening moving away from the Krafla central volcano due to deepening of the dyke tip (Hollingsworth et al., 2013). Dyke-induced grabens, in fact, become wider with the deepening of the dyke tip, as observed by analogue experiments and numerical modelling (Pollard et al., 1983; Rubin, 1992; Trippanera et al., 2015b; Al Shehri and Gudmundsson, 2018; Drymoni et al., 2023).

Although these interpretations fit with the overall distribution of the different types of structures along the Fremrinamar Fissure Swarm and with the dominant direction of fault slip profile tapering, we must take into account that a relevant number of faults shows a southward slip profile tapering. This observation does not fit with all the other results, and might be considered an uncertainty linked to the slip profile method and to the complexity of fault development in volcano-tectonic rifts, where both magmatic and tectonic stresses can contribute to creating fissure swarms. The complexity derives also from the fact that the slip profile geometry at dyke-induced faults depends on the shallow magma flow direction, which can be from horizontal to vertical and cannot be constrained if dykes do not crop out. Fault slip is also a function of the lithology of the offset rocks (e.g., lava vs. hyaloclastite piles) (Trippanera et al., 2019; Tibaldi et al., 2020c) and of the depth/distance of the intrusion compared to dyke thickness (Trippanera et al., 2015a). Furthermore, possible fault rejuvenation during successive dyking events can occur (Gudmundsson and Bäckström, 1991). Nevertheless, we suggest that at least some of the faults with slip profiles displaying southward propagation, might be the result of single episodes of southward dyke propagation from other magmatic sources (e.g., Thorarinsson, 1959). In our study area, we highlight the presence of the Sveinar graben, which clearly shows a southward tapering profile of the graben faults (Figure 14B). Such graben was studied by Tentler and Mazzoli (2005), who dated faulting after the deglaciation, starting from data of Thorarinsson (1959). Its formation is suggested to be prior to 6–8 ka (Thorarinsson, 1959), when an eruption started at the Randarholar eruptive fissure (Y coordinate 7,306,090) and propagated towards SSW (Thorarinsson, 1959), until it met the already existing graben and followed its direction southward. Bäckström and Gudmundsson (1989) suggested that the displacement along graben faults continued after the eruption. In view of the above, such graben could represent an example of southward fault propagation guided by an episodic southward propagating dyke from other sources in the Northern Volcanic Zone.

With regard to the 146 eruptive fissures identified in the present work, which represent about 7% of the overall dataset of structures, it was noticed that their direction is mostly NNE-SSW, and that they are all located south of the Grímsey Lineament intersection (Figure 9). Their strike are more compatible with the directions of plate spreading based on geomagnetic data (DeMets et al., 1994; DeMets et al., 2010; Hjartardóttir et al., 2012), than those based on GPS (Metzger et al., 2011; Metzger et al., 2013; Drouin et al., 2017; Sigmundsson et al., 2020). Since eruptive fissures can be used as indicators of the stress field (Tibaldi and Bonali, 2017), these structures suggest a post-LGM orientation of the intermediate principal stress (σ₂) and the least principal stress (σ₃) of N10°-20°E and N100°-110°E, respectively, at least for the southern and central part of the Fremrinamar Fissure Swarm. We refer to post-LGM time, since all of them were identified by the morphometry of well-preserved scoria cones or vent alignments (e.g., Tibaldi, 1995; Tibaldi and Bonali, 2017), related to post-LGM units (Sæmundsson et al., 2012). Such an indicator is mostly comparable with the geometry of extension fractures, which represents the post-LGM deformation field. The latter mainly strike N to NNE in the southern part of the rift, whereas they strike more NNW to the north (Figure 5C).

Regarding their distribution along the rift, compared to latitude, eruptive fissures totally disappear in correspondence of transect 125 (Figure 9A), about 44.8 km north of the central volcano, south of the Grímsey Lineament. This is quite common in rifts belonging to the Northern Volcanic Zone and has been observed both at the Theistareykir Fissure Swarm and the Krafla Fissure Swarm. With regard to the latter, a deepening of the dykes, propagating to the north and outward from the magma chamber (Hollingsworth et al., 2013) for as far as 70–80 km from the central volcano, has been observed during the Krafla Fires (1975–1984) (Einarsson and Brandsdottir, 1980). Observations at the Theistareykir Fissure Swarm show eruption vents as far as about 5 km from the central volcano (Tibaldi et al., 2019), and there is no data related to lateral dyke propagation in historical time.

From another perspective, very few (less than 1%) eruptive fissures from the present study were recognised to be along the prolongation of normal faults or to be emplaced onto them (Figure 14). Such observation suggests that very few dykes can be captured, during their lateral or vertical propagation, by previous normal faults in this context, as suggested by previous authors (e.g., Johnson, 1961; Currie and Ferguson, 1970; Gudmundsson, 1983; Delaney and Gartner, 1997). Surely, in our case study, a very good example is represented by the 1875 (Thorarinsson, 1959; Gudmundsson and Bäckström, 1991) eruption, as explained above. Additionally, in Iceland, Gudmundsson (1983) conducted a study in the eroded rift zones in the eastern part of the country, showing that only about 10.5% of dykes occupy normal faults, but rather they create new tensile fractures to continue their propagation. Our observations suggest that the majority of the dykes in the Fremrinamar Fissure Swarm can follow the regional stress field but not necessarily pre-existing structures and dyke-induced graben faults; as a consequence of this, most dykes create new fractures and follow those as main magma paths. This is also consistent with what was observed during rifting episodes in Iceland (e.g., Ruch et al., 2016) and by analogue experiments, from single dyke-induced graben to rift scale (Pollard et al., 1983; Trippanera et al., 2015a; Trippanera et al., 2015b).