A total of 10,729 fracture segments were identified from aerial images.

The features used for the preparation of the structural map (Figures 2-5) are: 1) The stratigraphy [54] , grouped into three relative rock ages. These are the older be drock (upper Miocene-late Quaternary), the postglacial lavas aged 15000 - 8000 yrs, and the youngest Theistareykir lavas dated 2400 yr BP [95] . 2) Fracture

segments, classified in terms of their relative age (younger and older), trace (major, minor, faint), types (open fracture, normal-slip, undifferentiated fractures, and then further analyzed geometrically for strike-slip motions) [14] . 3) Lava tubes and lava channels, mapped to distinguish them from possible fractures, along with Pleistocene and postglacial craters. 4) Surface geothermal manifestations, outlined and labeled as low to high.

Due to the shallow level of erosion, very few appear at the surface. In fact, less than 10 of all mapped segments are dykes, and they crop out far to the east and west of the mapped area (Figure 2), in accordance with the field mapping of [54] [92] . The mapped fractures are thus mostly a-magmatic, but this does not exclude that dykes can arrest at depth, creating surface extensional fractures or grabens, e.g. [36] [97] [98] .

It is emphasized that 10729 is not the total number of faults cutting the series as a single fault consists of several segments along its trace, regardless of age.

The mapped structures are at different stages of their evolution during which, segments coalesced, accumulated displacement, and some evolved into major structures (Figures 4(a)-4(d) and Figures 5(a)-5(d)). The length of individual segments forming a single structure can be as short as ~5 m and as long as 5000 m. Generally, older segments are the longest with sharper traces in the topography and a greater dip-slip in the series. Younger fractures are shorter, fainter in traces (hair-like) and with much less or no dip-slip. Hence, the difficulties in identifying them in outcrops. Minor segments have relatively well visible traces on images, but their slip has not lead to significant changes in the topography. Minor segments can be on the trace of a major and longer structure.

Segments labeled as faint/gash have the least prominent traces, mostly because they are young and have not been subject to major or repetitive deformation. Undifferentiated fractures can have a mild or a strong trace, but their displacement could not be determined because they are either prominent joints or even strike-slips without marker horizons to determine their sense of motion. In some cases, uncertainty did not allow estimating towards which direction a normal fault dips. Possible fractures are breaks in the ground where either their exact trace or their existence cannot be confirmed due to poor exposure.

Fracture frequency is one of the many important outcomes of this investigation (Figure 3(b)). An attempt was made to find the most common fractures by counting the strike of each of the 10729 segments, but the method gave an erroneous result. The reason is that a single major long segment was counted only once, but young minor fractures consisting of several short segments were counted as many times as the number of segments along the fracture trace. Therefore, a method conceived by [99] was used in which the fracture frequency in a given set is estimated by adding up the length of all segments in the same 10˚ strike-interval. The rose diagram obtained from this method (Figure 3(b)) reflects more coherently the frequencies in accordance with the visual assessment of fractures on the map (Figure 2 and Figure 3(b)).

It is emphasized that this section describes the fracture frequency only in terms of peaks of fractures that stand out compared to the rest of the fracture population (Figure 3(b)). At this stage, the statistical analysis does not take into account the sense of motions and the fracture types of rift and transform zones, which are discussed in the following section. Therefore, the fracture sets described here (Figure 3(b)) may not necessarily have the same wide strike-ranges as when the fracture sets are defined on the basis of their motions (Figure 8).

The present statistical analysis shows that fractures with four strike-ranges stand out (Figure 3(b)): N-S (N01˚-N20˚E), ENE (N51˚-N70˚E), WNW (N111˚-N130˚E), and NW (fractures ranging equally from N131˚-N160˚E). The peak of N11˚-N20˚E (8,52%) indicates that the most frequent individual fractures are N-S, but the N-S as a set represents only 16,6% of the total fracture population. The ENE set accounts for 14,5% with a peak at N61˚-N70˚E (7,41%), and the WNW 14,9% with a peak at N121˚-N130˚E (7,75%). The NW set covers three equal 10˚ strike-interval (6,19% to 6,24% of the fracture population), but combined, they seem the most frequent fractures (18,70%). A less outstanding peak shows a NNE set (N21˚-N30˚E) occupying some 5,17% of the population.

The E-W fractures appear clearly as a separate set on the structural maps with a strike-range of N81˚-N100˚E (Figure 2 and Figure 7(a)). Statistically, however, the E-W set represents 6,55% of the population although its peak (N91˚-N100˚E) is only 3,76% (Figure 3(b)), implying these are the least frequent fractures.

Extension is associated with all the six fracture sets and is expressed either as a dip-slip or as an opening along fractures (Figures 4(a)-4(d) and Figures 5(a)-5(d)). Open fractures are young, with an aperture up to 3 m at the surface. The dip-slips are 0,5 m to maximum 5 m in the 2400 yr PB lava, 5 - 20 m in the postglacial lavas, and up to 200 m in the bedrock. The N-S (N01˚-N20˚E), NW (N131˚-N160˚E) and E-W (N81˚-N100˚E) faults have the highest dip-slips in the bedrock, judging from the visual inspection of fault scarps on aerial images and the estimated values of topographic contours on maps.

Fracture geometry is a key for identifying the strike-slip component of fractures on images (Figure 5 and Figure 6) and the en échelon segmentation in particular is the criterion for determining the sense of shear motions. Left-stepping en échelon arrays, typical of dextral motion, were observed along the WNW (N101˚-N130˚E) (Figure 5(a) and Figure 6(a)) and NW sets (Figure 5(b)) and Figure 6(a)), which are respectively parallel to the HFF and the Tjarnarás Fault. Right-stepping en échelon arrangements, typical of sinistral motion, appear along the NNE (N21˚-N50˚E) to ENE (N51˚-N80˚E) fractures (Figures 5(c)-(d) and Figure 6(a)). The horizontal offsets of strike-slip faults are difficult to estimate due to lack of marker horizons but they are far less than the dip-slips. Data from SISZ indicate that offsets of strike-slip source faults are of the order of 2 m during a single earthquake [100] , but they can be cumulative up to 20 m in the older bedrock such as in west Iceland [101] [102] . However, no shear motion could be identified for E-W fractures on the basis of fracture geometry. Similarly, about 7,9% of the fracture population form a NNW set (N161˚-N180˚E), which has a dip-slip but a strike-slip could not be identified on them either.

The en échelon segments can be the surface expression of deeper and blind strike-slip faults that have not yet broken the surface (Figure 6(c)). The best examples are the WNW and NW open fractures of Sæluhúsveggur and Grísatungur that crop out locally on the trace of the HFF (Figure 5(a)-(b)).

“V” shape fault termination is another example of fracture geometry. Shorter segments branching out towards the end of particularly individual strike-slip faults form an angle with the main fault plane. Such structures are secondary horsetail splays, accommodating the shear motion of the main strike-slip fault segment (Figure 6(c)).

Regardless of the relative age or type of fractures in Theistareykir, individual segments of a single structure can dip to opposite directions along the overall fault trace due to the steepness of the fracture planes. This feature is also observed in other areas of Iceland [25] [26] .

The emerging structural pattern is a combination of pure extensional fractures parallel to the rift and five Riedel shear sets belonging to the transform zone (Figure 6(b)). The rift-parallel fractures are N-S, with strike-ranges identical to those on Figure 3(b). Shear motions identified from fracture geometry indicate that NNE to ENE Riedel shears act as sinistral oblique-slips, and the WNW to NW fractures display dextral motion. Although fracture geometry was not conclusive for the sense of shear motion along the E-W set, these fractures have likely a strike-slip component (either sinistral or dextral), similar to the adjacent NNE-ENE or WNW-NW sets.

The strike and motions identified here are identical to those of regional earthquakes within the TFZ [59] [60] , or even to those recorded during local earthquakes above the Theistareykir geothermal field [83] [103] [104] .

The stress map of Iceland combining focal mechanisms, boreholes breakouts, induced fractures, and geological features [65] shows that stress fluctuations occur both in and adjacent to the rift and transform plate boundaries. Such fluctuations result in the co-existence of normal faults and a wide range of strike-slips, regionally. Detailed stress analyses in this part of TFZ and NRZ [63] indicate that the spreading direction is roughly N102˚E near the NRZ, but ~N80˚E across TFZ, thus splitting the displacements into transversal (dip-slip motion) and lateral (strike-slip motion). Thorough analysis of kinematic indicators in the bedrock between the HFF and the DL identified severe fluctuations in the direction of S_Hmax, which explains strike- and dip-slip motions on a wide range of fractures [64] .

Although detailed stress analysis has not been carried out for any rock age in Theistareykir, the tectonic context, rock ages, and fracture sets are identical to those studied by [63] [64] [65] . Therefore, similar complex stress fluctuations can be expected in Theistareykir as well, explaining why extension is associated with fractures of all sets and why strike-slip occurs on such a wide strike-range in rocks of any age.

An attempt was made to identify the structural weak zones in the study area [14] . These zones are defined here based on one or several of the following criteria on aerial images: zones with high fracture density, prominent morphostructural geometries in the landscape, pronounced displacements, and control of lava structures (Figure 6(d)). As these zones form along any set, identifying them is critical for a better understanding of the tectonics of rift and transform zones and for their roles in the control of geothermal activity.

These structural weak zones display various configurations, regardless of whether they have broken the surface and been reactivated, or they are only surface expressions of deeper blind faults. The weak zones that have broken the surface appear as one dominant structure, or as a cluster of segments having more or less sharp traces and variable amount of dip-slips. The deep structures that did not break the surface consist mostly of minor or faint segments, open fractures, and occasionally of major segments along the fracture trace. Although the strike-range of fracture segments within a structural weak zone can be wide (Figures 3(b)) and Figure 6(b)), the strikes and motions of those individual segments are compatible with the outline and motion of the main structural weak zone. Several interesting features appear from the interpretation of these zones, some of which are shown for the first time by this study:

a) The weak zones belonging to the rift and transform zones appear tightly parallel and widely distributed throughout the Theistareykir Fissure Swarm. Together, they form a complex fracture pattern (Figure 7(a)), responsible for block compartmentalization at any scale. Along their lengths on aerial images, these weak zones cut the series of all ages at the surface (Figure 2), while well data (ÍSOR database) indicate highly fractured bedrock also under the thin postglacial and the 2400 yr BP lavas in Theistareykir. Most of the subtle structures in the younger lavas are likely reactivation of older underlying structures.

b) The Theistareykir Fissure Swarm is interpreted here as having an 11 km width (Figure 2 and Figure 7). This suggestion is based on the observation that the western boundary of this swarm could be on the trace of the last N-S postglacial eruptive cone and the northward continuation of the N-S normal fault segment to the north of Skessuskál, while its eastern boundary could be the N-S structure stretching from Ketilfjall southwards on the trace of the last open fractures at Kvíhólafjöll.

c) The N-S rift-parallel fractures are the most frequent individual set. However, they do not display a high number of weak zones compared to the other sets of the study area (Figure 3(b) and Figure 7(a)). Most N-S weak zones consist of young open fractures in the postglacial lavas, and of normal faults with conspicuous dip-slips (≥200 m) in the bedrock. From the HFF to the south, two short grabens occupy the central part of this fissure swarm (Figure 7(b)). The graben to the north cuts the 2400 yr BP lava, and the one to the south the postglacial lavas of Borgarhraun and the subglacial hyaloclastites of Bæjarfjall. Almost all the N-S weak zones are on the flanks of these two grabens, i.e., away from the centre of the fissure swarm (Figures 7(a)-(b)).

Although smaller horsts and grabens exist on the flanks of each graben, generally, N-S faults dip mostly to the west on the eastern flanks of the grabens, and to the east on their western flanks.

d) An interesting relation may exist between the NNE and ENE structural weak zones. Open fractures strike far more NNE than ENE (Figure 2 and Figures 5(c)-(d)). Similar to the WNW and NW dextral open fractures in Sæluhúsveggur/Grísatungur, the en échelon NNE sinistral fractures appear very near or above the ENE structures (Figures 5(a)-(d)), thus acting possibly as surface expressions of deeper ENE strike-slip faults.

e) Although E-W structures are the least common, the dip-slip of a few older E-W faults in the bedrock to the west of Skeiðin is among the highest observed. Both the dip-slips and the traces of those E-W faults decrease drastically in the adjacent younger lavas.

f) The WNW weak zones are as frequent as the N-S zones (Figure 7(a)). WNW faults generally dip to the southwest and display both dip-slip and dextral motion similar to the HFF. For the record, the WNW segments of the HFF (N121˚-N130˚E) are relatively straight, identical to all other WNW structures of similar strikes (Figure 2).

g) The NW weak zones contain faults and fractures that dip mostly to the southwest. These structures display some of the highest dip-slips in the bedrock, and the highest number of open fractures after the N-S set. The Tjarnarás Fault and the open fractures of Grísatungur are the most obvious NW structures (Figure 5(b)). Two hypotheses are suggested for at least these latter NW open fractures. First, their typical en échelon geometry could correspond to a local reopening of several parallel NW faults of the bedrock that intersect the HFF. Such an opening is along the NE direction and almost perpendicular to the WNW direction of spreading. Secondly, these NW en échelon open fractures could be short dextral segments that accommodate the shear motion of the HFF at the surface together with the neighbouring WNW open fractures of Sæluhúsveggur. Neither these open fractures nor the Tjarnarás Fault are the sole NW structures in the area as NW weak zones are widely spread, with a subtle left-stepping en échelon arrangement indicative of a dextral motion.

Due to their significance for the eruption and geothermal activity, some specific features of the N-S and WNW structures are further interpreted here.

As stated earlier, very few N-S dykes, and none with other strikes, crop out in the study area due to the level of erosion. N-S eruptive fissures typical of rift fissure swarms in north Iceland are not common here either. Furthermore, the two N-S grabens in the middle of the Theistareykir Fissure Swarm seem to be shifted in an apparent sinistral motion [51] [92] at the latitude of Stórihver (Figure 7(b)). Based on the number of N-S faults on the grabens’ flanks, fault geometry and geological elements at Stórihver, a new structural analysis [105] suggests that the N-S grabens are shifted by the dextral motion of a WNW strike-slip fault crossing the Stórihver crater, and that the 2400 yr BP eruption of the picrite lava likely occurred on this deep and mostly blind fault of the transform zone. A short horsetail splay of the main WNW fault, also with a dextral oblique-slip motion, stretches from north and east of Stórihver eastward into the geothermal field.

The eruption on a WNW fault perpendicular to the Theistareykir rift fissure swarm demonstrates that strike-slip faults are pathways for magma, and that transform zones can be leaky. Magma injection into strike-slip faults of transform segments is not so uncommon as it is reported from the Tertiary dykes of west Iceland [106] , the TFZ in north Iceland [42] , the young eruptive fissures in the RP oblique rift in southwest Iceland [107] , and even the 2014 fissure eruption of Holuhraun stretching from Bárðarbunga to the north of Vatnajökull [44] .

The mapped fractures are of different ages and at different stages of evolution during which they coalesced, accumulated slip, and were reactivated. As the N-S extensional faults parallel to the rift constitute only 16,6% of the fracture population (Figure 3(b)), a rough insight into how the fractures form and evolve is crucial particularly where rift and transform zones interact.

An absolute dating of these fractures requires other data than available in this paper. Therefore, a simple method is used here where the 10729 fracture segments are classified in three main groups according to the relative age of their host rocks (Figures 8(a)-(d)), using the method of fracture-strike vs. fracture-length [99] . This approach may be simplistic, because young fractures can also form in the older series. Yet, it provides an overall understanding of how the extensional fractures of the rift and the Riedel shears of the transform zone form and evolve in time [15] . The strike-ranges of the sets described below correspond to fractures for which pure extension and shear motions could be identified.

a) Group (1) represents the youngest fractures, as they are in the 2400 yr BP Theistareykir lava (Figure 8(b)). It is reiterated that fractures in this lava have generally minor or faint traces as they are still young and not fully developed. Clockwise, their frequency is: N-S (9,5%), NNE (12,6%), ENE (21,4%), E-W (7,7%), WNW (24,4%) and NW (17,8%). At this stage, the WNW, ENE and NW sets are the most widespread, while the N-S, NNE, and E-W sets are the least common.

b) Group (2) consists of fractures cutting the postglacial lavas (15,000 - 8000 yrs) that surround the young Theistareykir lava (Figure 8(c)). Their frequency is: N-S (17,2%), NNE (14%), ENE (17,2%), E-W (6,1%), WNW (18,6%) and NW (18,5%). At this stage in time, the percentage of N-S fractures has increased drastically, and along with the WNW, NW and ENE sets, they form the four most prominent fracture sets.

c) Group (3) corresponds to fractures mapped in the older series, i.e., Quaternary to upper Tertiary lavas and hyaloclastites (Figure 8(d)). In general, faults in these rocks are the oldest, and the percentage of fresh structures or young open fractures is not high. At this stage, the fracture frequency remains almost the same as in Group (2), with N-S (17,7%), NNE (13,2%), ENE (17,3%), E-W (6,7%), WNW (17%) and NW (19,2%).

This analysis shows two important features. First, the WNW to NW dextral and ENE sinistral strike-slip faults dominate Group (1), and the N-S faults double in the older Groups (2) and (3). This implies that young fractures form under the influence of the transform zone, and only with time the effect of rifting becomes significant. Secondly, regardless of age, the Riedel shears of the transform zone are more widespread than the extensional fractures of the rift.

Surface manifestations consist of soil modified to clay, as well as mineral deposition, all caused by hot water and steam in fumaroles, solfataras, steam vents, mud pools and springs. However, alteration can also be hidden below the soil due to circulation of hot water that has not yet reached the surface. Surface mapping from aerial images [14] (Figure 9(a)) and outcrops [51] [88] [96] [108] show that alteration is localized in Theistareykir, within 4,5 km from the eastern boundary of this fissure swarm (Figure 7 and Figure 9(a)). Almost all fumaroles, solfataras and steam vents are located in the area of the most severe alteration (Figure 9(b)).

The boundaries of the block containing the surface alteration are suggested on Figure 9(a). The block has an overall N15˚E elongation. However, it presents variations along its western edge, which coincide with the structural weak zones numbered on Figure 9(c): the N117˚E dextral oblique-slip segment of the HFF (No. 1); the N46˚E oblique-slip sinistral fault to the northeast of Stórihver (No. 2) and the westernmost N162˚E segment of the Tjarnarás dextral oblique-slip fault (No. 3); two N15˚E normal faults (No. 4) and a short N135˚E dextral strike-slip segment (No. 3) to the west/southwest of the block. Finally, the outer edge of the block to the east could be the last N-S structure stretching from Kvíhólafjöll northwards to Ketilfjall (No. 4).

The same fracture sets of rift and transform zones seem to control the configuration of alteration within the block. Parallel ENE fractures (No. 2b) dominantly control the alteration in the centre of the block. This central zone is also bounded by other WNW (No. 1b), NW (No. 3a) and NNE (No. 2c) segments, and it contains the WNW dextral strike-slip horsetail splay of the Stórihver Fault (No. 1c) in presence of a short N-S fault (No. 4b). The narrower alteration zones to the south of the block could be controlled by N-S (No. 4c) and NW (No. 3c) fractures, while the isolated possible alterations to the north could be aligned on other ENE and WNW segments (No. 1a and 2a). The majority of the fumaroles, solfataras and steam vents appear aligned on these sets and even the same structures as controlling the alteration (Figure 9(d)).

Evidence of upflow and well data support the existence of the “alteration block” and the dominant role of the Riedel shears in geothermal activity:

a) The hottest of the drilled wells, with boiling temperature of 300˚C at 1100 m depth (ÍSOR database) are all within the ENE zone in the centre of the “alteration block”. Even outside this block, the ENE structures play a role in deep fluid flow.

In a well drilled at Stórihver (ÞG-8), the N-S faults are dominant down to 1700 m but total circulation loss is associated with ENE fractures between 1700 m and 1770 m [109] , which is also identified from the surface in this study (No. 5 on Figure 9(c)).

b) In an isolated site west of the “alteration block”, steam (40˚C - 82˚C) escapes from the lava along a short NW line to the south of Stórihver (No. 1 on Figure 9(d)). As surface measurements did not reveal H2S there [51] , the steam is likely from shallow ground water [88] [96] heated by vapor emanating from the geothermal reservoir. The possible structural paths of the vapor from the reservoir could be successively from the western NNE and N-S boundaries of the “alteration block” (No. 2a and 4 on Figure 9(c)), the WNW Stórihver Fault and then the NW fractures. In winter, snow melts exactly along these NNE, NS and NW patches [51] .

c) Temperature logs of ÞG-8 indicate an outflow from the main reservoir into the west of the “alteration block”. The water is 200˚C at 250 - 500 m (MD) depth in the well, but below that, the formation temperatures decrease to 100˚C at 2500 m depth, implying a deep influx of colder water [110] . The structural path for this shallow westward outflow from the reservoir could be the splay of the Stórihver Fault (Figure 9(d)).

Gas monitoring in Theistareykir geothermal field has been ongoing for more than half a century [51] [90] [96] [111] . One of the monitorings used a sufficient number of measurement points and provided both maps of gas concentrations and an assessment of reservoir temperature [51] . The chemical components determined are CO₂, H₂, H₂S, CH₄, Cl, Rn, He, ratio of H₂/H₂S, oxygen isotopes (¹⁸O/¹⁶O ratio), as well as gas temperatures based on CO₂ and H₂S in steam. In this study, the original determinations of H₂/H₂S, H₂S, CO₂, radon and two of the gas temperatures (CO₂ and H₂S) are selected for a correlation with the structural pattern (Figures 10(a)-10(f)).

Overall, five temperature provinces are suggested based on gas geothermometry [93] [112] . The hottest area is interpreted to be to the northeast (Ketilfjall), but temperature decreases along NW parallel isolines from Ketilfjall towards Tjarnarás and Stórihver. An early attempt to interpret the structural control of the concentrations, ratio, and temperatures of the gases suggested that NW and ENE structures could have an influence on a couple of gases [51] . However, the structural map available for that interpretation showed a few N-S rift-parallel fractures and the NW Tjarnarás Fault as the sole structures present above the reservoir [51] .

In this study, the new structural pattern of both the rift and the transform zones (Figures 10(a)-10(f)) is used to interpret the structural control and distribution of several of the gases determined in steam [15] . Although the chemical monitoring data date from long before the new structural map was prepared, the two datasets fit strikingly. They show that major changes in the concentrations of gases and their isolines display a pattern, whose orientation and location coincide with the mapped structures.

Selected examples are the ratio of H₂/H₂S, as well as the concentrations of H₂S, CO₂, Rn, along with the gas temperatures of CO₂, and H₂S (Figure 10). Their isolines show that three Riedel shear sets striking ENE, WNW and NW, along with the N-S extensional fractures control these distributions. Of these, the ENE sinistral weak zone in the centre of the “alteration block” seems to have the most dominant role, followed by the WNW dextral oblique-slip splay segment of the Stórihver.

An in-depth analysis is not at hand to determine at what depths the gases originate before reaching the surface. However, assumptions are that the gases come from a depth of 2 to 7 km, where the CO₂ and H₂ are among the deepest and methane among the shallowest (Ármannsson, Pers. Comm., 2015). Even if the fractures of rift and transform zones have mild dip-slips or even faint traces above the reservoir at the surface, the structures of both plate boundary types are pathways for the emanation of deep gases.

Although not shown here, correlations were also made between the fracture pattern and data from ten of the wells located in the core of the alteration block. The results demonstrate that the main WNW, ENE, NW and N-S fault segments also create permeability for feeders at depth [16] , while the NNE Riedel shears play a lesser role and mostly as secondary fractures.

The recent resistivity survey of Theistareykir and surroundings is a 3D inversion of MT data with a high number of measurement points along N-S and E-W profiles (Figures 11(a)-11(h)). The survey provides revealing information on geothermally altered bodies (or anomalies) down to 12 km b.s.l. [89] . The data show a shallow cap rock down to 400 - 800 m depth, reflecting the zeolite/smectite alteration zone. An underlying high resistivity core corresponds to the chlorite/epidote alteration zone, which rises to 200 m b.s.l. below the Theistareykir geothermal field. Finally, a deeper low resistivity body, possibly indicative of the heat source and upflow zones of geothermal fluid into the system, appears below Ketilfjall, Bæjarfjall and north of Stórihver. Two distinct connected low resistivity bodies also exist to the northwest of the surveyed area (i.e., south of Sæluhúsveggur). They reach down to several kilometres b.s.l. (Figures 11(c)-11(h)), but it is unknown whether they are of alteration or magmatic origin [89] .

The width of individual resistivity anomalies can be from <1 km to ≥4 km. A clarification regarding these widths is necessary before interpreting their structural control. In this study, the resistivity anomalies are interpreted as being bounded by a series of fractures (narrow weak zones) along a given set and not controlled by single fractures. This interpretation stems from observations of numerous altered fault zones in the 1,5 km eroded upper Tertiary crust of west Iceland [25] [106] , the 0,7 km eroded Hreppar micro-plate of south Iceland [26] , and the >2 km eroded series of the Asal Rift in Djibouti [20] .

In those outcrops, alteration is mostly associated with fault zones of oblique-slip faults where the width of the altered zone ranges from 10 m to 50 m along a single fault. It is, therefore, unlikely that the width of the resistivity anomalies in Theistareykir results from fluid flow along a single permeable fracture in the middle of each body.

In the initial investigation [15] , the tectonic settings of the resistivity anomalies were interpreted at depth intervals of 0 m, 500 m, 1000 m, 1500 m, 2000 m, 2500 m, 3000 m, 4000 m, 5000 m, 6000 m and 8000 m b.s.l. by superimposing them on the structural map. Due to similarities of the structural settings, only the depth intervals with significant differences are presented here for discussion (Figures 11(c)-11(h)). Their main interpretations are:

a) Resistivity anomalies display strong left- and right-stepping en échelon arrangements, typical of dextral and sinistral structures and the boundaries of these anomalies coincide with individual structural weak zones.

b) From 0 to 4000 m b.s.l., the resistivity anomalies are controlled by the WNW and NW dextral oblique-slip, and by the NNE to ENE sinistral oblique-slip Riedel shears. The ENE set, however, appears as a clear and dominant set mostly from depths of 2500 to 4000 m b.s.l. (Figure 11(e) and Figure 11(f)). Two WNW segments including the HFF form a deformation zone to the west. In this WNW zone, shorter N-S resistivity anomalies are bounded by rift-parallel structures, much like extensional pull-aparts within Riedel shear zones (Figures 11(c)-11(f)). From 5000 m b.s.l., the E-W set becomes more established. Along with the NW, WNW and the ENE sets, the E-W set controls a major anomaly to the east of the study area (Figures 11(g)-11(h)). At 8000 m b.s.l., the E-W and N-S sets are dominant. The N-S set is confined to the west of the study area, and a few NW structures still bound the deepest resistivity anomalies to the east (Figure 11(h)). It is noteworthy that on maps, the resistivity anomaly below the geothermal area appears clearly only from 2000 m b.s.l., and it is controlled by the Riedel shears down to 8000 m b.s.l. (Figures 11(d)-11(h)).

c) The resistivity anomalies and the structural weak zones controlling them present a gradual clockwise rotation of up to 42˚ to the east between 0 m and 4000 m b.s.l. The most obvious changes take place between 1000 m to 4000 m b.s.l. The WNW lineament rotate to NW/NNW by 42˚ (median from N118˚E to N160˚E), the NNE structures to ENE by 41˚ (median from N23˚E to N64˚E), and the N-S structures only by 7˚ (median from N09˚E to N16˚E) (Figures 11(c)-11(f)). Compared to the strike ranges at 4000 m depth, an anticlock rotation of the structures occurs by 4˚ to 20˚ to the west at 5000 and 6000 m b.s.l. intervals. Thus the NNW/NW structures rotate to WNW by 20˚ west (median from N160˚E to N140˚E), the ENE structures rotate 14˚ west (median from N64˚E to N50˚E), and the N-S 18˚ west (median from N16˚E to N178˚E). The NNE structures are rare and short, and they rotate only 4˚ west (median from N35˚E to N31˚E) within the depth interval of 5000 to 6000 m b.s.l. A couple of E-W structures (N83˚E and N86˚E) bound a major resistivity anomaly to the east of the area. They are a part of an ENE structural zone at 4000 m b.s.l., but at 5000 m b.s.l., their eastern half rotates clockwise after intersecting a NW structural zone (Figure 11(g)). At 8000 m b.s.l., the structures present again a clockwise rotation of 3˚ to 14˚ compared to the strike ranges at 6000 m b.s.l. At the great depth of 8000 m, the deep E-W structure has rotated only 3˚ to the east (median from N86˚E to N89˚E), the N-S structure 5˚ (from N178˚E to N03˚E), and the NW set 14˚ to (median from N140˚E to N154˚E) (Figure 11(h)).

The above interpretation reflects the dominance of the Riedel shears of the transform zone in the alteration process down to 8 km depth in the crust, as well as the concept of rotation of the structures. These rotations are summarized on Figures 13(a)-13(c), along with the strike-ranges of the most common structures at each depth interval (Figure 13(d)).

Figure 12(a) shows the structural weak zones identified regionally, and highlights the interpreted structures controlling the geothermal activity. Two sets of relocated earthquakes from 1993-2011 [83] and 2014-2015 [104] are interpreted in the light of this tectonic setting.

a) The 1993-2011 earthquakes (SIL network) to the west and northwest of Bæjarfjall are with magnitudes of −0,6 ≤ M_l ≤ 3,2 and a depth range of 3 to 7 km. They are grouped into four clusters and interpreted with complex alternative fault motions in each cluster [83] . These clusters are numbered as 1 to 4 on Figure 12(b) where the structural pattern obtained in this study is used as a basis to evaluate which of the alternative fault motions suggested by [83] are most likely. This study interprets the fault motions as follows: In clusters 1 and 2 (Figure 12(b)), the fault plane solutions (FPS) for the NW Tjarnarás Fault and the WNW Stórihver Faults are dextral oblique-slip on both structures. The group of earthquakes on the western slope of Bæjarfjall (No. 3 on Figure 12(b)) coincides with the NNE segments identified there as having a sinistral oblique-slip motion. The cluster of earthquakes distributed to the north of Bæjarfjall (No. 4 on Figure 12(b)) likely aligns on the ENE sinistral oblique-slip weak zone identified there [14] .

b) The 2014-2015 earthquakes are mostly located under the top of Bæjarfjall and to a lesser degree to its northwest (SIL and Landsvirkjun-ÍSOR network) (Figures 12(a)-12(b)). Their magnitude is 1 to 1,9 M_l and their depth 2 to 4 km, although some reach down to 6 km depth [104] . The seismic lineations of these earthquakes likely fall on parallel ENE sinistral oblique-slip segments that form the southern boundary of the main ENE zone within the alteration block [17] (Figure 12(a)). Some of the seismic lineations in Bæjarfjall could also align on tightly parallel WNW segments associated with the Stórihver Fault there [17] (Figure 12(b)). Three of the six available FPS for these earthquakes [104] indicate a dextral oblique-slip on the WNW segment of the Stórihver Fault that cut the main crater of Bæjarfjall (Figure 12(a)). The fit between the hypocentres of these earthquakes [104] and the structures mapped from the surface stems from the fact that ruptures occur on steeply-dipping fractures (i.e., ≥75˚ from horizontal).

Finally, some of the 2014-2015 earthquake activity in Bæjarfjall has been attributed to the discharge of geothermal wells [17] [104] [113] . Even though these are “triggered” or “induced” earthquakes, they still indicate fault slips on existing fractures, most of which are mapped in this study.