To identify the overall structural architecture, the morphostructures from a hillshade map with 5 m contour-interval (database of Iceland GeoSurvey-ÍSOR) and a Digital Elevation Model (DEM) map with a 2.5 m contour interval (Loftmyndir) were analysed. These image types reflect mostly the overall regional and major structures. However, as they do not provide sufficient resolution for finer structures, the analysis was completed using monochrome aerial photographs, as well as orthomaps and Spot 5 images in several channels for enhanced shadow/lights. Pairs of aerial photographs provided stereographic observations and a means to see the structures in relief and assess the dip direction of the fractures. Individual fractures, however, may appear with a slightly different trace

and geometry on each image type, depending on the direction of the light. Thus drawn fractures are “best fit” of their traces on all images.

The number of sets and distribution of the fractures obtained from aerial images and selected outcrops are shown on the enlarged (Figure 2), and detailed views of the study area including mapped surface geothermal manifestations (Figure 3). The rose diagram on Figure 2 shows that the NE and NW fractures are the most frequent in the entire study area, but local statistical analysis may show other dominant frequencies, depending on the location of the studied area within the overall tectonic pattern.

Individual fractures are segmented along their trace, and segments can be as short as a few metre(s) or reach up to 2 km in length. Joints or very young fractures that have not yet coalesced or fully broken the surface have more subtle traces (hair line), particularly in lavas <2400 yrs, and may be mistaken as lava structures or go undetected. The traces of older structures with accumulated slip are stronger both on images and in outcrops.

Regardless of their age or length, shear fractures present en échelon arrangements, which were used to determine their strike-slip motions as discussed

below. This criterion shows that the tectonic fabric is made of purely extensional fractures as well as strike- and oblique-slip faults.

The mapped volcano-tectonic structures are faults (Figures 4(a)-(c)); major joints; rare dykes due to the level of erosion (Figure 4(d), Figure 4(e)); eruptive fissures and craters (Figure 5(b)); surface ruptures of earthquakes (Figure 3(b), Figures 5(a)-(c), Figure 6(a) and Figure 6(b)); and surface geothermal manifestations (Figures 6(a)-(g)). Micro-structures such as mineral veins (Figure 4(f) and Figure 4(g)) and secondary fractures (Figure 5(f) and Figure 5(g)) were also analysed, but striae were not found on any of fracture planes studied.

Six fracture sets emerge from the overall structural analysis: N-S (0˚ - N27˚E), NNE (N28˚ - N40˚E), ENE (N41˚ - N75˚E), E-W (on average N90˚E), WNW (N115˚ - N130˚E), and NW/NNW (N140˚ - N155˚E and N165˚E). In the geothermal field and its surroundings, the ENE, N-S and NW sets are the most common. The E-W fractures are the least common in the entire study area and their traces are very subtle, without convincing evidence as to their shear motion.

Dip-slip is associated with all sets (Figures 4(a)-(c)), ranging from <0.5 m to ≥20 m. The highest slips are along the ENE, N-S and NE faults. Fracture apertures are generally <0.5 m (Figure 6(a) and Figure 6(b)), rarely reaching 1 m in width in the study area.

The NNE/NE fractures parallel to the Reykjanes rift fissure swarm are purely extensional. The en échelon fracture geometries, visible along younger and older fractures, are typically left-stepping along the N-S, WNW and NW dextral faults, and right-stepping along ENE sinistral fractures (Figure 3(b) and Figure 3(c)). The N-S dextral and ENE sinistral strike-slips are the main conjugate source faults of earthquakes and are identical to those found within the pure transform zone of the SISZ farther to the east [94] [120] . The magnitude of offset along the shear fractures could not be identified due to the lack of marker horizons. However, data from the 2008 earthquakes within the SISZ itself indicate that the offset along an individual N-S dextral source fault is about 2 m [95] . In the 15 Ma old lavas of west Iceland, sets of dykes are shifted up to 20 m by strike-slip faults [121] [122] indicating cumulative displacement likely due to fault reactivation during their lifetime.

The source faults of earthquakes in Reykjanes consist of sink-holes, open fractures and push-ups. The mapping of these surface ruptures in the SISZ [94] [120] shows that sink-holes and open fractures are more frequent than the push-up along a single fault. But in this part of Reykjanes, push-ups are by far the most dominant structures both along the N-S and the ENE faults (Figure 3(a), Figures 5(a)-(c), Figure 6(a) and Figure 6(b)), and their significance is further analysed with the earthquakes in Chapter 4.3.2 below.

The only dyke segments cropping out in the study area are at Rauðhólar (Figure 4(d) and Figure 4(e)). Even if they are in soft hyaloclastite host rock, which

could influence their strikes, the mapped dyke segments strike dominantly ENE, and then N-S and NNE, and finally WNW (Figure 4(d)), reflecting the strikes of the faults (Figure 2).

Due to minimal level of erosion of the young series, deeper parts of fault zones do not crop out for structural analysis. However, mineral veins and secondary fractures were measured in rare sites.

Two sets of mineral veins cut the hyaloclastites of Rauðhólar above the geothermal reservoir (Figure 4(f) and Figure 4(g)) in the same outcrop as the mapped dykes (Figure 4(d)). One set consists of shallow dipping (0˚ - 45˚) veins parallel to the stratification of the hyaloclastites (Figure 4(e)). As some panels of hyaloclastites and veins are tilted by collapse, the shallow-dipping veins are rootless and do not represent the initial state of fracturing. A second set of veins dipping 45˚ to 90˚ is also present (Figure 4(f) and Figure 4(g)). Their aperture is from 1 mm to 3 cm, and their length from 1.5 to 3 m. About 78 such veins were measured (Figure 4(f)) in an area of 300 m × 50 m. The total vein population strikes dominantly N30˚ - N40˚E and parallel to the rift fissure swarm, although locally the veins strike N10˚ - N30˚E, N50˚ - N60˚E and even N110˚ - N140˚E similar to the shear fractures. A few vein cross-cuttings occur, but their order is not coherent enough to conclude a chronology of emplacement. A red contact aureole appears along a handful of veins (Figure 4(g)). Although these veins were not sampled, magma injection was excluded as the reason for the aureoles because these veins are at a significant distance of the mapped dykes at Rauðhólar. Localised alteration along the veins from fluid flow is assumed to cause the aureoles.

Another rare site to the west of Mölvík displays joints and secondary fractures in association with two series of N-S and NW-trending faults (Figures 5(d)-(g)). Three parallel N-S faults by the shoreline strike N22˚E and bend gently to N28˚E in their northern parts. They displace the basaltic lavas by 1 to 2.5 m (Figure 5(e)). About 68 joints measured along these faults (Figure 5(d) and Figure 5(e)) strike dominantly N20˚ - N30˚E, then N10˚ - 20˚E, and to a lesser extent N40˚ - 50˚E and N70˚ - N80˚E. Only a few joints strike N100˚ - N110˚E and N150˚ - N170˚E. Some of the joints display geometries typical of shear motions (Figure 5(f)). For example, a metre-scale block of lava is bounded by two deformation bands striking N20˚ - N30˚E parallel to the main faults. Each band is at least a metre thick and consists of unbroken lavas alternating with parallel brecciated zones where small dip-slip faults with up to 7 cm vertical displacement are present (Figure 5(f)). The central block of lava between the two deformation bands is cut by series of joints trending at 50˚ oblique to the deformation bands (Figure 5(f)). These joints strike N40˚ - N60˚E and show a sinistral horizontal offset of 5 cm. They are cut by other joints striking N130˚ - N140˚E (Figure 5(f)). The overall structural settings here reflect Riedel shears where the N-S deformation bands act as dextral strike-slip (R’) and the ENE set as sinistral strike-slip (R), much like the conjugate sets of surface ruptures in the SISZ.

Finally, at a few metres from this outcrop, a short fault segment strikes N163˚E and dips near vertical, displacing the lava by 1 m to the west (Figure 5(d) and Figure 5(g)). No breccia is associated with the fault and several regularly-spaced joints, striking N160˚ - N170˚E and N110˚ - N130˚E crop out in the hanging wall of this NW fault (Figure 5(g)). The NW fault is the continuation of a ridge similar to a horst, which has a km length and stretches northwest to Sýrfell (Figure 2 and Figure 3(a)). The edges of this ridge are gentle and covered by sand. The northeast and southwest scarps strike N139˚E and N144˚E and stand higher than the surrounding by >5 m and >3 m, respectively. The ridge could be misinterpreted as a lava structure. However, the freshly ruptured NW faults (Figure 5(g)) and joints (Figure 5(d)) by the shoreline are in the continuation of this ridge, suggesting that the structure is a horst and that it has been activated recently. Similar NW structures with left-stepping segmentation typical of dextral strike-slip are relatively common in the entire study area.

These qualitative observations from rare outcrops of secondary fractures show that the rift extensional fractures and the Riedel shears of the transform zone are both active even at the micro-structural level, which is relevant for fluid flow in the geothermal reservoir.

Surface geothermal manifestations crop out only above the reservoir (Figure 3(a)). All surface alteration was mapped, along with the fumaroles, steam vents and mud pools within the alteration zones. Only one extinct silica sinter was found. It dates from 1919 and is located in the southern part of the alteration area. Additionally, soil temperatures were measured with a hand-held thermometer to a maximum depth of 1 m. The structural control of all these manifestations was also assessed [65] . It is emphasised that geothermal activity changes rapidly and may increase or disappear with time. The interpretation discussed here is based on the data gathered during the summer of 2014.

Mapped ground temperatures were as high as 91˚ - 100˚C and as low as 11˚ - 21˚C (values <11˚C measured in the summer are not representative of geothermal heat). Surface alteration was mapped by visual inspection and ranged from very mild (Figures 6(a)-(c)) to intense (Figure 6(d)-(g)). The mildest alteration appears as unaltered rock with occasional steam escaping from lava or surface ruptures (Figure 6(a) and Figure 6(b)), leading to greener moss (Figure 6(c)) due to warm ground (<40˚C) and humidity.

In more intensely altered areas (Figure 6(d)), ground temperature is 60˚C to 80˚C, and local white/red/orange clay is widespread along with small fumaroles and steam vents. In areas of intense alteration where ground temperature is 80˚C to <100˚C, the rock is entirely altered to clay, and fumaroles are more frequent and large. Mud pools are quiescent to vigorously boiling and vegetation is totally absent (Figures 6(e)-(g)). Fumaroles and mud pools are generally <1 m in diametres except a few mud pools that are up 1.5 m in size (Figure 6(f)). At the time of mapping, Gunnuhver was the biggest fumarole and strongly venting water and steam from craters of 2 to 2.5 m diametre in size (Figure 6(g)).

Based on the above observations, alteration is moderate to the north of Lón, mimicking the lava lobes, but more intense to the east, southeast, and mid central area to the south of Lón (Figure 3(a)).

Several structural weak zones are interpreted within each fracture set (Figures 7(a)-(f)). These zones appear either as areas of high fracture density, or as single fractures that are prominent compared to their surroundings due to their trace, length, geometry, or higher slip. These zones reflect where most of the deformation is concentrated and fault reactivation occurs.

Individual N-S dextral, ENE sinistral, and WNW to NNW dextral weak shear zones lie parallel to each other with a relatively regular spacing (Figure 7(a), Figure 7(c), Figure 7(e) and Figure 7(f)). The NNE extensional rift structures are, however, fewer and occupy specific locations within the study area. These NNE zones consist of two grabens with 1 to 1.5 km width at Haugur and Vatnsstæði (Figure 4(c)), then a series of narrow structures from Skálafell to Rauðhólar and Sýrfell (50 m width), and a series of narrow grabens at Kinn (<100 m width) that merge with ENE structures there (Figure 7(b)). Of the fracture sets, the E-W is the least frequent and has the shortest and narrowest structures. The structural weak zones of the other five sets are the most prominent, and play a role in geothermal activity as thoroughly discussed in the following chapters:

a) The four westernmost N-S weak zones stretching from Skálafell to north of Lón at Rauðhólar, from east of Skálafell to the northwest of Sýrfell, from the coast to Melur, and from Langihryggur to the northeast of Sýrfell (Figure 7(a)). The most prominent N-S faults within these zones are series of parallel dextral oblique-slip faults with <10 m vertical slip covering the area from Langihryggur (Figure 4(b)) towards the geothermal reservoir where a number of young surface ruptures and most of the push-ups (Figure 5(c)) are aligned on them (Figure 3(a)).

b) The NNE segments of Sýrfell/Sýrfellsdrög/Rauðhólar (Figure 7(b)) on the trace of which the only dykes were observed (Figure 4(d) and Figure 4(e)) and mineral veins measured (Figure 4(f) and Figure 4(g)).

c) The six southernmost ENE weak zones starting from Sýrfell/Sandfellshæð

extend all the way to the shoreline. These zones are at Litla-Vatnsfell, from Melur to Rauðhólar, from south of Melur to north of Skálafell, from Klofningar to Háleyjabunga and Skálafell, and that of Mölvík (Figure 7(c)). The best example of ENE sinistral oblique-slip faults belonging to these zones are the fault segments from Klofningar to south of Skálafell, displaying a systematic right-stepping en échelon arrangement, along with some of the highest metre-scale dip-slips (Figure 4(a)).

d) Three short WNW weak zones between northwest of Sýrfell and north of Rauðhólar, stretching to the shoreline (Figure 7(e)).

e) Four NW/NNW weak zones, stretching from land to shore, crossing Litla-Vatnsfell/Skálafell, Rauðhólar, Sýrfellsdrög to Melur and Háleyjabunga (Figure 7(f)).

The left-stepping en échelon arrangement of the WNW to NNW dextral oblique-slip segments is clear on maps but faults of these weak zones have <5 m dip-slip on average, which is among the least dip-slip values.

The significance of the fracture population for the style of deformation, and the structural setting of eruptions are analysed here.

a) The most relevant structures of each of the six sets (Figures 7(a)-(f)) were combined and highlighted on a single structural map (Figure 8(a)). The emerging tectonic pattern is clearly that of an ENE Riedel shear zone with a higher number of sets and fractures than previously mapped on RP [16] . Two adjacent shear zones were observed and labelled as the Northern (NRSZ) and the Southern (SRSZ) Riedel shear zones (Figure 8(a)). Each of these zones is at least 3 km wide but their total dimension is unknown due to the limits of the study area in RP. The NRSZ has an overall strike of N74˚E and the SRSZ N64˚E. The northern boundary of the NRSZ coincides with a blend of NNE extensional faults and ENE sinistral oblique-slip faults crossing Kinn. The southern boundary of the NRZS and the northern boundary of the SRZS is the same. This boundary falls on the trace of very subtle ENE fractures that stretch from Sandfellshæð to Sýrfell and have a possible sinistral character (Figure 8(a)). The southern boundary of the SRSZ could be the ENE lineament passing through Mölvík. There, fresh ENE surface ruptures indicate recent reactivation (Figure 5(a)). In the NRSZ, the N-S set is lacking and there are fewer fractures belonging to the other sets, while the SRSZ is populated by a large number of fractures belonging to all the six sets (Figure 8(a)). The geothermal reservoir is in the western portion of the SRSZ. In the NRSZ, the Haugur Graben seems to stop on a smaller WNW fault to the north of Sýrfell, which is one of the reasons why this rift structure does not appear to continue as a wide graben in the SRSZ.

b) The second striking feature is the structural pathways of eruptions. The diametres of the eruptive edifices range from roughly a metre such as east of Rauðhólar (Figure 3(a) and Figure 5(b)) to 100 - 150 m for the crater at Sandfellshæð (Figure 2 and Figure 8(a)). Eruptions are generally considered to occur on rift extensional fractures in Iceland [88] , and on RP [16] . Consequently, the main eruptive dyke and the possible heat source of the geothermal reservoir is likely to be attributed to the NNE structure of Rauðhólar/Sýrfell. However, the mapped dykes at Rauðhólar strike dominantly ENE, and secondarily N-S, NNE and WNW (Figure 4(d)), and are thus a mixture of extensional and shears fractures. Furthermore, a closer analysis of the structural map shows that the crater of Melur is on the trace of a N-S dextral oblique-slip fault, and the smaller craters to the east of Rauðhólar are aligned on an ENE lineation in the southwestward continuation of the ENE push-ups. The craters and the push-ups are arranged in right-stepping ENE en échelon segments similar to sinistral conjugate source faults of earthquakes (Figure 3(a)). Similarly, farther to the east, the eruptive fissures of Klofningar in Eldvörp are aligned ENE with a clear right-stepping en échelon arrangement in the continuation of the ENE oblique-slip fault passing through north of Mölvík and Háleyjabunga, and south of Skálafell (Figure 2 and Figure 8(a)). These examples show that an important number of the RP eruptions occurs along non-rift structures.

The concept of oblique rifting on RP where extension and shear motion occur in equal proportion was introduced by [123] , although opinions diverge regarding the detailed dynamic behaviour. These fault modes are explained by permutations of the axes of two principal stresses in which normal faulting occurs when the maximum principal stress (σ₁) is vertical, and strike-slip faulting when the intermediate principal stress (σ₂) is vertical [124] . The minimum principal stress (σ₃) remains the same in each case and is perpendicular to rift segments. Based on this idea, it is suggested that oblique rifting and magmatic phases alternate in RP, and as the last eruption dates from the 13th century [106] , RP presently undergoes oblique rifting [103] . Accordingly, the maximum compression is vertical during magmatic phases, shifting to horizontal (NE-SW) during oblique rifting and strike slip faulting, with the minimum compression perpendicular to the rift (NW-SE) in both phases. However, the earthquake swarms of 1972 [107] and 2013 [116] [125] demonstrate that normal, strike-slip and oblique-slip fault ruptures occur during the same earthquakes in a-magmatic phases (Figure 10(a) and Figure 10(b)). Stress inversion of focal mechanisms shows that the average direction of (σ₃) on RP is N120˚E, fitting well with the directions of the greatest extensional strain rate [126] . Analyses of aerial images and clay models of RP conclude that the predominant strike of eruptive fissures, non-eruptive fissures, and normal faults in the centre of the rift zone is perpendicular to N133˚E [16] and not to the plate spreading direction at N103˚E [64] .

The observations presented here shed new lights on the above discussions. The major (Figure 8(a)) and secondary fractures (Figure 8(b)) mapped in this part of RP are dominantly shear fractures, combined with one set of extensional fractures, similar to the SISZ [127] and its immediate zone of influence in the Hreppar micro-plate [120] . In SISZ and the mapped area in RP, the N-S dextral and ENE sinistral conjugate source faults act respectively as (R') and (R), and the NNE set as (T), which is also parallel to rift. These sets along with the WNW/NW and E-W fractures accommodate together the sinistral motion of the transform zone. The stress field required to explain such a structure is more complex than previously discussed, but a part of the deformation can be explained kinematically. Four points reflect this:

a) In RP, dykes are observed to inject into NNE rift extensional fractures as well as into N-S, ENE and WNW shear fractures. Magma injection into shear fractures has been reported from other areas of Iceland under the influence of rift and transform zones such as west Iceland [86] , Theistareykir onshore [128] , and GOR offshore north Iceland [62] . Magma injection into shear fractures is relatively frequent although this mechanism receives less attention than magma injection into rift structures.

b) Within the SRSZ, the Litla-Vatnsfell and Rauðhólar/Sýrfell structures present a bend in their orientation from NNE to ENE. Rauðhólar/Sýrfell has two parallel narrow segments and Litla-Vatnsfell displays a single fault trace. Both structures have local metre-scale dip-slip and en échelon arrangements typical of sinistral motion. In case of Rauðhólar/Sýrfell, the segmentation could also result from the intersection with some of the WNW/NW weak zones (Figure 8(a)). The strikes of these structures change from N35˚E to N47˚E and to N54˚E for Litla-Vatnsfell, and from N25˚E to N35˚E to N50˚E for Rauðhólar/Sýrfell. This gradual change reflects a clockwise rotation of the rift structures at the approach of the ENE sinistral boundary between the two Riedel shear zones. Rotations of rift across transform segments are not rare and are also seen in north Iceland [129] and on the Iceland shelf to the south of RR [130] .

c) Further discussion on the direction of (σ₃) is necessary. Under a regional spreading at N103˚E [64] , and assuming that (σ₃) is perpendicular to the rift fissure swarm, pure extensional fractures should strike N13˚E (Figure 8(c)), but this study and that of [16] show that fractures striking N13˚E are dextral oblique-slips. In fact, none of the shear fractures observed in this study can be explained by a single (σ₃) perpendicular to rift because the N-S, ENE, WNW and NW sets have a wide strike-range, requiring a (σ₃) at N113˚E to N128˚E (Figure 8(e)). The suggested variation in the direction of (σ₃) is compatible with the results of [16] and [126] , and also observed across SISZ [127] . Along with permutations in (σ₁) and of (σ₂), these results indicate that fluctuations occur also in the direction of (σ₃).

d) The significance of the WNW to NNW fractures in Reykjanes is puzzling and their existence cannot be explained by stress fluctuations. The two sets act as secondary fractures within the SISZ where they display a possible sinistral motion [127] . In Reykjanes, these sets do not appear as second-order fractures because they have prominent trace and length and they display a dextral motion (Figure 8(e)). Prior to this study, a few WNW to NW fractures were pointed both on RR near Reykjanes [131] and onshore near the study area [111] , and attributed to transfer/transform zones. However, these sets cannot belong to a second active WNW-NW transfer/transform zone since the transform zone strikes ENE on RP (Figure 1(b)). Furthermore, major WNW and NW structures are present as far north as in Snæfellsnes Volcanic Zone [121] and in the older lavas of west Iceland [122] . There, the structures are attributed to a Tertiary transform zone that was reactivated during the Quaternary [85] [86] . Although the origin of these sets cannot be resolved with the data at hand, a likely kinematic explanation could be that they act as horsetail splay structures (Figure 8(e)).

The surface alteration, fumaroles and mud pools, as well as soil temperature distribution are interpreted in light of the structural pattern (Figures 9(a)-(c)).

The mapped alteration above the reservoir is organised in two overall groups. The rare high alteration spots to the north of Lón are arranged in a left-stepping en échelon array along a N-S line similar to N-S dextral strike-slip surface ruptures (Figure 3(a) and Figure 6(d)). This N-S zone extends to the south of Lón where it is wider and it separates the areas of intense and high alterations (Figure 9(a)). The alignments and the changes in the degree of alteration in these zones coincide mostly with the ENE sinistral, N-S and NW dextral structures. In particular, the intense alteration occurs dominantly along the ENE set, while the high alteration is controlled by the NW set and by series of tightly parallel N-S strike-slip faults crossing the Lón from north to south (Figure 9(a)).

Although much smaller in size, the fumaroles, mud pools, and steam vents similarly appear on or at the intersection of fractures (Figure 9(b)). The most prominent fumaroles align on five parallel N-S segments, two of which stretch into the described N-S “high alteration zone” to the north of Lón. Other fumaroles, mud pools, and steam vents line up on the same ENE sinistral (Figure 6(e)) and NW dextral fracture segments of the alteration zones. A few align on the NNE segments of Rauðhólar (Figure 9(b)). Such manifestations are absent west of the Litla-Vatnsfell Fault, and the last steam vents to the east of the study area are aligned on the ENE push-ups and N-S sink-holes of the surface ruptures and not on the small crater row of east Rauðhólar (Figure 3(a), Figure 6(a) and Figure 6(b)). The ENE sinistral and N-S dextral segments are likely the structural barriers of the reservoir to the east and the last permeable structures into which the steam from the reservoir can circulate.

The highest soil temperatures (91˚ - 100˚C) appear in small spots to the north and east of Lón but mostly occur to the south of this lagoon (Figure 9(c)). Except for the N-S structures crossing the middle of Lón, temperature distribution is controlled by the same fracture sets as those governing the alteration and fumaroles. However, the dominant sets playing a role in the temperature distribution are different. The main areas with highest temperatures (91˚ - 100˚C) are constrained by the NW structures and the NNE segments of Rauðhólar. These structures isolate a small block where temperatures are 10˚C to 20˚C lower. The cooling of the soil beyond the central hot core seems to occur on the same structures as above, i.e., the Litla-Vatnsfell Fault to the west, and the ENE sinistral segments acting as the last structure where steam escapes from the reservoir to the east (Figure 6(a), Figure 6(b), Figure 9(b), Figure 9(c)). Additionally, the dominant WNW boundaries are those to the south of Rauðhólar and southwest of Lón (Figure 9(c)). The monitoring of soil temperature above the reservoir using a grid of 25 m interval [132] [133] shows a rapid fluctuation of surface temperatures between years. The suggested fracture sets controlling the surface geothermal manifestations are evident in the yearly independent monitoring.

Four earthquake swarms occurred in Reykjanes since the 1960’s. The 1967 events had a magnitude up to M_L4.9 [134] [135] , and 5 - 8 cm oblique-slip on the fault at Valbjargargjá (Figure 2) [136] [137] . The 1972 swarm had 17.000 events, mostly at 2 - 5 km depth and M_L up to 4 [107] . The 2006 swarm occurred south of Reykjanes geothermal field offshore, and that of 2013 took place in the same area and was well recorded by the network of HS Orka-ÍSOR and that of the Icelandic Meteorological Office (SIL). GPS and InSAR [119] revealed co-seismic displacements up to 3 cm in the 2013 events, but there was no significant influence on the geothermal field.

Although the 1972 and 2013 earthquakes have not been subject to relative relocation of hypocenters, the correlation of their seismic lineations (Figure 10(a)) and fault plane solutions (FPS) (Figure 10(b)) with the structural map (Figures 10(c)-11(d)) yields critical information:

a) The epicentral zone of the 1972 earthquake swarm is ~1.2 to ~2.6 km wide and its overall ENE trace coincides with the area separating the NRSZ and the SRSZ (Figure 10(a), Figure 10(c), Figure 10(d)). This swarm acts as the boundary of the two ENE Riedel shear zones. The subtle N65˚E fractures with possible sinistral motion to the north of Sýrfell could be the discreet surface expression of this deep earthquake swarm (Figure 10(a) and Figure 10(c)). The 2013 swarm is contained in a ~1.5 km wide band that includes the suggested southern boundary of the SRSZ (Figure 10(d)), including the fresh ENE ruptures found during the 2014 field mapping (Figure 5(a)). Analysis of the 2013-2015 earthquakes [117] reveals an aseismic body under the geothermal field (Figure 10(a)) above the brittle-ductile boundary of 6 km. The body is attributed to episodic activity, pore pressure reduction due to production, or to a hot body (>600˚C ± 100˚C) at depth.

b) The epicentral zone of the 2013 earthquake swarm is short and strikes N60˚E, but that of the 1972 swarm could be interpreted as striking ~N60˚E up to Haugur Graben and bending to ~N80˚E from there to Klofningar with two clusters of earthquakes shooting off its middle (Figure 10(c)). Seismic lineations align on three sets within the 2013 swarm: (1) ENE offshore; (2) WNW-NW offshore in the continuation of the dextral weak zones of the SRSZ (Figure 5(d) and Figure 5(g)); (3) N-S from sea towards Haugur Graben on the trace of the weak zone of Langihryggur (Figure 10(c)). Within the SRSZ onshore, seismic lineations align on two additional N-S dextral oblique-slip faults. They are on some of the N-S faults with push-ups to the east of Rauðhólar, and the N-S structure crossing the middle of Lón (Figure 10(c)). These seismic lineations appear on faults, which have already been interpreted as weak zones based on structural features (Figure 7, Figure 8(a) and Figure 10(d)). In addition to the lineations coinciding with the subtle ENE sinistral fractures of north Sýrfell, earthquakes do cluster on three additional structures within the 1972 swarm. Clockwise on Figure 10(c), one of the clusters appears as a short E-W/WNW lineation immediately to the north of Sýrfell. It is in the continuation of an E-W fracture mapped farther east of Sýrfell. A second earthquake cluster protrudes with an overall WNW elongation farther north and seems to be bounded by the mapped WNW dextral strike-slip segments there. The third cluster is the most prominent and aligns N25˚E at the exact location of Haugur Graben (Figure 10(c)).

c) The FPS of the 1972 swarm are an average of several earthquakes, and their suggested fault traces [107] fit our mapped structures (Figure 10(b)). These FPS are normal, strike- and oblique-slips, in absence of reverse-slip. When strike- and normal-slip FPS are adjacent, strike-slips are interpreted as being above 3 km depth and normal-slips below that depth [107] . The 2013-2015 FPS are at 1.6 to 6 km depth, with M_L 0.61 to 3.08, oblique-slip is rare but three reverse-slip FPS exist [117] [125] . Strike-slips in the FPS are on N-S dextral and/or E-W sinistral fractures, while normal and reverse-slip occur on N10˚E, N40˚E, N110˚E, and N155˚E planes (Figure 10(b)). No apparent relation exists between the depths, magnitudes and mechanisms of these FPS. The FPS from 2013-2015 were accurately positioned on topographic map but without suggested fault traces [117] [125] . The location and motions of 23 of the 2013-2015 FPS correlate well with the strike and motions of the mapped fractures (Figure 10(b)). Furthermore, the correlation of all FPS with the mapped structures shows that the 1972 earthquake swarm occurred only in area (A) in the NRSZ, as well as on the ENE eruptive fissure of Klofningar (Figure 11(a)). This confirms the suggestion that the eruption there occurred on oblique-slip fault segments. The distribution of the 2013-2015 earthquakes, on the other hand, is influenced by three of the major weak zones (Figure 11(b)). A blend of dip-slip and strike-slip FPS are in areas (A1) to the west of Rauðhólar in the SRSZ. Almost all of the dip-slip FPS are located in area (A2) of the NRSZ. On the other hand, all reverse-slip and almost all strike-slip FPS are concentrated in area (B) of the SRSZ (Figure 11(b)). These dominant types of FPS are clearly separated by the major structures identified earlier, i.e., the ENE boundary between the NRSZ and the SRSZ, the rotated segments of Rauðhólar/Sýrfell, and the western portion of the ENE fault to the south of Lón (Figure 11(b)).

The above correlations (Figure 10) with earthquakes allow five additional seismo-tectonic interpretations of regional structures (Figure 11).

a) The existence of the NRSZ and SRSZ and their roles in the distribution of earthquakes and their FPS suggest that the transform zone consists at minimum of two adjacent ENE Riedel shear zones. Consequently, the transform zone is at least 7.5 km wide (Figure 10(c)), and is much wider than previously thought. The SRSZ is presently the most seismically active part of the transform zone as indicated by the 2013 swarm.

b) The epicentral zone of the 1972 earthquakes can be explained without a bend and as a single zone of N70˚E with an average width of 1.3 km (Figure 11(c) and Figure 11(d)). Most earthquake clusters are within this ENE zone, including those of Klofningar-Eldvörp, which seem organised in right-stepping en échelon segments indicative of sinistral motion (Figure 11(c)). Within this ENE zone, local earthquakes also align on short WNW planes (Figure 11(c)). A rare example of reactivation of a rift structure by the transform zone appears to be the NNE earthquake clusters shooting off the ENE zone into the NNE Haugur Graben. As this earthquake cluster is confined to the NNE graben, it indicates that during episodes of transform faulting, earthquakes “leak” into extensional rift structures and reactivate them (Figure 11(c)).

c) In addition to the individual weak zones interpreted earlier, one particular WNW structure also contributes to block compartmentalisation, but at a more regional scale (Figure 11(c) and Figure 11(d)). Haugur Graben merges with the ENE shear zone of 1972 earthquakes and the southern border of the graben stops on one of the described WNW seismic lineations to the north of Sýrfell (Figure 11(c)). To the southwest of this WNW structure, another parallel subtle WNW seismic lineation exists and coincides with the trace of many of mapped WNW weak zones that extend from Sýrfell southeast to Mölvík (Figure 11(c)). This particular WNW structure separates the highly fractured Block (a) from the lesser deformed Block (b). The geothermal reservoir lies in the heavily broken Block (a) within the SRSZ.

d) The bend of the Rauðhólar/Sýrfell segments and the Litla-Vatnsfell Fault from NNE to ENE clearly occurs across the ENE boundary between the two Riedel shear zones and is compatible with sinistral motion of this structure. Compression is expected in association with such clockwise rotation. The evidence for this compression appears as the dominance of push-ups concentrated in the eastern compartment of the rotated Rauðhólar/Sýrfell structure as well as the location of the three reverse-slip FPS of the 2013-2015 earthquakes (Figure 11(b) and Figure 11(d)). Reverse-slip in Iceland is not a result of regional compression by shortening but occurs due to local bends of steeply-dipping normal faults or in association with dyke injections [138] . Additionally, about 20% - 30% of seismic events in the western part of RP show reverse-slip mechanism, and have been explained as displacements on non-optimally oriented faults or block rotations [139] . The localised push-ups and reverse-slip FPS in the SRSZ show an example of compression associated with such rotation. More rotated rift structures similar to Rauðhólar/Sýrfell and Haugur Graben should be present on RP within the transform zone (Figure 11(c) and Figure 11(d)).

e) A wide graben similar to Haugur is absent from the SRSZ likely because the amount of sinistral shift in the NRSZ has been larger over hundreds or thousands of years than in the SRSZ. This greater displacement in the NRSZ led to graben formation. The GPS measurements of a 6-year period (2000-2006) indicate the direction of differential horizontal movement across the transform zone [104] , but this time period is much shorter than the geological time necessary for the graben formation.

Along with an overall increase in the CO₂ flux [115] , there has been a steady increase in the extent and activity of surface geothermal manifestations. This increase occurred since the start of the power plant, but can also be attributed to seismicity [115] [132] . As an example of yearly fluctuations, CO₂ flux changed from 51.4 tons in 2013 to 78.5 tons in 2014 [115] , and steam flow estimated from all fumaroles increased from 0.82 kg/s in 2004 to 6 - 7 kg/s in 2007 [132] . After an episode of earthquakes in 2014, a few mud pools erupted and a powerful fumarole appeared to the west of Gunnuhver. Based on local well discharge and visual comparison with the steam discharge of Gunnuhver in 2014, the steam flow of the new fumarole was estimated at 18 - 19 kg/s.

Surveys of 2005, 2006, 2008, and 2011 show the CO₂ flux to the south of Lón (Figures 12(a)-(d)). The measurements of 2012 and 2014 included an area to the north and east of the lagoon (Figure 12(e) and Figure 12(f)). Despite minor changes in the shape of the areas, the overall changes in the CO₂ flux in 2005, 2006, 2011, 2012, and 2014 [115] occur within blocks bounded by tectonic fractures. These blocks are bounded dominantly by WNW dextral and ENE sinistral strike- and oblique-slip fractures, and secondarily by the N-S dextral strike- and oblique-slip faults and the NNE extensional structures of Rauðhólar. Even if the measured CO₂ flux is more subtle in 2008 (Figure 12(c)), the same structural control is visible among the faint distributions of the gas during that year.

Overall, these are the same structures as identified locally and regionally from other criteria above (Figure 8(a)). As example, the N-S, NNE, ENE, and WNW-NW controlling the surface geothermal manifestations, temperatures and CO₂ flux (Figure 9 and Figure 12) also ruptured during the 2013 swarm as shown by the FPS of area (A1) (Figure 11(b)).

Prior to this structural analysis of tracer recovery, alteration and temperature from a fewer wells than in the present study were used to determine the internal reservoir characteristics [111] . Those results indicated a shallow (200 m) NE upflow zone stretching from southeast of Lón to Sýrfellsdrög, and a N-S structure crossing the middle of Lón. These two structures were also suggested by TEM resistivity anomalies [140] .

To study the flow channels and fracture permeability in the reservoir, tracers were injected into two production wells, i.e., 2,7-NDS into well RN-20b in August 2013, and 2-NS into RN-33 in January 2014 (Figures 13(a)-(d)). At the time of these investigations, the tracer from RN-20b took between 77 and 133 days to reach the monitored wells, and that from RN-33 between 24 to 333 days (Table 1, Figure 13(a) and Figure 13(c)).

Data from the tracer experiment were summarised by [141] , supporting the existence of the hypothesised N-S and NE zones. From these studies, the N-S zone was deemed semi-permeable, and the NE permeable structure was placed slightly north of that suggested by [111] . Both these interpretations were made without the benefit of the present structural interpretation.

In this study, three datasets were used for the structural interpretation of tracer flow paths. These are: (a) results of the distribution and timing of tracer arrival (Table 1) in wells [118] , (b) the pressure drawdown at 1600 m depth (Figure 13(a) and Figure 13(c)) calculated for the period of 2008 to 2012 (courtesy of Ómar Sigurðsson, HS Orka, 2014), and (c) the structural map. The tracer is

assumed to have entered the formation at the depths of 1180 m (MD) in RN-20b and 2250 m (MD) in RN-33 [118] . The rate of tracer propagation was calculated from its first arrival in wells. Overall, tracer moved faster from RN-33 than from RN-20b (Table 1), and a large part of the injected water (and tracer) dispersed and diffused through the reservoir outside of direct flow channels [118] .

The structural interpretation of tracer flow paths took into account these criteria: travel time, sequence of tracer arrival from injection well to the receiving well, the possible directions for such flow, the amount of tracer recovered, depth of the receiving feeders, the buoyancy of tracer, and the pressure drawdown. Where a receiving well had more than one feeder, the shallower feeder was chosen (Table 1) because colder injected tracer water is quickly heated by the hot formation and, therefore, drawn laterally and upwards into the zone of lowest pressure.

Figure 13(a) and Figure 13(c) summarise the travel paths and arrival sequences of tracer from RN-33 and RN-20b to individual wells. Figure 13(b) and Figure 13(d) show the interpreted structures that likely facilitated (or hampered) tracer flow based on their strong, moderate or weak concentration in the receiving wells (Table 1).

The tracer from RN-33 flowed from east-northeast towards the west-southwest and appeared successively along paths (1) to (5) without going farther than east of Lón (Figure 13(a) and Figure 13(c), Table 1). The tracer likely travelled first along two sets of carrier structures, i.e., the NNE segments of Rauðhólar/Sýrfell and a NW-striking fault, resulting in a high concentration and rapid appearance in wells just north of Rauðhólar (Figure 13(a) and Figure 13(b)). The fact that the tracer appeared later and in less concentration just to the south of Rauðhólar may be explained by the presence of the ENE fault along the trace of the push-ups. This ENE fault likely dips to the southeast thus acting as barrier to flow from RN-33 coming from the northeast. The longest tracer travel time and minimum concentration occurred in well RN-15 located in the footwall of the NW-striking fault. The explanation for these occurrences is that the NW fault likely dips to the northeast. This structure facilitates fluid flow on its hanging wall but hampers the flow to its footwall. A second possible cause could be that the westernmost NNE segment of Rauðhólar/Sýrfell dips to the northwest and acts as a structural barrier to the tracer flow coming from the northeast (Figure 13(a) and Figure 13(b)). Finally, no tracer was found in the western half of the geothermal field. It is suggested that there is a N-S structure dipping to the west crossing the middle of Lón and hindering tracer flow. This barrier is in the same location as the series of N-S segments identified by the analysis of surface geothermal manifestations and CO₂ surface flux (Figure 9, Figure 12, Figure 13(a) and Figure 13(b)).

As to the test from well RN-20b (Figure 13(c)), the tracer flowed from east to west and northwest, appearing successively along paths (1) to (9). Evidence of tracer was not detected farther west than the middle of Lón (Figure 13(c) and Figure 13(d), Table 1). The tracer first travelled along a WNW segment (path 1) and arrived quickly and in high concentration at the hanging wall of the ENE fault (Figure 13(c) and Figure 13(d)). This is the same ENE fault that acted as a barrier in the case of the flow from RN-33. The tracer from RN-20b then travelled southwest (path 2) where it appeared with lower concentration in well RN-28. This well is located in the hanging wall blocks of both the ENE fault and the NNW segments of Rauðhólar. The tracer travelling along the paths (3) to (7) and (9) to the west and northwest arrived with much lower concentration and later in wells located in another compartment bounded by the NNE Rauðhólar/ Sýrfell, the ENE fault and the N-S barrier (Figure 13(c) and Figure 13(d)). The weak tracer signals there are likely caused by the same structural features as mentioned above, which are the dip directions of the ENE fault towards the southeast, and that of the NNE segment of Rauðhólar to the west, both acting as semi barriers. The flow to this compartment, however, could have occurred along another WNW segment belonging to the carrier structure of path (1). Similar to RN-33, no tracer from RN-20b was found in the wells located on the western half of the geothermal field and the reason is likely the existence of the N-S barrier structure there.

These analyses indicate that the same structure can act either as a carrier or barrier to flow, depending on the provenance of incoming flow and the dip direction of the fault.

In this section, the structural control of the pressure drawdown, the subsidence revealed by InSAR, and the direction of fluid recharge towards the reservoir (Figure 14(a) and Figure 14(b)) are analysed.

a) The geothermal reservoir has been subject to a pressure decline due to production, which value was assessed for the period of 2008 to 2012 (Figure 14(a)). The drawdown at 1600 m depth corresponds to the difference between the reservoir pressure before the start of large-scale production in Reykjanes and the year 2012. The pressure drop is as high as −40 bars in the centre of the geothermal field, decreasing gradually to −10 bars some 1.5 km away from the centre of production (Figure 14(a)). Although the analysis of the 2013 earthquakes shows a local aseismic body under the centre of the production field [117] , there is no clear relation between the distribution of the 2013-2015 earthquakes and the general shape of the pressure drawdown (Figure 14(a)). The pressure drawdown has an overall NE elongation and is mostly confined to the SRSZ, banking against the ENE area of the 1972 earthquake swarm (Figure 14(a)). Both the shape and the detailed changes in the isolines of pressure drawdown are governed by the structures identified from the seismo-tectonic and the structural analysis of the geothermal reservoir. The centre of the drawdown occupies a narrow NNE zone where the −40 bars isolines align on the NNE segments of Rauðhólar/Sýrfell. The adjacent isolines of −30 to −37.5 bars, however, are oriented ENE and not NNE (Figure 14(a)). More precisely, the −30 bar isoline matches the trace of the Litla-Vatnsfell Fault to the northwest. To the southwest, the isolines of −37.5 and −35 bars coincide with the ENE sinistral surface ruptures of earthquakes, which are the last structure to the east from which steam escaped from the reservoir during the 2014 field mapping [65] . Within the western half of the pressure drawdown area, the isolines of −35 to −25 bars bend locally from ENE to N-S, exactly on the traces of the N-S dextral faults crossing the middle of Lón (Figure 14(a), Figure 15(a), Figure 15(b)).

b) The effect of the drawdown due to reservoir fluid extraction (Figure 14(a)) is also visible as land subsidence between 2009-2013 on InSAR images (Figure 14(b)) [119] , as well as in gravity and GPS data [142] . The InSAR images (TerraSAR-X and TanDEM-X radar satellites) show a 3 km wide elliptical subsided area with an overall NE/ENE-SW/WSW elongation (Figure 14(b)) roughly matching the area of pressure drawdown (Figure 14(a)). The vertical subsidence rate coinciding with the centre of the production zone is on average 43 mm/yr, and the subsidence totals 20 cm between 2009 and 2013. The horizontal ground motion is up to 15 mm/yr towards a relatively narrow NNE-SSW zone in the centre of the reservoir. Although the overall subsided area is wider on InSAR

than the shape of the pressure drawdown, the centre of both is to the east of Lón in the SRSZ. Additionally, their overall shapes fit with the ENE sinistral oblique-slip boundary faults at Litla-Vatnsfell and north of Skálafell. Furthermore, both the maximum drawdown and maximum subsidence occur on the NNE segments of Rauðhólar/Sýrfell (Figure 14(a) and Figure 14(b)).

c) Fluid inclusions [143] [144] and chemical analysis [145] [146] [147] indicate that the geothermal fluid in Reykjanes has a dominant seawater origin although since the Pleistocene meteoric water has also recharged the reservoir. The replacement rate of the reservoir fluid is estimated to range from once every 5 years [148] to once since the last ice age [149] . This natural recharge is considered to occur along two NE-striking rift-parallel structures, which could be vertical semi-permeable boundaries formed by the precipitation of anhydrite [1] . The two boundaries form a NE structure of some 2.5 km width, stretching from the shore to the latitude of Sýrfell, whose centre coincides with the Rauðhólar/Sýrfell hyaloclastite ridge. Numerical modelling suggests that 2/3 of the natural recharge enters the geothermal system from the southwest quadrant [150] .

The structures identified in this study support a dominant recharge of the reservoir fluid from the southwest and show the role of tectonic block compartmentalisation by rift and transform zone structures in the direction of natural recharge. The fluid coming from the northeast encounters the 1972 seismic belt, which acts as a boundary (Figure 14 and Figure 15(a)). Due to this ENE barrier, less fluid enters the reservoir within the SRSZ as a part of it is dissipated through the fracture system within the NRSZ. The fluid coming from the southwest quadrant on the other hand enters the SRSZ within an open area between the 1972 and 2013-2015 seismic belts in a narrow NE corridor. This dominant southwest direction of recharge also fits the shape of pressure drawdown bounded by the ENE sinistral oblique-slip faults of north Skálafell and Litla-Vatnsfell (Figure 14(a) and Figure 15(a)), thus channelling the water from the sea, deep into the reservoir. These are however, regional structural configurations (Figure 15(b)), and different from local barrier and carrier structures within the reservoir (Figures 13(a)-(d)).