3.a. Alternative tectonic models

Published tectonic models fall into two classes, the first of which assumes a constant regional strain during Miocene–Recent time and a second which assumes changes in strain with time.

Strike-slip models. Several authors have inferred that the Hatay Graben formed in a sinistral strike-slip-controlled system involving extension at a releasing bend and the formation of, variously, a pull-apart basin (Perinçek & Çemen, Reference Perinçek and Çemen1990), a half-graben (Lyberis et al. Reference Lyberis, Yürür, Chorowitcz, Kasapoğlu and Gündoğdu1992) or a graben (Muehlberger, Reference Muehlberger1981; Perinçek & Çemen, Reference Perinçek and Çemen1990). Rojay, Heimann & Toprak (Reference Rojay, Heimann and Toprak2001) proposed that the Hatay Graben (their Antakya–Samandağ depression) represents a releasing bend on a larger-scale strike-slip zone represented by the Karasu Rift (Figs 1, 2a). Most of their data came from the Karasu Rift, but they also reported some data for faults in the Hatay Graben. Other authors envisaged the Hatay Graben as a strike-slip, transtensional or extensional basin related to an eastward extension of the Cyprus Arc (Kempler & Garfunkel, Reference Kempler and Garfunkel1994; Ben-Avraham et al. Reference Ben-Avraham, Tibor, Liminov, Leybov, Ivanov, Tokarev and Woodside1995; Yürür & Chorowicz, Reference Yürür and Chorowicz1998). This lineament might link through the Hatay Graben to the Karasu Rift to the northeast (fig. 2b, c), and relate regionally to the westward tectonic escape of Anatolia (Yürür & Chorowicz, Reference Yürür and Chorowicz1998).

Figure 2. The four different models proposed for the development of the Hatay Graben: (a) extension at a releasing bend in a strike-slip-controlled fault system (Muehlberger, Reference Muehlberger1981; Perinçek & Çemen, Reference Perinçek and Çemen1990; Lyberis et al. Reference Lyberis, Yürür, Chorowitcz, Kasapoğlu and Gündoğdu1992); (b) a graben resulting from the westward escape of Anatolia (Yürür & Chorowicz, Reference Yürür and Chorowicz1998); (c) a strike-slip or transtensional basin related to the Cyprus Arc to the west (Kempler & Garfunkel, Reference Kempler and Garfunkel1994; Ben-Avraham et al. Reference Ben-Avraham, Tibor, Liminov, Leybov, Ivanov, Tokarev and Woodside1995; Yürür & Chorowicz, Reference Yürür and Chorowicz1998); (d) a transtensional graben formed as a result of Pliocene-Quaternary strike-slip superimposed on older basement faults of the Arabian Platform (Boulton, Robertson & Ünlügenç, Reference Boulton, Robertson, Ünlügenç, Robertson and Mountrakis2006).

Multi-phase model. Över, Ünlügenç & Bellier (Reference Över, Ünlügenç and Bellier2002) interpreted the Plio-Quaternary to present-day stress regime of the Hatay Graben based on field measurements of faults and earthquake focal plane solutions. They concluded that there was a change from an older, strike-slip to a younger, normal fault regime that reactivated pre-existing faults. However, they gave little explanation of how this could fit into the regional tectonic setting.

Transtensional model. Boulton, Robertson & Ünlügenç (Reference Boulton, Robertson, Ünlügenç, Robertson and Mountrakis2006) proposed that the Hatay Graben formed as a result of Pliocene–Quaternary transtension generally related to westward tectonic escape of Anatolia (fig. 2d). This deformation followed on from the development of the Miocene Hatay basin as the distal part of a foreland basin related to collision of the African (Arabian) and Eurasian (Anatolian) plates.

Each of the above tectonic models can be tested utilizing the structural dataset we have assembled during this work, especially when combined with other evidence including seismicity and GPS data (summarized below).

5.a. First-order structures

First-order normal faults are characterized by NW-dipping faults that are present along the SE margin of the Hatay Graben (Figs 4, 5a). Fault planes are generally eroded but still have a strong topographic expression. In the NE of the field area, adjacent to the city of Antakya, basin-scale faults displace Upper Cretaceous or Eocene limestone, and Upper Cretaceous ophiolitic rocks against Plio-Quaternary graben sediment fill (fig. 5b). Two large normal faults define the edge of the graben in this area where the width of the graben margin, defined as the area between the innermost and outermost boundary faults, is estimated as 4 km. The innermost fault has the greater throw (>500 m), although the outer fault also has significant throw (>200 m). Further south, the width of the graben margin increases to about 10 km. First-order faults form two en échelon arrays of sub-parallel fault segments that step away from the axis of the graben (fig. 4b, c). The outer array comprises three main segments, whereas the inner array is shorter, comprising two fault segments. This second array is truncated, possibly by a transverse fault that is orientated at a high angle to the graben in this area (Fig. 4). Near Antakya, two normal fault segments overlap to form a relay ramp that exhibits relay-breaching faults, which are at an oblique angle and link the two boundary faults (fig. 4c). This relay ramp forms a minor half-graben (<10 km²) on the flank of the main graben and is cut by dominantly NE–SW-trending faults, although NW–SE faults also cross-cut these faults.

Figure 4. (a) Inset graphic in top left shows the zones used to divide fault data. (b) Geological map of the Hatay Graben; lines A–A' and B–B' indicate the position of the cross-sections (c) and (d) below, respectively; TF – inferred transfer fault; modified from Boulton, Robertson & Ünlügenç (Reference Boulton, Robertson, Ünlügenç, Robertson and Mountrakis2006). Positions of the photographs in Figures 5 and 7 are indicated by stars and a letter. Boxes indicate spatial extent of Figures 6 and 8.

Figure 5. (a) View to the north along the innermost fault plane on the southwestern margin, near the present coast; note a prominent fault scarp in the foreground, width of view about 1 km. (b) View along major fault plane (~10 m high) east of Antakya cutting Upper Cretaceous and Eocene limestone, downthrowing to the left/west towards the graben floor; orientation of main fault plane 160°/67 °W. (c) View of limestone horsts (furthest is 500 m away) bounded by normal faults near Sebenoba; the softer marl is covered with crops. (d) Normal faults in Pliocene sandstone and marl in the axis of the graben, outcrop about 8 m wide. (e) View to the NW of the first-order oblique-slip fault observed near the coast cutting Middle Miocene limestone, width of view ~2 km. Figure 6 is a topographic map of this fault. (f) Photograph of the vertical shear zone (1 m wide) observed near Kesecik.

As noted above, first-order normal faults are uncommon on the northwestern margin of the graben (fig. 4b). None were observed in the northwest but a few do occur on the coast near Çevlik, striking NE–SW. The outer faults expose Middle Miocene limestone in the footwall and Upper Miocene marl in the hangingwall. The innermost faults cut Upper Miocene sediments at the surface. The magnitude of the throw on these faults is unknown as there are no exposed stratigraphic constraints.

The majority of first-order faults trend parallel or sub-parallel to the graben. In addition, a large NW–SE-striking transverse fault is inferred, running along the course of the Karaçay (fig. 4b). The evidence for this is: (1) numerous faults strike NW–SE in the vicinity; (2) there is a stratigraphic offset on the northern margin of the graben across the Karaçay valley; (3) there is a difference in the height of the graben margin on either side of the Karaçay and (4) to the south, one of the graben-parallel boundary faults appears to terminate at the Karaçay.

5.b. Second-order normal faults

Second-order normal faults are exposed along the margins of the graben and also cut the graben sediment fill. Movements along normal faults striking at a high angle to the trend of the graben have resulted in back-rotation of fault blocks. For example, a second-order fault strikes E–W through the Harbiye gorge at a high angle to the graben margins near Harbiye (Fig. 4). This fault, a probable dip-slip or oblique-slip fault, displaces Upper Cretaceous and Eocene limestone by tens of metres. In addition, many third-order E–W-striking faults are present in the vicinity.

A zone of intense deformation is present between two basin-bounding faults near the village of Sebenoba in the southeast (Figs 4, 5c). Two main populations of faults are observed there: NE–SW-striking (basin-parallel) and NW–SE-striking (basin-perpendicular). Together, these faults form a series of horst–graben structures, in which blocks of well-lithified Middle Miocene limestones have been uplifted, whereas Upper Miocene marls have been downthrown. Both sets of faults are generally extensional, although a few show reverse slip. Fault planes are planar and are associated with narrow zones (<25 cm) of fault breccia composed of hard limestone clasts. The two fault populations (NW–SE and NE–SW-striking faults) repeatedly cross-cut each other, but despite careful examination, do not exhibit any systematic relative chronology. Furthermore, it has not been possible to establish any systematic chronology of faulting throughout the Hatay Graben as a whole (Boulton, Robertson & Ünlügenç, Reference Boulton, Robertson, Ünlügenç, Robertson and Mountrakis2006).

Another zone of map-scale faulting that is situated between two basin-bounding faults was observed in the north, east of Antakya (fig. 4a). In this area, faulting has resulted in back-rotation of a large fault-block (~4 × 1 km) between two first-order faults and the formation of an intervening sub-basin. Smaller (second- and third-order) synthetic and antithetic faults are mostly orientated parallel to the boundary faults. These are planar dip-slip faults with a throws of <5 m metres, placing Eocene limestones against Upper Miocene marls and thereby forming a series of small horst and graben structures. Cross-cutting third-order faults trend NW–SE, with measured offsets of <5 m.

5.c. Third-order normal faults

Third-order normal faults are common at outcrop-scale and in sediments of all ages (latest Cretaceous–Pliocene), on both the floor and the flanks of the graben (fig. 5d). For example, a first-order graben-bounding fault cuts latest Cretaceous to Eocene sediments in a quarry near the village of Dursunlu (fig. 4a), downthrowing to the west. The exposed fault plane has been quarried (fig. 5b), revealing a series of third-order conjugate dip-slip and oblique-slip faults that strike at a high angle to the boundary fault; unfortunately, the relationship between the boundary fault and the smaller faults has been concealed by the quarrying.

5.d. First-order oblique-slip faults

First-order oblique-slip faults are uncommon in the Hatay Graben, although one was identified on the NW margin, near the village of Yoğunoluk (Figs 4, 5e). The sense of movement cannot be determined directly from structural measurements and there is no direct evidence of the age of this fault. However, landscape features provide an indication of the sense of movement and age of the fault. First, there is evidence of sinistral offset of streams by at least 100–150 m. Strike-slip faulting has separated the upper and lower stream courses creating a left-lateral ‘dog-leg’ across the fault (Fig. 6). Second, there is a vertical component of motion along this fault as shown by noticeable differences in height (~150 m) between the footwall and hangingwall; also, streams are more incised on the footwall compared to the hangingwall. There is also evidence of ‘beheaded streams’, where fault motion has been greater than the capacity of the streams to down-cut so that the catchments and downslope segments of the stream have become separated across the fault zone. The presence of stream offset and ‘beheaded streams’ suggests that this first-order oblique-slip has been active at least during later Quaternary times.

Figure 6. SPOT image of the fault plane shown in Figure 5e. Note the geomorphological evidence for lateral motion on the fault plane, for example, offset and beheaded streams. Excerpt of SPOT image (shown top left) 5 118–278 06/11/23 08:21:49 2 J, copyright CNES (2007), Distribution Spot Image, S.A., France, all rights reserved.

In addition, some first-order normal faults along the SE margin could be oblique-slip in nature, but the kinematics of these faults is unknown because slickenlines are not preserved.

5.e. Third-order oblique-slip faults

Third-order oblique slip faults are only recognized in the field where there is a three-dimensional exposure or the presence of slickenlines. One such fault was observed near the village of Güneysöğüt (fig. 4b), where a vertical fault plane (117/90°) has a throw of 2 m, offsetting Pliocene sandstone and Quaternary conglomerate along a 40 cm wide fault zone (fig. 7a).

Figure 7. (a) Third-order oblique-slip fault in Plio-Quaternary sediments in the graben; the fault plane is vertical (117°/90°) with 2 m vertical offset; the horizontal motion is unknown. (b) Series of third-order strike-slip faults (174°/90°W, 168°/77°W, 180°/56°W) with low-angle slickenlines (6–7°) in Pliocene sandstones adjacent to the basin-bounding fault in the southwest. Note the sinistral motion on faults. Compass is 20 cm long.

5.f. Third-order strike-slip faults

Third-order strike-slip faults are common at an outcrop scale especially in sediments of Pliocene age. Good examples were observed near Kesecik (fig. 4b), where a small-scale flower structure is exposed. In this outcrop the main fault strikes at 140° with a dip of 90° and splays upwards into three smaller faults. Local changes in bedding orientation indicate that the sediments between these faults have been rotated. Another vertical shear zone was observed at the same location (Fig. 5f). Although the geometry is less clear in this instance, the sediments on either side of the shear zone are deformed and do not correlate across the shear zone. In many other locations, strike-slip faults are inferred based on the presence of small offsets (<10 cm) or lack of vertical offset across a vertical fault plane. For example, small strike-slip faults are observed south of Samandağ (fig. 4b), where three sub-parallel sinistral strike-slip faults (fig. 7b) are present within a deformed zone adjacent to a first-order basin-bounding fault. Where a sense of motion can be determined in the field, it was found that sinistral strike-slip faults (N = 31) are more common than dextral faults (N = 10). The sinistral strike-slip faults show no preferred orientation (that is, all points of the compass are represented), whereas the dextral faults are generally orientated NE–SW or NW–SE.

In addition, significant variation was observed in the orientations of bedding planes within Pliocene sediments near Samandağ; locations <200 m apart exhibit differences in dip direction of >90° (Fig. 8). This variation is considerably more than can be accounted for by primary depositional processes and thus a tectonic cause is inferred, possibly relating to block rotation along a strike-slip zone (see below).

Figure 8. Topographic map (derived from SRTM data; source for this dataset was the Global Land Cover Facility, www.landcover.org) showing the location of Pliocene sediments and the strike of the bedding planes at various locations (see Fig. 4 for location within the Hatay Graben); inset rose diagrams show the orientation of third-order faults measured at those locations.

5.g. Third-order reverse faulting

Reverse faults are uncommon within the Hatay Graben. Where present, they are always small, third-order structures with a maximum displacement of <10 m, and commonly associated with normal, oblique-slip, or strike-slip faults. For example, reverse faults associated with normal faults were observed in the Sebenoba area (fig. 4b). In addition, reverse faults are spatially associated with strike-slip faults near Kesecik (fig. 4b).

5.h. Evidence of folding

There is little evidence of compressional folding in the Hatay Graben. Folding was not generally observed in post-Eocene units. A gentle fold was observed on the main road between Antakya and Samandağ; however, this is closely associated with normal faults and is interpreted as a drag fold.

By contrast, Eocene strata are extensively folded, as seen to the east of Antakya, where medium-bedded calcarenites are folded into disharmonic, asymmetrical, gentle folds. The dip of the bedding in this area is nearly vertical in places. Further north, around the towns of Belen and Kırıkhan, the Eocene strata are more deformed. Folding there is highly variable, but fold vergence was observed as always being towards the NW or W. Fold geometry ranges from open to tight and angular folds; fold axial planes are widely dispersed and the fold wavelengths are <5 m. The largest fold observed in the study area is a N-vergent monocline deforming the boundary between the Hatay Ophiolite and the Upper Cretaceous limestones south of Harbiye.

6.a. Fault trends and patterns

In total, more than 850 measurements of fault planes were made in the Hatay Graben. Most of these are of third-order faults, which are generally better exposed than the first- and second-order faults. The data include measurements of strikes and dips, fault displacements and the orientations of slickenlines on fault planes (Appendix; available online as supplementary material at http://www.cambridge.org/journals/geo). In addition, thirteen fault planes showed two sets of lineations, but no relative chronology could be determined in these cases.

When all the data are considered together, the majority of the faults strike between 60° and 140°. There are three main trends in strike direction: NE–SW, NW–SE, N–S. The majority of the faults dip at >50°. There is no clear relationship between fault orientation and angle of dip (fig. 9a). When the strike of a fault is compared to the rake of the lineations on its fault plane (fig. 9b), there is no obvious trend. The rakes of the lineations are highly variable and show no preferred relationship with the dip angles of the fault planes (fig. 9c). This lack of a clear relationship between dip and rake implies these faults do not satisfy coulomb theory that predicts an inverse relationship between the magnitude of rake and dip of the fault plane, and thus these faults instead record 3D-strain (Reches & Dieterich, Reference Reches and Dieterich1983; ten Veen & Kleinspehn, Reference ten Veen and Kleinspehn2002).

Figure 9. (a) Graph showing the strike of faults against the dip, for all data (n = 880). (b) Graph showing the strike of faults against the rake of any lineation measured upon the fault (n = 244). (c) Graph showing the dip of the fault plane against the rake of lineations present (n = 244).

To refine the interpretation, the structural data were divided into sub-groups and analysed in several ways. This approach has been shown to be effective in other regions (e.g. Rhodes, Greece: ten Veen & Kleinspehn, Reference ten Veen and Kleinspehn2002) in order to counterbalance the limitations of fault data due to heterogeneity of fault populations.

First, the data were divided into structural ‘zones’ (Fig. 10) to assess any spatial variation in fault population. Six zones were accordingly defined for the Hatay Graben, three for the graben floor and three for the flanks of the graben (labelled 1–6; fig. 4a). Zone 1 covers the NW graben margin and zones 2 and 3 the graben floor. The recognition of zones 2 and 3 takes into account the proposed transverse fault that strikes perpendicular to the graben margins along the course of the Karaçay valley (fig. 4b). This structure also defines the NE edge of zone 4 that is represented by an uplifted block between the graben floor and its SE margin. Zones 5 and 6 are delineated on the SE graben margin, again recognized by the location of the possible transverse fault.

Figure 10. Chart showing breakdown of fault strike data into different types of fault for each spatial zone defined; see Figure 3. Note that the total number of faults is greater than the numbers of specific faults due to the motion on the fault plane not always being identifiable.

Second, the faults were considered according to the inferred age of the faulting (Fig. 11). For example, a fault measured in a Middle Miocene unit is necessarily Middle Miocene or younger and is placed in a ‘Middle Miocene or younger’ age category; a fault cutting the Middle-Late Miocene boundary is Late Miocene or younger and is placed in the ‘Late Miocene or younger’ category, etc.

Figure 11. Chart showing breakdown of fault strike data into different types of fault, depending on the age of the sedimentary rock in which the fault was measured. Note that the total number of faults is greater than the numbers of specific faults due to the motion on the fault plane not always being identifiable.

6.b. Spatial distribution of faults

Zone 1 is dominated by normal faults striking 030–070°, 160–180° and 090–140°. However, a few faults are of oblique-dip or strike-slip type (both sinistral and dextral) with strikes of 050–070° and 150–160°. By contrast, zones 2 and 3, covering the axial zone of the graben, exhibit numerous normal and strike-slip faults, plus a few oblique and rare reverse faults. In both of these areas, the strikes of normal faults exhibit two main trends (040–060° and 150–190°) with a subordinate trend striking 110–130° (Fig. 10). Similarly, strike-slip faults also exhibit a range of strikes (140–200° in zone 2; 100–120° in zone 3). Zone 4 covers an uplifted block encompassing the mountainous Samandağ and is bounded to the south and southeast by the River Asi. The faults of this zone are again dominantly normal and strike-slip, but oblique-slip faults are also present. The general trends of faults are similar to elsewhere; however, the dominant orientation of fault strike is different, as NW–SE faults are more common. Normal faults mostly strike 090–130°, although there are N–S- and NE–SW-striking faults as well. By contrast, oblique-slip and strike-slip faults exhibit two dominant trends: 050–070°, 100–120° and 020–080°, 140–170°, respectively. Zones 5 and 6 cover the SE margin of the graben. The majority of faults of zone 5 are normal faults, generally striking 140–180°. Oblique-slip, strike-slip (both mainly striking 150–180°) and reverse faults (striking 050–080°) are also present. Zone 6 again shows a spread in the orientation of faults; normal faults predominantly strike 0–020° and 040–060°, with a sub-set striking 120–140°. Oblique faults, the next most common type, show a dispersed fault pattern. Small numbers of reverse faults and strike-slip faults (sinistral and dextral) are also present (Fig. 10).

In summary, normal faults predominate, commonly with trends of 030–070°, 090–140° and 150–190°. Oblique-slip and reverse faults are uncommon, as are strike-slip faults in zones 1, 5 and 6. However, strike-slip faults, generally trending 020–080° or 100–170°, are numerous in zones 2, 3 and 4, representing the floor of the graben.

6.c. Temporal distribution of faulting

Further analysis depends on assessing the maximum age of faulting. To determine the age of any particular fault or set of faults it is necessary to define an unconformably overlying unit that is not affected by this faulting. This is the case for the Upper Cretaceous/Eocene sediments in the area, which were deposited and deformed prior to the formation of the Hatay Graben. These sediments were folded and faulted and then covered by much less deformed Miocene sediments. In addition, intraformational syn-sedimentary faulting has been documented from Middle Miocene limestone related to the inferred early foreland basin phase, again prior to formation of the neotectonic graben (Boulton, Robertson & Ünlügenç, Reference Boulton, Robertson, Ünlügenç, Robertson and Mountrakis2006).

No such clear stratigraphic relationships are seen within the Miocene–Recent sediments of the Hatay Graben. However, it is useful to determine the abundances, trends and fault types in sediments of progressively younger age. Any fault types not seen in a younger unit in the same area could indicate that this faulting had ended before deposition of this particular sedimentary unit. Using this approach, the fault populations were separated into those in Upper Cretaceous, Eocene, Lower Miocene, Middle Miocene and Upper Miocene and Pliocene sediments, respectively (Fig. 11). Very few faults were identified in Quaternary deposits and no slickenlines were identified, presumably because of the coarse, unconsolidated nature of the sediments. Occasional measurements of Quaternary faults are combined with Pliocene faults.

Faults cutting Pliocene–Quaternary sediments exhibit three clear trends. The dominant fault orientations are NE–SW (040–065°) and N–S (0–010°) with a subordinate 120–140° trend (Fig. 11). Normal and strike-slip faults predominate with similar trends. There are also a few oblique and reverse faults. Faults considered to be Late Miocene or younger (that is, observed in Upper Miocene sediments) show similar trends in fault strike to the Pliocene dataset. However, Middle Miocene sediments contain far fewer identified strike-slip faults than those of Pliocene age. This is also true for the Middle–Early Miocene sediments, where there are three main trends of faulting (NW–SE, N–S, NE–SW), dominated by normal faults (Fig. 11).

Faults measured within Eocene and latest Cretaceous lithologies differ noticeably. All of the faults measured in the Eocene rocks have similar orientations to those observed within younger sedimentary rocks. However, when fault types are considered, it is seen that normal faults dominate in the Eocene rocks but have only one main trend in strike (340–040°), in contrast to two or three different trends in younger units. The Upper Cretaceous dataset is different again. When all of the fault data from Upper Cretaceous rocks are considered, a dominant trend emerges: 330–000° (Fig. 11). This pattern results from a combination of oblique-slip faults trending N–S and normal faults with two dominant strike directions (030–070° and 120–150°).

In summary, analysis of the fault data indicates that faulting in the pre-Miocene rocks (Upper Cretaceous & Eocene) is dissimilar to the younger rocks. This is to be expected, as the older rocks experienced a prior period of deformation as mentioned above. In addition, the margins of the graben, mainly composed of Middle Miocene and older rocks, are dominated by normal faults in three main trends (NE–SW, NW–SE and N–S). By contrast, the graben floor that is composed of Upper Miocene and younger sediments also exhibits numerous normal faults, but strike-slip faults are more common than observed elsewhere.

8a. The Hatay Graben represents a releasing bend on a regional sinistral strike-slip fault (Fig. 2)

In one scenario (Yürür & Chorowicz, Reference Yürür and Chorowicz1998), the Karasu Rift should be a pull-apart basin on the East Anatolian Fault, with the Hatay Graben as the surface expression of the Anatolian/African plate boundary. In this case, the Hatay Graben would be a NE–SW sinistral strike-slip fault with NNE–SSW normal faults and sinistral strike-slip faults, N–S and NW–SE dextral strike-slip faults and E–W reverse faults.

In reality, we observe NE–SW-trending normal faults forming a NE–SW-orientated half-graben. Sinistral strike-slip faults show no preferred orientation (that is, all points of the compass are represented); dextral faults are generally orientated NE–SW or NW–SE and rare reverse faults are N–S-trending, which in effect excludes this hypothesis.

In other, similar, scenarios, motion along a N–S-trending strike-slip fault should result in the development of a NW–SE-orientated pull-apart graben (Muehlberger, Reference Muehlberger1981; Perinçek & Çemen, Reference Perinçek and Çemen1990; Lyberis et al. Reference Lyberis, Yürür, Chorowitcz, Kasapoğlu and Gündoğdu1992; Rojay, Heimann & Toprak, Reference Rojay, Heimann and Toprak2001). In this case, normal/oblique-normal faults should trend parallel to the graben/half-graben margins (NW–SE). Three sets of strike-slip faults should develop with sinistral faults orientated NW–SE, and dextral faults orientated E–W and NE–SW; reverse faults should also develop which should be NE–SW-trending. However, the Hatay Graben is NE–SW-trending and faults are not consistent with these predictions, again excluding this model as a viable hypothesis.

Rojay, Heimann & Toprak (Reference Rojay, Heimann and Toprak2001) recognized four main sets of faults with NE–SW-strikes dipping to the NW or SE in the Hatay Graben, and with kinematic indicators suggesting an oblique-normal-slip motion with a left-lateral component. These data are consistent with our observations for the Hatay Graben using a much larger dataset. However, two lines of evidence oppose the interpretation of Rojay, Heimann & Toprak (Reference Rojay, Heimann and Toprak2001), and other authors (Muehlberger, Reference Muehlberger1981; Perinçek & Çemen, Reference Perinçek and Çemen1990; Lyberis et al. Reference Lyberis, Yürür, Chorowitcz, Kasapoğlu and Gündoğdu1992; Yürür & Chorowicz, Reference Yürür and Chorowicz1998), that the Hatay Graben formed as a pull-apart basin on a releasing bend of a strike-slip fault system represented by the Karasu Rift/Dead Sea Fault Zone. First, the orientation and geometry of the present graben are not consistent with this model. A pull-apart basin along the sinistral Dead Sea Fault Zone in this area (fig. 2a) should be orientated NW–SE, whereas the Hatay Graben is NE–SW-trending (Fig. 1), and therefore should be an area of transpression if controlled by motion on the Dead Sea Fault Zone. Second, the Miocene initiation of normal faulting in the area took place before the Dead Sea Fault Zone intersected this region during the Pliocene or later (Ben-Avraham, Reference Ben-Avraham1978), and thus the formation of the Hatay Graben cannot be directly related to the initiation of the Dead Sea Fault Zone.

In addition, if a dextral lineament associated with the Cyprus arc (the Larnaca Ridge) extends through the Hatay Graben (fig. 2c) as suggested by Kempler & Garfunkel (Reference Kempler and Garfunkel1994), Ben-Avraham et al. (Reference Ben-Avraham, Tibor, Liminov, Leybov, Ivanov, Tokarev and Woodside1995) and Yürür & Chorowicz (Reference Yürür and Chorowicz1998), it should result in E–W-trending normal faults, NE–SW-trending sinistral strike-slip faults, NW–SE and N–S dextral strike-slip faults, and N–S reverse faults. Again, these predicted fault orientations are not compatible with our observations of fault patterns and stress analysis.

8b. The Hatay Graben changed from strike-slip to extension with time

In this interpretation the field evidence would be expected to document early strike-slip faults cut by a later generation of extensional faults. Strike-slip faults should be present in large numbers on the flanks of the Hatay Graben but should be uncommon within the graben, which should be dominated by graben-parallel normal faults. This is not consistent with the patterns of faulting observed, but is consistent with the seismicity data that point to a present-day extensional setting (see Section 4).

Över, Ünlügenç & Bellier (Reference Över, Ünlügenç and Bellier2002) reconstructed the Plio-Quaternary to Recent stress regime of the Hatay Graben using fault measurements (<300) and earthquake focal plane solutions. They inferred a strike-slip stress regime (σ₃ to NE) and two extensional stress regimes (σ₃ to the NE and E). They also reported two lineations, high angle and low angle, on individual fault planes, and considered normal striations to be younger, based upon superposition. These authors concluded that the present-day stress state is dominantly extensional with a small strike-slip component. They further inferred a change from an older, strike-slip to a younger, normal fault regime.

A model of strike-slip followed by extension is not supported by the data obtained during this work. First, the majority of our strike-slip faults were measured in rocks of Pliocene or younger age, whereas many of Över, Ünlügenç & Bellier's (Reference Över, Ünlügenç and Bellier2002) fault measurements were in Eocene or unspecified (Miocene?) units. Only five of their locations are in the Pliocene–Quaternary graben fill, two of which were considered together. Our data instead indicate that normal faulting was followed by a greater incidence of strike-slip faults. We were also unable to confirm any systematic cross-cutting of faults or slickenlines of fault planes.

Över, Ünlügenç & Bellier's (Reference Över, Ünlügenç and Bellier2002) conclusion of recent normal faulting is, however, consistent with focal plane mechanisms determined from recent earthquakes (see Section 4). These earthquakes are shallow and only two have had strain ellipses calculated for them and so may not be representative of the area as a whole. For example, the Marmara Sea segment of the North Anatolian Fault Zone exhibits transtension partitioned between seismic and aseismic slip in different areas (Aksu et al. Reference Aksu, Calon, Hiscott and Yaşar2000), which may also be applicable to the Hatay Graben. Given that we record numerous strike-slip faults within Pliocene sediments, the absence of calculated strike-slip focal mechanisms is surprising. Possibly, the earthquake record is simply too sparse or ill-constrained to detect evidence of strike-slip. Alternatively, the stress regime may have changed to more extensional after the Pliocene, which is quite possible as we have very little fault data for the unconsolidated Pleistocene sediments.

Över et al. (Reference Över, Kavak, Bellier and Özden2004) identified a large strike-slip lineament in the Amanos Mountains based on the location and apparent trend of recent earthquakes. However, the earthquake epicentres may well be mislocated, as noted in Section 4, by as much as 10–15 km; if so, the strike-slip lineament could instead be located near or within the Hatay Graben. Över et al. (Reference Över, Kavak, Bellier and Özden2004) further identified a possible strike-slip lineament within the Amanos Mountains using SPOT images. However, this lineament appears to run along the crest of the mountain range about 15 km north of the Hatay Graben, and is likely to mark a watershed rather than a tectonic lineament.

8c. The Hatay Graben is an oblique extensional (transtensional) structure

In this model it is predicted (assuming α is ~ 30°) that three populations of dip-slip, oblique-slip and strike-slip faults should be present. One population should form sub-parallel to the graben (approximately NE–SW) coevally with two sets of faults orientated at a high angle to the rift and displacement direction (fig. 14a). Our structural data are mainly consistent with this transtension, which is therefore explored in the next section.

Figure 14. (a) Diagram showing the types and orientations of structures expected to develop in a sinistral transtensional zone with the orientation of the Hatay Graben. Note the similarities with the fault orientations actually observed, that is, three orientations of normal faults, conjugate strike-slip faults and E–W-trending thrusts and fold-axes. (b) Sketch showing the development of a two-stage graben where a strike-slip zone cuts through an existing graben. (c) Sketch showing partitioned transtension with a central wrench-dominated domain with adjacent extension-dominated domains (modified from De Paola et al. Reference De Paola, Holdsworth, McCaffery and Barchi2005). (d) Sketch showing the geodynamic/neotectonic setting of the Hatay Graben (HG) within the Eastern Mediterranean. Note the directions of GPS vectors acting to create net extension in the Hatay region. NAFZ – North Anatolian Fault Zone; other abbreviations as in Figure 1.

10a. Extension followed by strike-slip?

Normal faults dominate the Miocene sediments on the flanks of the graben, whereas strike-slip faults are numerous in the axial Pliocene sediments. Possibly extension dominated during the Miocene (on graben-bounding faults), followed by strike-slip faulting in a second phase during the Plio-Quaternary (fig. 14b). However, this seems unlikely for several reasons. If the graben had switched from extensional to strike-slip with time, we would expect to find only normal faults in Miocene and older rocks orientated parallel to the graben (NE–SW), plus possibly a conjugate fault set at 60°. The fault pattern associated with Pliocene–Quaternary strike-slip faulting would depend on whether the fault zone was dextral or sinistral; both have been suggested (Muehlberger, Reference Muehlberger1981; Perinçek & Çemen, Reference Perinçek and Çemen1990; Lyberis et al. Reference Lyberis, Yürür, Chorowitcz, Kasapoğlu and Gündoğdu1992; Kempler & Garfunkel, Reference Kempler and Garfunkel1994; Ben-Avraham et al. Reference Ben-Avraham, Tibor, Liminov, Leybov, Ivanov, Tokarev and Woodside1995; Yürür & Chorowicz, Reference Yürür and Chorowicz1998). Assuming a sinistral strike-slip regime, we would predict NE–SW and NNE–SSW sinistral faults, E–W and NW–SE dextral faults and N–S normal faults in the axial zone of the graben. This is similar to what is observed, but the transtensional model in which faulting became more oblique over time fits the fault data better for the following reason.

Where the acute angle between the rift trend and direction of extension (α) is ≥45°, all faults are expected to be of dip-slip striking parallel and obliquely to the graben trend. This conforms to the fault patterns observed in pre-Pliocene sediments. However, where α ≤ 30°, dip-slip, oblique slip and strike-slip faults all develop as three discrete fault populations with one population parallel and two populations at a high-angle to the rift trend. This corresponds closely to our observations made in the Pliocene sediments. In addition, the stress analysis for the Miocene sediments shows σ₃ (approximate direction of extension) as ~45° to the rift orientation, whereas σ₃ from the Plio-Quaternary sediments is variable. This could reflect faulting sub-parallel to the rift, as commonly observed, coupled with faulting at a high angle to the graben with (unconstrained) block rotations.

10b. Partioned transtension?

Another option that we favour is that the two apparently different ages and locations of faulting associated with the graben relate to strain-partitioning or ‘partitioned transtension’ (De Paola et al. Reference De Paola, Holdsworth, McCaffery and Barchi2005). This allowed two styles of faulting to co-exist in different areas: extensional faulting on the rift margins (extension-dominated transtension) and strike-slip faulting (wrench-dominated transtension) near the graben axis (fig. 14c). This interpretation explains many observations, including the overall fault orientations, the presence of dip-slip faults, possible block-rotations near the graben axis, the presence of a significant topographic graben, and the deformation of Pliocene sediments by graben-bounding faults. Similar interpretations have recently been proposed for part of the North Anatolian Fault Zone (Aksu et al. Reference Aksu, Calon, Hiscott and Yaşar2000), the North Aegean Trough (McNeill et al. Reference McNeill, Mille, Minshull, Bull, Kenyon and Ivanov2004) and the Northumberland Basin, northern England (De Paola et al. Reference De Paola, Holdsworth, McCaffery and Barchi2005).

Partitioned transtension can occur where pre-existing crustal anisotropies, such as faults or shear zones, are present, controlling the location and orientation of the different structural domains. As noted above, localized E–W faults parallel large faults in the underlying kilometres-thick Hatay Ophiolite. In addition, the northern margin of the Arabian plate is known to be strongly affected by extensional faults that relate to Triassic rifting of the Tethyan Ocean to the north (Brew et al. Reference Brew, Lupa, Barazangi, Sawaf, Al-Imam and Zaza2001). Pre-Cretaceous rocks are not exposed in the study area, but it is likely that a zone of pre-existing structural weakness exists beneath the Hatay Graben that was exploited by strike-slip during Plio-Quaternary times.

Finally, it is possible the two alternatives may be mutually exclusive, that is, a trend towards more oblique transtension over time could have been accompanied by strain partitioning and basement reactivation. This is also consistent with previous interpretations of the Hatay Graben as a part of a diffuse plate boundary between the African and Anatolian plates (Kempler & Garfunkel, Reference Kempler and Garfunkel1994; Robertson, Reference Robertson, Robertson, Emeis, Richter and Camerlenghi1998; Vidal, Alvarez-Marron & Klaeschen, Reference Vidal, Avarez-Marron and Klaeschen2000). We infer that westward extrusion of Anatolia from a post-collisional area of Arabia to the east to a pre-/syn-collisional area in the Eastern Mediterranean to the west is expressed as oblique–extensional strain at the interface between these two areas, including the Hatay Graben (fig. 14d).