Origin and kinematics of a basin‐scale, non‐polygonal, layer‐bound normal fault system in the Levant Basin, eastern Mediterranean

Polygonal, layer‐bound normal faults can extend over very large areas (>2,000,000 km2) of sedimentary basins. Best developed in very fine‐grained rocks, these faults are thought to form during early burial in response to a range of diagenetic processes, including compaction and water expulsion. Local deviations from this idealised polygonal pattern are common; however, basin‐scale, layer‐bound faults with non‐polygonal map view are not well‐documented and accordingly, their genesis is not well understood. In this study, we use 3D seismic reflection data, biostratigraphy and well logs from the Southern Levant Basin, offshore Israel, to develop an age‐constrained seismic‐stratigraphic framework and determine the geometry and kinematics of such basin‐scale fault system. The faults tip out downwards along an Eocene Unconformity, but unlike layer‐bound faults in the Northern Levant Basin, they do not reach the base of the Messinian evaporites, instead tipping out upwards at the top Langhian. On average, the faults in the Southern Levant Basin are 6.3 km long, have an average throw of 120 m, and consistently strike NW‐SE. Throw‐depth plots, accompanied by thickness changes, indicate that the faults accumulated growth strata during the Late Burdigalian and are spatially and kinematically associated with a WSW‐ESE‐striking strike‐slip fault. Unlike true polygonal faults, these faults propagated through ca. 2 km‐thick sandstone‐prone Oligocene‐Miocene strata. Whereas previous studies from the Northern Levant Basin associate fault nucleation and growth with burial‐related diagenesis, the sandstone‐prone character of the Oligocene‐Miocene suggests that this process cannot be readily applied to the Southern Levant Basin. Instead, we highlight potential tectonic events that occurred during and may have triggered thin‐skinned extension at times of fault growth.


| INTRODUCTION
Layer-bound normal faults (defined as faults that are vertically restricted within discrete stratigraphic units and do not offset the basement) are found in sedimentary rocks throughout the geological record. One of the most common types of layer-bound faults are polygonal faults. These low-displacement (<100 m) normal faults are found in >150 basins around the world, forming broadly polygonal plan-view systems covering extensive areas (>2,000,000 km 2 ) (Cartwright, 2011). The faults are confined within discrete stratigraphic units called tiers, detached from the acoustic basement . These tiers are commonly tens to hundreds of meters thick, predominantly often dominated by very fine-grained, smectite-rich claystone or chalk, and are bounded by sandstone-prone units or other types of detachment layers (Cartwright & Dewhurst, 1998;Cartwright & Lonergan, 1996;Dewhurst et al., 1999). Their unique polygonal planform suggests growth within an isotropic stress field (i.e. 1(vertical) > 2 = 3 ), with this polygonal planform being highly sensitive to local changes in the prevailing stress regime (Roberts et al., 2015). For example, changes in host rock dip, and stress perturbations around salt diapirs, pockmarks and even deep-water channels all can alter the faults polygonal planform, causing them to become locally aligned or radially disposed (Carruthers et al., 2013;Goulty, 2002Goulty, , 2008Ireland et al., 2011;Morgan et al., 2015). Even in these cases, fault nucleation and growth are assumed to be triggered by the same, early burial-related diagenetic process as inferred for true polygonal faults. Whereas their kinematics are fairly well understood, and there is a general agreement they form by post-burial diagenetic processes involving dewatering in fine-grained clays and chalk, the exact mechanism responsible for their development is still under debate (Cartwright, 2011;Cartwright & Lonergan, 1996;Goulty, 2008;King & Cartwright, 2020;Wrona et al., 2017).
One particularly striking example of a basin-scale (ca. 70,000 km 2 ), non-polygonal layer-bound fault system, for which the diagenetic model has previously been proposed, is the 'piano-key' fault system of the Levant Basin, eastern Mediterranean (Ghalayini et al., 2017;Kosi et al., 2012). Comprised of NW-striking, mostly nonpolygonal, linear (i.e. one single dominant orientation) faults, this system covers the entire Levant Basin, displacing the >2-km-thick Oligocene-Miocene strata (Ghalayini et al., 2017). By integrating seismic attribute analysis and throw measurements along the faults surface, the spatial and temporal evolution of the fault system has been analysed in the Northern Levant Basin (Figure 1) (Ghalayini et al., 2017;Ghalayini & Eid, 2020;Hawie et al., 2013;Kosi et al., 2012). On the basis of the faults layer-bound geometry, a small portion of polygonally shaped faults in the system, and the lack of known extension events at times of presumed fault nucleation, Ghalayini et al. (2017) associated the piano-key faults with the same dewatering, diagenetic mechanism often inferred to drive polygonal fault nucleation and growth. The dominant NW strike of the piano-key faults reflected their development in an anisotropic (rather than isotropic) stress field (Ghalayini et al., 2014(Ghalayini et al., , 2017. Piano-key faults are also documented in the Southern Levant Basin, displacing a thick sandstone-prone, Oligocene-Miocene sequence (see detailed description in sub-section 2.0) (Craik & Ben-Gai, 2019;Gouliotis, 2019;Karcz et al., 2019;Needham et al., 2017;Ortega et al., 2019;Steinberg et al., 2011). The presence of these layer-bound faults within the sandstone-bearing sequence challenges the application of the diagenetic model fault development for the Southern Levant Basin. In this study, we aim to propose a mechanical model for the formation of the unusual piano-key faults in the Southern Levant Basin by constraining their structural evolution, the lithological variability of the faulted sequence and the broader geodynamic setting of the basin at the time of fault formation. To do this, we use high-quality, 3D seismic reflection data, chronostratigraphic markers and well logs from six offshore wells. This allows us to (1) create a detailed, age-constrained, stratigraphic framework; (2) constrain the lithological variability of the different units within the faulted Oligo-Miocene; (3) measure the geometrical properties of individual faults and the fault system as a whole; (4) determine the fault kinematics; (5) discuss • Non-polygonal, layer-bound faults offshore Israel consistently strike NW, perpendicular to the basin margin. • Unlike the North Levant Basin, the faults tipout upwards at the Langhian, not reaching the Messinian Evaporites. • Nucleated as syn-depositional faults between ca. 15 and 17.5 Ma, the faults are kinematically related with a WSW-ENE-striking strike-slip fault. • Propagated through ca. 2 km-thick sandstonedominated strata, their mechanism cannot be readily applied to burial-related diagenesis. • We show geodynamic events that could have triggered thin-skinned extension at times of fault growth.
F I G U R E 1 (a) A regional map of the Levant basin. Zoomed area shows the Ghalayini and Eid (2020) published fault system offshore Lebanon and the three different fault types described by them. (b) Throw-depth profiles of these three fault types described offshore Lebanon (Modified from Ghalayini & Eid, 2020). (c) The diagenetically induced mechanical model suggested by Ghalayini and Eid (2020) for offshore Lebanon (a-c) are modified from Ghalayini and Eid (2020). possible mechanical models for the formation of the piano-key faults, while also considering the geodynamic events occurring at the basin during fault nucleation and growth. We argue that the previously proposed diagenetic mechanism may not be applicable in this specific case (Ghalayini et al., 2017;Ghalayini & Eid, 2020), and those late Miocene regional tectonic events which shaped the Levant Basin may have played a role. More generally, we also argue that the presence of layer-bound normal faults should not be taken to necessarily infer very fine-grained sedimentary sequences in the absence of good data. As shown here, such structures could be erroneously interpreted as polygonal faults formed in an anisotropic stress field (Ghalayini et al., 2017), while actually forming in a sandstone-prone host rock. Our new interpretation for the origin of this enigmatic basin scale, non-polygonal (i.e. unidirectional) layer-bound normal fault systems encourage re-examination of the origin and kinematics of other unidirectional normal fault systems to see if they have similar origins. Finally, as these layer-bound faults 'fossilise' the strain distribution in sedimentary basins, they can help us reconstruct the regional stresses and geodynamics of these basins.

| GEOLOGICAL SETTING
The Levant Basin is located in the eastern Mediterranean and is bordered by the Cyprus subduction arc to the north, the Eratosthenes Seamount to the west and the continental margin of Egypt, Israel, Lebanon and Syria to the south and east (Figure 1a). We here follow the arbitrary division of the basin into the Southern and Northern Levant Basins approximately along the Israel-Lebanon maritime border (Ben-Gai, 2018) ( Figure 1a). The basin's unique location within a triple junction of the Eurasia, Arabia and Africa plates means that it evolved in response to a complex series of tectonic-stratigraphic events. The basin initially formed in response to Permian, Triassic and Early Jurassic rifting, associated with multiphase, NW-SE-oriented extension, thinning of the continental crust and the formation of NE-SW-striking normal faults (Gardosh & Druckman, 2006;Gardosh, Druckman, Buchbinder, & Rybakov, 2008;Gardosh et al., 2010;Garfunkel, 1998;Garfunkel & Derin, 1984;Granot, 2016;Robertson, 2007;Sagy et al., 2015). Following this, passive margin conditions prevailed until the Late Cretaceous, during which time a shallow marine carbonate platform was established along the basin margin (Gardosh & Druckman, 2006;Gardosh, Druckman, Buchbinder, & Rybakov, 2008;Gardosh & Tannenbaum, 2014;Garfunkel, 1998;Garfunkel & Derin, 1984). Late Cretaceous convergence between the African and Eurasian plates resulted in the formation of a north-dipping subduction zone along the Cyprus Arc (Gardosh, Druckman, Buchbinder, & Rybakov, 2008;Morag et al., 2016;Robertson, 1998a). Within the Southern Levant Basin, compressional stresses related to ongoing subduction caused large-scale folding above the pre-existing, rift-related normal faults (Cohen et al., 1990;Druckman, 1994;Freund, 1975;Garfunkel, 2004;Krenkel, 1924;Sagy et al., 2018). Forming part of the 'Syrian Arc' (Krenkel, 1924), these folds are most prominent onshore and along the basin's eastern margin, where high-amplitude, short wave-length anticlines are developed (i.e. 10-30 km long, 5-10 km wide and amplitude of >1 km) (Eyal, 1996;Gardosh, Druckman, Buchbinder, & Rybakov, 2008;Walley, 1998). This first pulse of Syrian Arc-related folding stopped by the Eocene, during a time characterised by the deepening and the deposition of deep-water chalk and marls across much of the Middle East (Bar et al., 2013;Garfunkel, 1998;Sagy et al., 2018;Steinberg et al., 2018;Ziegler, 2001).
In addition to witnessing a second Syrian-Arc-related folding event (Syrian Arc II) (Gardosh, Druckman, Buchbinder, & Calvo, 2008;Needham et al., 2017;Robertson, 1998a;Sagy et al., 2018;Walley, 1998), the Oligo-Miocene recorded a drastic change in the depositional environment within the basin, from deep-water carbonates to deep-water clastics, resulting in the Eocene Unconformity . This change created a drastic increase in sedimentation rates, peaking between ca. 24 and 12 Ma (ca. 900 m/Myr), two orders of magnitude higher than the pre-Oligocene period (ca. 5 m/Myr) (Torfstein & Steinberg, 2020). The cause for this drastic and immediate change was linked to a series of geodynamic events that exposed large expanses of previously submerged areas, which then formed significant clastic sediment sources. These events include (1) regional uplift of the eastern margin that exposed the Arabian Plateau, initiating large-scale, NW-directed drainage system into the retreating Levant (Bar et al., 2016;Faccenna et al., 2019;Gvirtzman et al., 2011;Zachos et al., 2001;Ziegler, 2001); (2) regional doming south of the Levant basin, created by the Afar Plume, which elevated the Ethiopian Plateau (31-29 Ma) (Bosworth et al., 2015); (3) Red-Sea rifting, which was initiated by the Cairo Plume (23 Ma) (Bosworth et al., 2005(Bosworth et al., , 2015; (4) the final stages of closure of the Indian Ocean-Mediterranean Seaway (IOMS) in the Aquitanian (Bialik et al., 2019;Torfstein & Steinberg, 2020); (5) the activation of the Continental Margin Fault Zone along the Levant eastern margin in the Early Oligocene (Gvirtzman & Steinberg, 2012); (6) the development of the Dead Sea transform in the Burdigalian (Freund, 1975;Garfunkel, 1997;Nuriel et al., 2017;Segev et al., 2014) and (7) the local uplift of the Judea Hills, onshore Israel (Bar et al., 2013). It is not yet clear if or how all these events kinematically interacted, but it does highlight that during the Oligo-Miocene, the Levant Basin was tectonically very active and that this activity could have influenced the formation and growth of the fault system considered here.

| The piano-key faults of the Northern Levant Basin
The piano-key fault system is composed of an NWstriking normal fault system that covers ca. 70,000 km 2 offshore Lebanon, Israel and Cyprus (Ghalayini et al., 2017). Offshore Lebanon, in the Northern Levant Basin, the faults are bounded below by the Eocene Unconformity and above by the base of the Messinian evaporites, displacing the Oligo-Miocene sedimentary sequence (Ghalayini et al., 2017;Ghalayini & Eid, 2020;Kosi et al., 2012). Based on their geometry and how fault throw varies with depth, the faults in the system offshore Lebanon were divided into three main 'types' by Ghalayini and Eid (2020) (Figure 1). Type-1 (T1) faults are predominantly located in the deep basin ( Figure 1a). They are tall (ca. 3800 m), long (6-12 km), linear, strike NW, and have a maximum displacement of 200-350 m (Ghalayini & Eid, 2020) (Figure 1b). Throw vs depth analysis, which highlights the depth of fault nucleation as a function of maximum throw, revealed two throw maxima separated by a local minimum, creating a 'B-Type' profile (Ghalayini et al., 2017;Muraoka & Kamata, 1983). The analysis indicates that T1 faults had nucleated in separate tiers, later connecting by fault tip propagation (Figure 1b). The presence of growth strata in the 'Lower-Middle Miocene Interval' suggests faults breached the surface during the Early-Middle Miocene (Reiche et al., 2014), even though, we note, the faults could have nucleated at greater structural depths. Found in the northernmost part of the basin, adjacent to the Latakia Ridge, Type 2 (T2) faults are small (ca. 1000 m tall), short (2-3 km in length) and have smaller displacements than T1 faults (<60 m). Unlike T1, T2 faults have no observable growth strata, and their throw-depth analysis creates a symmetrical, 'C-type' profile that lacks any local minima (Ghalayini & Eid, 2020;Muraoka & Kamata, 1983) (Figure 1b). Unlike the other types of faults, T2 faults are not co-linear in planform, but rather form a semi-polygonal planform (Figure 1a). Type 3 (T3) faults are linear, striking NW-SE. They are found along the eastern basin margin and do not displace the Eocene Unconformity or the base-Messinian (Ghalayini & Eid, 2020) (Figure 1b) being the smallest faults in the basin, that is they are <800 m tall, have <90 m of displacement and are <3 km in length. T3 faults are characterised by 'C-type' throw profiles and lack growth strata, similar to the T2 faults (Ghalayini & Eid, 2020).
The vast areal extent of the faults, alongside their layer-bound character and the polygonal planform of the T2 faults, led Ghalayini et al. (2017) to suggest that they formed due to compaction and dewatering during shallow burial (e.g. Cartwright, 2011). In the absence of borehole data, these authors inferred that the faults developed in a mudstone-dominated, very fine-grained sedimentary sequence, typical of polygonal fault systems. They argued that the throw-minimum on the T1 faults and their 'Btype' throw-depth profiles is associated with a sandstoneprone, basin floor fan (Ghalayini & Eid, 2020) (Figure 1c). Similar observations were made offshore Norway, both Ormen Lange Field, Møre Basin (e.g. Möller et al., 2004;Stuevold et al., 2003), and in the exploration well 35/9-3T2 located in the Måløy Basin (Jackson et al., 2014). In the case of the latter, a 92 m-thick sandstone-dominated body separated two mudstone-dominated tiers of polygonal faults, leading to a local minimum on throw-depth profiles (Jackson et al., 2014).

| DATA SET
The available data set consists of seven deep-water wells, and one 3D Pre-Stack Depth-Migrated (PSDM) seismic reflection volume covering 2355 km 2 in water depth of ca. 1.5 km offshore Israel (Figure 2a). The seismic data were acquired in 2009 and processed in 2010 by Petroleum Geo-Services. Reprocessing of the survey in 2019 by WesternGeco focused on the faulted Oligo-Miocene sequence, with a final bin size of 25 × 25 m. Inlines and crosslines are oriented NE-SW and NW-SE respectively (i.e. parallel and perpendicular to the faults orientation). The seismic data are zero phase, 'normal' SEG polarity where a positive amplitude peak indicates an increase in acoustic impedance with depth (red in figures), and a negative amplitude troughs indicate a decrease in acoustic impedance (blue in figures). F I G U R E 2 (a) The location of the study area in the southern Levant basin, overlaid by the outline of our seismic data (white), the location of available wells (colour coded) and the outline of the profile displayed in c. (b) The seismic-stratigraphic framework for the southern Levant basin used in this study. (c) A depth-migrated seismic cross-section through the available wells. Interpretation highlights the seismic-stratigraphic framework and the geometry of the piano-key layer-bound faults.
The available wells targeted the Oligo-Miocene sequence, with X-2 terminating just above the faulted sequence. X-1 is the deepest well, reaching as deep as the Eocene unconformity (i.e. near the basal tips of the studied faults; Figure 2c). We had access to gamma-ray (GR), Neutron, Density logs (all three measured every 10-15 cm, depending on the well), the computed volume of shale (Vsh) (see Methodology), cutting samples and lithostratigraphic markers.

| METHODOLOGY
We used lithostratigraphic markers, and chronostratigraphic data from dated cutting samples (Torfstein & Steinberg, 2020) to constrain the age of nine sub-evaporite reflections ( Figure 2b). The deepest reflection mapped in this study was not penetrated by the wells, but following other seismic-stratigraphic frameworks, which correlated onshore data to the shallow offshore in the Southern Levant Basin (Gardosh, Druckman, Buchbinder, & Rybakov, 2008;Steinberg et al., 2011Steinberg et al., , 2018, we interpret it as the Upper Cretaceous Unconformity (previously labelled 'Senonian Unconformity'), based on its characteristic seismic expression (Steinberg et al., 2018) (Figure 2b). We also used spectral decomposition to highlight the subtle structural elements, most importantly the WSW-ENE-striking fault and its associated splays. Spectral decomposition involves decomposing the seismic reflection data into three frequency-band limited copies. Then the seismic amplitude envelopes are mapped from each copy along the interpreted structural surface and blended as red, green and blue channels, respectively, to form a single composed full-colour map. The obtained spectral decomposition maps highlight primarily frequency-modulated seismic tuning due to subtle resolution thickness changes along the interpreted surfaces and are useful for delineating fine geological structures (Othman et al., 2016;Partyka et al., 1999). More specifically, we here used the Geoteric HDFD spectral decomposition workflow which further enhances colour resolution and vertical resolution within the RGB blend (e.g. Eckersley et al., 2018).
The Oligo-Miocene succession comprises three (from youngest to oldest) main units: (i) smectite-rich mudstone, which contains thin sandstone beds; (ii) marls, which contain thin beds of smectite-rich mudstone and (iii) a mud-rich sandstone, which forms the host rock to the fault system studied here (see Section 5.1 for further details). Given these units are all located below a thick, halite-dominated unit, a well-log-only analysis of their detailed composition is potentially problematic. This was highlighted by Christensen and Powers (2013), who raised two main issues based on their study of the nearby Tamar gas field. First, contamination of the rock units by hypersaline, barite-rich, water-based drilling fluids, which were used to drill through the Messinian evaporites and subsequently used to hinder the swelling of smectite-rich mudstone units. Second, in most cases, a relatively thick, sandstone-poor and fully resolved by well-logs mudstone located either directly above or below the depth of interest is used as a baseline for 100% clay content. However, the bounding mudstones in our study area are marls, rather than the clay-rich mudstone that comprises the faulted sequence of interest. Additionally, the faulted mud-rich layers within the overall heterolithic sub-salt sequence may also be too thin to be fully resolved by the logging tools compared to the bounding marls-mudstone sequence, causing a traditional GR-based analysis to overestimate shale volumes. Another commonly used technique to differentiate between mudstone and sandstone is the crossover between Neutron and Density logs (i.e. in the case of sandstone, neutron logs are deflected to the right and density logs are deflected to the left; see Figure 3). An issue arises here as these logs were mostly measured within the faulted, gas-prone, mud-rich sandstone unit and are therefore affected by the gas content. The gas content causes underestimation of the pore space within the rocks and therefore affects the separation of the neutron and density logs, inhibiting a log-based lithological interpretation (see an overview by Rider & Kennedy, 2014). In a similar way to the GR log, the Natural Gamma Ray Spectrometry (NGS) log also measures the natural radioactivity emitted from rocks, but unlike a GR log, it also breaks the radioactive energy into the three main radioactive elements (i.e. Potassium (K), Thorium, and Uranium). This allows for potentially better minerology identification, but as the drilling fluid was high in Potassium, it masks the real rock mineralogy.
To overcome these difficulties, Christensen and Powers (2013) suggest that Nuclear Magnetic Resonance (NMR) can be used to correlate between the 'clay-bound water' and the total porosity to create a so-called 'Volume of Shale' log (Vsh). Once the Vsh log was calculated, it was correlated to a GR baseline so that sandstone and mudstone could be differentiated. In areas where neutron and density readings were not affected by gas, the separation between these two logs (i.e. see above) also helped guide the Vsh log-driven lithological analysis.
In our study, we used a similar analysis to that outlined by Christensen and Powers (2013), that is creating a simplified lithology column for each well based on the Vsh log and neutron density, which were used to calibrate the GR log. As we are not able to show the Vsh log due to confidentially reasons, we have also calibrated our analysis with cutting samples. Along our depth range of interest, these samples were collected every 3 m, then washed on the drilling rig, with the lithologies averaged by the well site geologist (Figure 3). Our analysis combined data from all available wells, which include the two wells utilized by Torfstein and Steinberg (2020). At depths where >1 well penetrated the depth of interest, the logs and resulted lithologies were compared, but besides slight bed-thickness F I G U R E 3 Lithological interpretation along the X-1 well. Integrating (from left to right) the seismic signature, GR log, sample cuttings, derived simplified lithology and neutron-density log. The simplified lithology column represents the lithological variability of the faulted Oligocene-Miocene section. The depth axis in this and subsequence figures was removed according to the confidentiality agreements. variations, the interpreted lithologies are consistent between the wells.
Kinematic analysis was performed on 136 faults in the study area. The spatial and temporal evolution of the different structural elements, including the NW-SEstriking faults, were determined by following the methodology of Jackson et al. (2017): (1) depth structure maps were used to highlight the current geometry of the sedimentary sequence. These maps were then used to generate thickness (isopach) maps that highlight the timing of syn-depositional structural activity: acrossfault thickening indicates syn-sedimentary fault growth, and thinning across the Leviathan High indicates periods of syn-depositional folding Thorsen, 1963); (2) strike parallel throw profiles (T-X) were used to visualise the spatial distribution of strain within the fault system (Childs et al., 1995(Childs et al., , 2019Peacock & Sanderson, 1991, 1996Walsh & Watterson, 1990). By measuring the throw along a fault length (we measured throws every 250 m, regardless of the fault length), T-X profiles can help indicate kinematic interaction between and the linkage of faults within the system (Childs et al., 2019;Dawers & Anders, 1995;Nicol et al., 2010;Peacock & Sanderson, 1991, 1996. This analysis is specifically beneficial when the piano-key faults are compared to polygonal faults, as polygonal faults are thought to have a higher degree of fault interaction and linkage (i.e. the system is more mature) with depth (Cartwright, 2011); (3) dip-parallel throw profiles (T-Z) were used to understand the role dip-linkage and mechanical stratigraphy had on fault growth and ultimate geometry (Baudon & Cartwright, 2008b;Childs et al., 1996;Jackson et al., 2017;Muraoka & Kamata, 1983;Peacock & Sanderson, 1991;Roche et al., 2012;Rotevatn et al., 2019;Rykkelid & Fossen, 2002). T-Z plots also help us infer the depth and correlative geological period at which the faults nucleated (Barnett et al., 1987;Nicol et al., 1996;Walsh et al., 2003;Walsh & Watterson, 1988;Wrona et al., 2017). We extracted T-Z plots from the position of maximum throw, as identified on the T-X plots. Similar techniques have been applied in previous studies to highlight how sandstone intervals separate polygonal fault tiers and how they are themselves characterised by local minima (Cartwright, 2011;Ghalayini et al., 2017;Jackson et al., 2014;Lonergan et al., 1998;Stuevold et al., 2003;Turrini et al., 2017;Wrona et al., 2017); (4) expansion index (EI) plots (i.e. hanging wall vertical thickness of a stratigraphic package divided by its footwall vertical thickness) were constructed to identify growth strata and hence determine if faults breached the surface during their development . Growth strata are highlighted, where EI > 1 Robson et al., 2017;Thorsen, 1963;Tvedt et al., 2013). EI plots were constructed at the same sites where throwdepth plots were taken.

| RESULTS
Here we integrate our observations of seismic facies variability with drilling data to constrain the age and lithology of our new, sub-evaporite, seismic-stratigraphic framework. Thickness changes within different units are also highlighted, which helps infer the timing and pattern of deformation. We then integrate this with our detailed analysis of the geometry and kinematics of the NW-SEstriking, layer-bound faults (Sections 5.3 and 5.4), such that we can ultimately propose a mechanical model for fault development (Section 6).

| Seismic-stratigraphic framework and integration with drilling data
In addition to the base-evaporite horizon, we interpreted 10 pre-evaporite horizons to constrain the 10 seismicstratigraphic units (Figure 1). Each section below begins with a description of the unit's seismic facies and lithology, the latter derived from drilling data. Then, the current geometry of its bounding upper surface, and if present, any thickness changes within the unit are also characterised. This structural framework provides the foundation for the kinematic analysis linking the timing of layer-bound faulting and other regional tectonic events in the Southern Levant Basin.

| Unit 1: Pre-Coniacian
Unit 1 is characterised by sub-horizontal, continuous, moderate-amplitude reflections and is capped by the bright, continuous, 'Upper Cretaceous Unconformity' horizon ( Figure 2c). On the basis of published onshore and shallow offshore wells (i.e. no wells penetrated this unit in the deep-offshore), the Mid-Jurassic to Turonian (Mid Upper Cretaceous) unit comprises deep-water clastics and pelagical and hemipelagical carbonates (Gardosh, Druckman, Buchbinder, & Rybakov, 2008;Gardosh et al., 2011). The top of unit 1 outlines the large, triangular Leviathan High, which is located at the centre of the study area. The high is bounded to the north by an ENE-WSW-striking fault and to the south by an NE-striking, SE-dipping monocline (Figure 4a). No thickness analysis is presented for this unit as we did not have any lower boundary reflection to constrain this unit. 5.1.2 | Unit 2: Coniacian-Eocene (33.9) Unit 2 is characterised by chaotic, mostly transparent or low-amplitude seismic reflections and is capped by the F I G U R E 4 Structural maps of the horizons used in the study, indicating the present-day geometry of the Oligocene-Miocene. Note the different depth ranges of the colour scales used for enhancing the structural elements in each map. bright, continuous, 'Eocene Unconformity' (33.9 Ma) ( Figure 2). Our lithological analysis of the X-1 well indicates that Unit 2 is composed of deep-water chalk and marls, which is in agreement with previous studies of the deep Levant Basin (Figure 3) (Gardosh, Druckman, Buchbinder, & Calvo, 2008;Gardosh & Tannenbaum, 2014;Steinberg et al., 2011). The Leviathan High is still well-expressed at the top of Unit 2, with NW-striking faults also developed at this level (Figure 4b). Unit 2 (Figure 5a) thins across the Leviathan High and it gently thickens from the footwall to the hanging wall of the faults. The unit age corresponds to the same age as the Syrian Arc I, therefore the thickness changes seen here suggest an uplift/folding of the Leviathan High alongside fault activity during this time. 5.1.3 | Unit 3: Rupelian-Early Chattian (33.9-24.07 Ma) Characterised by sub-horizontal, semi-transparent, moderate amplitude seismic reflections, Unit 3 is capped by the bright, semi-continuous Intra-Chattian horizon (24.07 Ma) (Figure 2). The lithological analysis from the X-1 well shows that this deep Oligocene unit is composed of thin (ca. 5-m-thick) sandstone beds within a mostly mudstonedominated sequence (Figure 3). The Leviathan High is also well expressed at the top of Unit 3, with the NW-striking faults also well developed (Figure 4c). Degradation in the imaging quality at this depth interval makes the Intra-Chattian horizon difficult to map (Figure 4c), as expressed in the southern portion of the thickness map for Unit 3 (Figure 5b). However, it is still clear that Unit 3 is broadly tabular and of uniform thickness, indicating the main tectonic event(s) occurring during Unit 2 had largely stopped (Figure 5b).

| Unit 4: Late Chattian
(24.07-23.02 Ma) Unit 4 is characterised by a sub-horizontal, mostly continuous, moderate to high amplitude seismic facies, and is capped by the semi-continuous, moderate to low amplitude Top Oligocene horizon (23.02 Ma) (Figure 2). Like Unit 3, Unit 4 is composed by alternations of sandstone and mudstone (Figure 3). The top of Unit 4 still shows the Leviathan High and the NW-SE-striking faults ( Figure 4d). As with Unit 3, Unit 4 is isopachous (Figure 5c) Figure 2). Penetrated by five of the six wells, Unit 5 is also composed of alternating sandstone and mudstone, like Units 3 and 4 ( Figure 3). The Leviathan High, the NW-SE-striking faults and the WNW-ENE-striking strike-slip faults are all very clearly expressed at the top Unit 5 map (Unit 4E). Like Units 3 and 4, Unit 5 is broadly isopachous (Figure 5d) (Figure 2). Unit 6 is more sandstoneprone than the deeper units with a Net-to-Gross of 70% (Karcz et al., 2019), and it contains the stratigraphically youngest sandstones present within the faulted units ( Figure 3). Similar relatively high Net-to-Gross sandstone units are described in neighbouring fields (Christensen & Powers, 2013;Stearman et al., 2021; see Figure 1). The high Net-to-Gross of this unit has the substantial impact on the fault-growth model we present later in this manuscript (see Section 6.1.2). The top of Unit 6 continues to show the Leviathan High and the NW-SE-striking faults (Figure 4f). Unit 6 gently thins towards the WSW-ENEstriking strike-slip fault, but no thickness changes are seen across the faults (Figure 5e). This NW thinning trend towards the strike-slip fault may suggest a renewed tectonic activity in the study area. 5.1.7 | Unit 7: Late Burdigalian-Middle  Unlike the units below, Unit 7 is characterised by semitransparent, low to medium amplitude seismic reflections, capped by the continuous, bright Intra Langhian horizon (14.4 Ma) (Figure 2). Unit 7 is mudstone-dominated and contains thin (<5-m-thick) carbonate beds; sandstone is notably absent (Figure 3). Because the carbonate beds are relatively thin, they are not clearly detected in well logs; however, they are observed in all six well-site analyses, documented in cutting samples and composite logs. In addition to the triangular Leviathan High and the NW-SE-striking faults, the Intra-Langhian structural map also shows a system of polygonally arranged depressions (Figure 6), which locally become concentric around the Tamar anticline (Figure 6c). Besides thickness changes associated with the NW-striking faults (Figure 7), Unit 7 also F I G U R E 5 Thickness maps of the seismic-stratigraphic units used in this study. Note the different thickness ranges of colour scales used for covering the entire thickness range of each unit. The maps indicate that the study area had experienced two main kinematic events. The first was during the Eocene, where thinning is seen across the Leviathan High. This was followed by a hiatus in tectonic events during Units 3-5 seen by isopachous maps. The second kinematic event peaked in Unit 7, where thinning across the high, alongside across-fault thickening showed faulting was associated with folding. Faulting had stopped in Unit 9, but folding seems to continue until the deposition of the Messinian evaporites. Contours are shown for every 100 m.
shows thinnings across the Leviathan High (Figure 5f), indicating a significant tectonic activity period.
In detail, flattening the Top Langhian horizon reveals a significant intra-formational onlap horizon within Unit 7 (Figure 8). This horizon, which is dated as Late Burdigalian (ca. 15 Ma), divides Unit 7 into two (Figure 8). The lower sub-unit 7 (7a) is broadly tabular and seismic reflections are continuous over the Leviathan High (Figure 8c,e), whereas the upper sub-unit (7b) on plans this Late Burdigalian horizon on both sides of the Leviathan High (Figure 8c Figure 6). Like sub-Unit 7b, Unit 8 thins across the Leviathan High, with no faulting-related thickness changes, suggesting that faulting was no longer syndepositional (Figure 5g).
We note that the top of Unit 8 defines an unconformity, with Serravallian strata missing in the two wells studied by Torfstein and Steinberg (2020). Those authors show that Unit 8 is capped by a mudstone-rich, carbonate-poor Tortonian unit (Unit 9), suggesting the top of the Langhian coincides with the global Miocene Carbonate Crash event and concluding that the unconformity resulted from a large-scale carbonate dissolution event. This dissolution event may be responsible for the polygonal pattern identified in Unit 8. Stratigraphically, these depressions are confined to Unit 8 and are present

F I G U R E 7
Cross-section through two locations within the study area indicating growth strata across the faults during the Late Burdigalian. Sections are located within (a), and away from the structural high (b), to indicate thickness changes occurred within Unit 7 regardless of the relative location to the structural high. Syn-depositional onlaps and wedge-shaped thickening are highlighted (b). Throwdepth plot for the L7 fault is superimposed. above the growth strata associated with the NW-SEstriking, layer-bound faults (Unit 7), meaning their formation and deformation post-date at least the main, initial period of faulting. It is therefore possible that displacement of the polygonal fabric occurred due to subsequent upward propagation of the faults when the faults were not surface-breaking. 5.1.9 | Unit 9: Early Tortonian (13.82-9.18 Ma) Unit 9 is characterised by chaotic, low-amplitude seismic-facies which are capped by a moderate amplitude, continuous Intra-Tortonian horizon (9.18 Ma) (Figure 2). Torfstein and Steinberg (2020) note that Unit 9 is mudstone-rich and foraminifera-and CaCO 3 -poor, indicative of carbonate dissolution (see above). Because of its chaotic seismic signature, we cannot say for certain whether the faults extend through Unit 9, although the top of the unit does not appear to be deformed by these structures (Figure 4i). Unit 9 clearly thins across Leviathan High (Figure 5h) indicating the second tectonic activity which started at Unit 7 is continuous here. The origin of this chaotic section is beyond the scope of this manuscript, but we do suggest a possible correlation to similar observations made by Papadimitriou et al. (2018), where they suggested a similar chaotic section on the flanks of the Eratosthenes Seamount, triggered by the collision between the Seamount and Cyprus. 5.1.10 | Unit 10: Late Tortonian (9.18-5.96 Ma) Unlike Unit 9, Unit 10 is characterised by sub-horizontal, continuous, moderate amplitude seismic reflections, capped by the base evaporites bright and continuous seismic horizon (Figure 2). Unit 10 is lithologically similar to Unit 9, comprising foraminifera-and CaCO 3 -poor mudstone (Torfstein & Steinberg, 2020). The top of Unit 10 dips gently north-westwards, although three large channels are present (Figure 4j). The NW-striking faults are absent. Similar to Unit 9, Unit 10 thins across the Leviathan High (Figure 5e).
In summary, our data set is dominated by the large, triangular-shaped Leviathan High and numerous NW-SE-striking, layer-bound (i.e. by the Top Langhian and Base Oligocene horizons) normal faults. Thickness changes are seen in two main stratigraphic intervals and corresponding time periods: the first during the Coniacian-Eocene, where thinning across the Leviathan High is most dominant, and the second during the Burdigalian and Langhian, where marked thickness changes occur not only across the Leviathan High but also across the NW-SE-striking faults. These two phases of deformation appear to have been separated by a period of relative quiescence.

| Other prominent structural elements
In addition to the NW-striking piano-key faults and the Leviathan High described above, a prominent ENE-WSW-striking fault exists across our study area along the northern edge of the Leviathan High. Cross-sections across the fault indicate that it corresponds with a deep, single stem which cross-cuts the entire Coniacian to Oligo-Miocene sedimentary sequence (Figure 9a). From its single stem, splays spread in a negative flower structure along the Top Aquitanian horizon (Figure 8a). Spectral decomposition along the Top Aquitanian horizons highlights this WSW-ENE-striking fault, which is composed of several, similarly striking, segments (Figure 9c). Adjacent to these segments, the otherwise NW-SE-striking piano-key faults change their strike to N-S, perpendicular and locally physically linked to the ENE-WSW-striking fault system (Figure 9c). A similar geometric relationship is seen in the adjacent Karish gas field (ca. 50 km east of our study area). There, NWstriking faults abut against the 'Karish Shear Zone' (Gouliotis, 2019), a WSW-ENE-striking, dextral strikeslip fault that could be the along-strike extension of the geometrically similarly fault found in our study area (Stearman et al., 2021). Additionally, we note two other smaller (ca. 5 km long), ENE-WSW-striking faults at the centre of the study area, where the intensity of NW-striking faults is locally higher than elsewhere (Figures 8e and 9c). Finally, in terms of their age, the thickening of Units 7 and 8 indicates the ENE-WSWstriking structure was active in the Late Burdigalian to Late Langhian (Figures 7 and 9b).
Similar geometrical relationships between otherwise NW-SE-striking piano-key faults and WSW-ENEstriking faults are documented in the Northern Levant Basin (Ghalayini et al., 2014). There, the faults change their orientation to strike in an almost N-S direction and they are inferred to represent Riedel-like structures orientated at 60° from the dextral strike-slip fault (Ghalayini et al., 2014). The origin of these faults is not yet clear, but Ghalayini et al. (2014) suggested that they may be related to a strike-slip reactivation of buried riftrelated faults by the sinistral movement along the Dead-Sea transform.

| NW-SE-striking fault geometry and distribution
We have identified, mapped and undertaken a geometric and kinematic analysis of 136, predominantly NW-SEstriking normal faults present within the Oligo-Miocene succession, bounded above by the Base Oligocene and below the Top Langhian ( Figure 10). The faults have an average length of 6.3 km and an average throw of 116 m (see Section 5.4.2 for more details) and are normally displaced relative to their length (Figure 9b). Most faults (61%) dip to the SW, with seemingly no relationship between faults dip direction and their location, except to the north of the WSW-ENE-striking fault described above, where all the layer-bound faults dip SW (Figure 9a).

| Throw-length (T-X) analysis
Of the 136 mapped faults in the study area, 16 were not included in this analysis (or the T-Z analysis described below) because they extended outside the seismic data set and thus, we could not constrain their true length. Based on their throw vs. length profile shape, the faults were classified into four groups (TX1-4) (Figure 11). TX1 and TX2 are asymmetrical, with maximum throw offset to the SE or the NW, respectively, of the fault centre. TX3 are symmetrical, with maximum throw at the fault centre, whereas TX4 is defined by a profile containing two throw maxima (Figure 11). We do not see any direct spatial correlation between these groups and other structural elements; however, we do note a change in the distribution of strain with depth. For example, our analysis shows that symmetrical profiles are more common with depth, that is whereas 37% of the faults displacing the upper boundary (Late Langhian horizon) have a symmetrical throw distribution, 67% of the faults displacing the lower boundary (Base Oligocene) have a symmetrical throw distribution ( Figure 11). Given that symmetrical profiles typify less mature faults that have developed in kinematic isolation from surrounding structures, we infer a greater degree of fault interactions and higher fault maturity at a shallower depth (Nicol et al., 2010;Walsh & Watterson, 1990).

| Throw-depth (T-Z) analysis
Throw-depth profiles were constructed for the same 120 faults analysed in Section 5.4.1. Our analysis shows that the average T-Z profile is asymmetric, with maximum throw across the Intra-Burdigalian (17.54 Ma) horizon, decreasing upwards and downwards towards the fault tips ( Figure 12a).
The faults were divided into two main groups based on their vertical extent (TZ1 and TZ2). TZ1 faults displace the entire Oligo-Miocene sequence, with an average length of 7.2 km, an average height of 1.9 km and an average vertical throw of 128 m (Figure 12). TZ1 throw profiles are asymmetrical, with a prominent maximum throw along the Intra-Burdigalian horizon (Figure 12a). From this maximum, the throw profile decreases almost linearly both upwards to the base of the Lower Tortonian chaotic unit (Unit 9) and downwards to the Upper Chattian/Eocene units. TZ2 faults are smaller (average length of 4.2 km and a maximum throw of 80 m), their lower tip does not displace the Intra-Chattian horizon and they exhibit a more symmetrical throw profile ( Figure 12). Spatially, 70% of the mapped faults in the F I G U R E 1 0 Geometrical properties of the faults. (a) The geographic location of the 136 mapped faults in the study area. Colours represent dip direction to the SW (red) and to the NE (blue) (b) Max throw vs fault length relative to the global database (Lathrop et al., 2022). The faults are located within the global database and are not anomalous in that regard.
F I G U R E 1 1 Strike-parallel throw profiles of 120 faults along the top-most horizon (Top Langhian; top) and the base horizon (Base Oligocene; bottom) with the profiles arranged into groups based on the profiles symmetry (see text for details). The resulting maps are colour coded, matching with the profiles, based on the throw profiles type (centre left and right, respectively). The relative abundance of different types is shown in a pie diagram next to the respective maps. We note that unlike polygonal faults, the faults in our study area show more symmetrical profiles with depth, indicating less strain connectivity between the faults in the system. study area are TZ1, with TZ2 mostly located along the high's flanks (Figure 12c).
Compared to throw-depths plots by Ghalayini et al. (2017) and Ghalayini and Eid (2020) from the Northern Levant Basin offshore Lebanon, TZ1 faults are similar to their Type 1 faults and TZ2 are similar to their Type 3 faults (Figures 1 and 12). Whereas some similarities could be seen with regard to their throw-depth plots, the faults in the Northern and Southern Levant Basin do have their differences. Unlike Type 1 faults offshore Lebanon, TZ1 faults offshore Israel do not offset the base-Messinian evaporite, making them smaller than the Type 1 faults offshore Lebanon (height of 1.9 km vs. 3.8 km), and with smaller vertical throw (120 m vs. ca. 250 m). TZ2 and Type 3 faults do have very similar geometrical properties, but unlike Type 3 faults offshore Lebanon which are located along the basin margin (Figure 1), TZ2 faults are located in the deep basin ( Figure 12).

| Expansion index
Expansion index (EI) for the 120 faults analysed yielded EI > 1 for Unit 7 (17.54-14.4 Ma) and Unit 8 (14.4-13.82 Ma), with EI = 1 for Unit 3 (33.9-24.07 Ma) (Figure 13a). Values <1 are seen in the other units, possibly highlighting the difficulty associated with the interpretation of the bounding horizons (Figure 13a) (see further details in Section 5.1). EI results strengthen our observations from Section 5.4.2, whereby all the faults, regardless of bottom tip depth, accumulated growth strata during the Late Burdigalian, with possible continued activation during the Langhian (Figure 13b).
In summary, our thickness maps, seismic crosssections, throw-depth profiles and expansion index data suggest that piano-key faults in the Southern Levant Basin breached the seabed during the Late Burdgialian. The correspondence between the T max position (inferred to represent the position of fault nucleation) and the base of the syn-kinematic sequence also implies the faults nucleated at or very near the seabed (see also Baudon & Cartwright, 2008a for a comparable example offshore Israel). Assessing the exact mechanics of near-seabed fault nucleation, which may be considered problematic given that near-seabed sediments would be weakly unlithified and thus unable to sustain a shear fracture, is beyond the scope of this manuscript. However, it is possible that these faults nucleated at deeper depths, before rapidly attaining their final shape and size and breaching the seabed. As such, the faults may have rapidly transitioned from being blind to syn-depositional (see Baudon & Cartwright, 2008a). 6 | DISCUSSION

| Mechanical model for the formation of the piano-key faults
We have shown that the non-polygonal, layer-bound faults identified in our study area breached the seabed during the Late Burdigalian. Regardless of their depth of nucleation, the timing of faulting raises questions regarding their origin: (1) how can a diagenetic-induced fault system, so strongly linked to very fine-grained sediments and sensitive to changes in host rock composition, propagate through a ca. 2-km-thick sandstone prone host rock?
(2) what occurred in the basin during the time of fault growth that caused their initial nucleation? (3) why are the faults so linear, striking NW-SE, perpendicular to the basin margin? To address these questions, we here describe possible mechanical models for their formation and discuss their implications.

| Diagenetic model
Previous studies from the Northern Levant Basin, offshore Lebanon suggested that the piano-key faults are nucleated and grew within mudstone-dominated host rock in accordance with the same diagenetic mechanism as the one typically associated with polygonal faults (Figure 1c) (Ghalayini et al., 2017;Ghalayini & Eid, 2020). Based on their relative geographic proximity, and the geometrical similarities between the piano-key faults in the Northern and Southern Levant Basins, we here test the role of such proposed diagenetic model in the latter.
Our lithological analysis shows the layer-bound faults in our study area had propagated through a ca. 2 kmthick sandstone-prone host rock. Sandstone-prone intervals of similar age are described in other neighbouring fields, suggesting that a relatively extensive, fan-like deep-water system was deposited in this part of the basin during the late Miocene (Christensen & Powers, 2013;Gouliotis, 2019;Karcz et al., 2019;Stearman et al., 2021). At greater depths, only X-1 penetrated the faulted strata, and within this interval, the seismic facies do not change away from the borehole. Regional studies also show that the Southern Levant basin experienced rapid sediment accumulation rates since the Oligocene. These coincide with incision events onshore Israel, and major progradation of the Nile Delta, suggesting that significant amounts of siliciclastic, sandstone-prone material reached the basin during the Oligocene and Early Miocene (Buchbinder et al., 1993;Gardosh, Druckman, Buchbinder, & Calvo, 2008;Gvirtzman et al., 2014;Steinberg et al., 2011;Torfstein & Steinberg, 2020).
These local and regional observations all suggest that the Oligocene-Miocene sequence in the Southern Levant basin was sandstone-prone, challenging the application of the diagenetic model for the Southern Levant Basin layer-bound faults. Our interpretation is that nucleation and growth of these faults were unrelated to near-surface diagenesis of fine-grained sediments, as supported by our kinematic analysis. T max typically occurs along the Intra-Burdigalian for all fault types, suggesting that they nucleated along the Intra-Burdigalian horizon (17.54 Ma) (assuming the depth of maximum throw corresponds to the depth of fault nucleation; e.g. Kim & Sanderson, 2005;Nicol et al., 1996). As the Intra-Burdigalian horizon also defines the base of the fault-related growth strata (i.e. Unit 7), we infer the faults nucleated as syndepositional faults, displacing the seabed, during the Late Burdigalian. Thus, our throw-depth plots (e.g. Figure 12) and thickness maps ( Figure 5) suggest the faults nucleated near their final upper tips, with significant down-dip propagation of their lower tips responsible for their vertical height.
We note that the T max position described here differs to that characteristic of polygonal faults, where T max (and the inferred site of fault nucleation) is located either at the centre or near the base of the fault surface (Cartwright, 2011;Wrona et al., 2017). This difference may reflect the contrasting origins and styles of growth of true polygonal faults and the layer-bound fault system described here. Specifically, the very fine-grained sediments must first be buried to activate the diagenetic processes and thus nucleate polygonal faults. In contrast, based on the assumption that T max represents the site of fault nucleation, the layer-bound faults studied here apparently nucleated as syn-depositional faults close to the seabed and propagated downwards (Cartwright, 2011;Seebeck et al., 2015) ( Figure 11). Morgan et al. (2015) challenge this assumption, proposing that polygonal faults can nucleate at the lower parts of a tier, as buried faults, but due to mechanical constraints imposed by an underlying mechanical barrier, which inhibits downwards propagation of the fault basal tip, T max can migrate upwards the tier centre. Even if this was the case in our layer-bound fault system, this would mean the faults still nucleated in sandstone-prone units, challenging the link present between fault formation and the diagenesis of very fine-grained sediments (i.e. the diagenetic model). Moreover, nucleation of the faults in our study area (TZ2) do not reach the base of the tier but instead, stop at different depth within the tier in the absence of any apparent mechanical barrier. Therefore, based on their nucleation and propagation within a sandstone-prone unit and their atypical distribution of throw compared to 'true' polygonal faults, we suggest that the layer-bound faults in our study area did not form in response to diagenesis of very fine-grained sediment, bringing us to an alternative, tectonic-related model (see Section 6.1.2). 6.1.2 | Tectonically induced layer-bound faulting Given our arguments against a diagenetic model for fault development, we here present an alternative model that is summarised in Figure 14. Our model uses our ageconstrained seismic-stratigraphic framework and refers to the tectonostratigraphic events that shaped the basin during times of fault nucleation and subsequent growth.
First, we note that Unit 2 acts as a basal décollement layer for the layer-bound faults across not only the Southern Levant Basin, as demonstrated here, but across much of the eastern Mediterranean (Gao et al., 2020;Ghalayini et al., 2014Ghalayini et al., , 2017Hawie et al., 2013). To the best of our knowledge, X-1 is the only well in the basin to penetrate Unit 2. This well encountered the Late Eocene strata but was aborted due to overpressure at that level. The exact reason for this overpressure is not known, but it is possible that the overpressure was developed by the rapidly buried Unit 2, leading to trapped fluids in the chalk and marls, eventually creating favourable conditions for overpressure to build. From a geodynamic perspective, compressional stresses associated with Syrian Arc folding, which were highest during the Coniacian-Eocene, are thought to have declined during the Oligocene (Sagy et al., 2018). This decrease in tectonic deformation is recorded in our study by a broadly isopachous, Early Oligocene unit (i.e. Unit 3) (Figures 5 and 14a). The rapid deposition continued throughout the Oligocene and Early Miocene (i.e. Units 4 and 5). This could therefore have caused pore pressure to build in the now-buried chalk and marls, driving disequilibrium compaction to the point of overpressure development in the impermeable unit, eventually leading to the formation of an intra-stratal décollement (Cosgrove, 2001;Jolly & Lonergan, 2002;Morley et al., 2008) (Figure 14b,c).
Following this period of tectonic quiescence since Eocene, evidence for deformation appears again in the Burdigalian. Thinning of the sandstone-prone, Lower Burdigalian (Unit 6) towards the WSW-ENE-striking strike-slip fault, suggest that this large fault was active at this time (Figure 5e). Initial activation of this strike-slip fault was followed by intense layer-bound faulting during the Late Burdigalian, in the mudstone-dominated Unit 7a (Figures 8e and 14d). By the end of the Langhian, both the strike-slip movement and the layer-bound normal faulting had stopped, while the uplift of the Leviathan High became the most prominent deformation event (Figure 8d). The nucleation and subsequent growth of the faults prior to the culmination of any large-scale uplift allow us to disregard the folding of the Leviathan High as a mechanism for the development of normal faults.
We do note, however, an apparent kinematic relationship between the WSW-ENE-striking strike-slip fault and the layer-bound normal faulting, as both faulting systems were most active during the Late Burdigalian and had mostly ceased by the end of the Langhian. The origin of this strike-slip faulting is beyond the scope of this manuscript. However, we do highlight several significant geodynamic events that occurred in and around the basin during times of fault activation: (1) a landward jump of strain from the Continental Fault Zone along the Levant eastern margin, to the sinistral movement along the Dead Sea transform (Gvirtzman & Steinberg, 2012;Nuriel et al., 2017); (2) the development of the Dead Sea transform (Freund et al., 1968;Nuriel et al., 2017;Segev et al., 2014); (3) the final closure of the Indian Ocean-Mediterranean Seaway (Bialik et al., 2019;Torfstein & Steinberg, 2020); (4) change in the subduction rates and slab angle beneath the Cyprus Arc subduction zone (Aksu et al., 2021;Gao et al., 2020) and (5) uplift of the Eratosthenes Seamount by >1 km at the Early Miocene (Gao et al., 2020;Papadimitriou et al., 2018;Robertson, 1998b).
A geodynamic outcome of these tectonic events may have been a counter-clockwise rotation of the basin, created by the non-subsiding Eratosthenes Seamount (Aksu et al., 2021;Papadimitriou et al., 2018;Robertson, 2007) (Figure 14e). As the Eratosthenes Seamount was stuck in place, the Levant Basin and its onshore segments, which continued to move northwards, rotated counter-clockwise around Eratosthenes. This counter-clockwise rotation could have therefore caused the formation of the offshore dextral strike-slip faults found in our data set and offshore Lebanon (Figure 14e). A similar interpretation is made by Ghalayini et al. (2014), with these authors suggesting that continued sinistral movement along the Levant Fracture System onshore Lebanon caused (dextral) strike-slip reactivation of Cenozoic, rift-related normal fauls. They then propose that the relative movement along the strike-slip faults eventually created onshore counter-clockwise block rotation, absorbing any extension in the Levant Fracture System pull-up structures onshore Lebanon.
We here build on the model of Ghalayini et al. (2014) and propose that thick-skinned, dextral strike-slip movement along WSW-ENE-striking strike-slip faults occurred in response to the large-scale, geodynamic reorganisation of the Levant Basin. This strike-slip movement induced local extensional stresses and strain, with one expression of this being the NW-SE-striking, layer-bound normal faults ( Figure 14). After nucleating, the NW-SE-striking faults propagated through the Oligocene-Miocene units until their lower tips decoupled within the overpressured Coniacian-Eocene unit. By introducing this tectonic-driven model, we can explain (1) the direct kinematic relationship we presented between the WSW-ENE-striking strike-slip fault and the NW-SE-striking faults; (2) the change in orientation from NW-SE-striking to a more E-W-striking, Riedellike orientation close to the strike-slip faults, as observed here and offshore Lebanon; (3) how the NW-SE-striking, layer-bound faults propagated through a ca. 2-km-thick sandstone-prone unit and (4) why the faults are so linear, and strike almost perpendicular to the current basin margin.
We do note that, unlike the Eocene, which is the lower boundary for the entire fault system across the Levant basin, the faults upper boundary varies; from the base-Evaporite (5.96 Ma) unit in the Northern Levant Basin, to the Top Langhian (13.84 Ma) at the Southern Levant Basin. We do not have a clear explanation for this discrepancy. One possibility is that the strike-slip faults, and their kinematically related normal faults, remained active for longer in the Northern Levant Basin (Ghalayini et al., 2014). Another possibility for this discrepancy is the presence of the chaotic section in Unit 9 in our data set, which is not present in the Northern Levant Basin. It is therefore possible that upper fault propagation was inhibited by the more ductile nature of the chaotic section.

| CONCLUSIONS
We use high-quality 3D seismic reflection, biostratigraphy and well-log data to characterise the spatial and temporal evolution of a layer-bound fault system in the Southern Levant Basin, offshore Israel. We present a new, ageconstrained, pre-Messinian seismic-stratigraphic framework for the basin, discussing the lithological variability and prominent thickness changes occurring within key intervals. This seismic-stratigraphic framework allows us to describe the prominent structural elements in our study area, which include the NW-striking, layer-bound faults, the triangular Leviathan High and a prominent, WSW-striking, dextral strike-slip fault. Throw-depth profiles, expansion index and thickness changes all indicate the layer-bound faults nucleated as syn-depositional faults during the Late Burdigalian (ca. 15-17.54 Ma) in a mudstone-dominated unit. The faults then propagated downwards through sandstone-prone Oligocene-Miocene units, tipping out within overpressured Coniacian-Eocene strata. The NW-striking faults also appear to be kinematically linked to the WSW-striking strike slip.
Based on (1) their direct kinematic relations to the strike-slip fault; (2) their propagation through sandstoneprone strata; (3) throw-depth profiles which show maximum throw at the top of the faults, differing from other documented polygonal faults; and despite nucleating in a mudstone-dominated unit, we suggest the faults did not develop through a diagenetic process as previously suggested, but as a thin-skinned response to a thickskinned tectonic reorganisation of the basin. The precise mechanics and kinematics of these geodynamic events are not clear, but they may relate to a possible counterclockwise rotation of the basin, with the spatially limited extension being accommodated by the layer-bound faults. This model suggests that basin-scale layer-bound normal faults can develop not only through a diagenetic model as proposed for polygonal faults, but also by tectonic-related processes. Therefore, we suggest that linear, layer-bound normal fault systems should be investigated in the context of the basin in which they formed.