Seismic stratigraphy of the southern Eratosthenes High, eastern Mediterranean Sea: growth, demise and deformation of three superposed carbonate platforms (Mesozoic–Cenozoic)

Isolated carbonate platforms form some of the world's largest, most prolific petroleum reservoirs. Their morphological characteristics and the fact that they are commonly buried and sealed by terrigenous mudstones or evaporites makes them prime exploration targets in tropical marine basins of all ages. However, such platforms vary enormously in size and are not always straightforward to interpret seismic stratigraphically as they may be confused with other prominent features, such as volcanoes, basalt accumulations at continental margins and faulted basement highs (Burgess et al. 2013; Jiménez Berrocoso et al. 2021). In many cases, it is often only after internal seismic stratigraphic details have been resolved that a carbonate origin can confidently be determined where no reliable ‘ground truth’ exists.

Recent major hydrocarbon discoveries in the deep-water Levant Basin, namely the Tamar, Leviathan and Aphrodite gas fields (Fig. 1), have contributed to a surge in exploration activity in the eastern Mediterranean Sea (Lie et al. 2011; Peace 2011; Needham et al. 2017) and to the acquisition of large volumes of subsurface data in a region that had previously seen only limited industry interest. The Zohr biogenic gas discovery in 2015 (Fig. 1), an apparent isolated Cretaceous carbonate build-up in a province hitherto regarded as dominated by deep-water Cenozoic siliciclastic reservoirs, stimulated speculation that similar Mesozoic or Cenozoic carbonate reservoirs might exist across the region (Bertello et al. 2014; Esestime et al. 2016; Cozzi et al. 2021; El-Orbi et al. 2022; Lottaroli and Meciani 2022). Outcrops of isolated and attached Tethyan carbonate platforms with reservoir potential are widespread around the eastern Mediterranean Basin (Fig. 2) and lend support to this hypothesis (Rusciadelli and Shiner 2018). Moreover, in the period since the Zohr discovery, further rumoured carbonate discoveries have been made, although details are sparse (Fig. 1).

Within the eastern Mediterranean Sea, particular focus has been given to the area surrounding the prominent, basin-centre Eratosthenes High (Fig. 1), because its promontories and outliers formed subsidiary elevations upon which Zohr-like isolated build-ups might have seeded (e.g. Roberts and Peace 2007; Sagy et al. 2015; Burchette et al. 2019; Van Simaeys et al. 2019; Lottaroli and Meciani 2022). Indeed, the Eratosthenes High as a whole has long been interpreted as a large, enduring isolated carbonate platform (Robertson et al. 1998; Montadert et al. 2014; Papadimitriou et al. 2018), a conclusion that our study supports.

This contribution provides a clear view of the seismic stratigraphy over the southeastern portion of the Eratosthenes High in the offshore Cyprus exploration Block 12 (Fig. 1). We show the presence of three distinct carbonate platforms of widely differing styles, outline the seismic criteria for their identification, and examine the controls on their growth and demise. We also suggest how these platforms may relate to regional analogues, and offer additional seismic stratigraphic and structural observations relevant to understanding the nature and evolution of the Eratosthenes High and associated potential petroleum reservoirs.

The evolution of the complex eastern Mediterranean Basin has been studied for over half a century (e.g. McKenzie 1970; Lort 1971; Aksu et al. 2005; Netzeband et al. 2006a; Robertson and Mountrakis 2006; Robertson et al. 2012; Klimke and Ehrhardt 2014; Inati et al. 2016; Steinberg et al. 2018; Nirrengarten et al. 2023; and references therein) and the reader is directed to those sources for details. The following brief summary provides context for our discussion.

The Levant Basin is located within the eastern Mediterranean Sea and formed during the early Mesozoic break-up of Pangaea when the Troodos Plate rifted away from the African–Arabian Craton, resulting in opening of the Neo-Tethys Ocean. Rifting along the southern Neo-Tethys margin occurred in several phases, each involving extension, subsidence, and widespread intrusive and extrusive magmatism (Wilson et al. 1998; Gardosh and Druckman 2006; Skiple et al. 2012); all of these events are reflected in the present tectonic architecture of the region (Fig. 3).

Rifting occurred in the Late Triassic and earliest Jurassic, accompanied by widespread, periodic volcanic activity (May 1991; Laws and Wilson 1997; Robertson et al. 1998, 2012; Wilson et al. 1998; Gardosh et al. 2010). More pronounced extension occurred at around the Jurassic–Cretaceous boundary, although the timing of these events remains poorly constrained (Laws and Wilson 1997; Wilson et al. 1998, 2000; Homberg et al. 2009). Regional uplift, erosion, subsidence, and extrusive and intrusive magmatism continued through the Early Cretaceous (Homberg et al. 2009). Garfunkel (1991), Wilson et al. (1998), Segev (2000) and Segev and Rybakov (2010) have proposed that a mantle plume or plumes drove these Mesozoic events, partly on the basis that significant regional uplift (the ‘Levant–Nubia dome’) and magmatism occurred over this period in northern Israel, Lebanon and Syria, and volcanic seamounts are present in Turkey (Robertson et al. 2012, 2016); such events typically generate large volumes of intrusive and extrusive basic magma (Roberts et al. 1979; White and McKenzie 1989; White and Lovell 1997; Geoffroy 2005). Despite protracted, multiphase extension during evolution of the Levant Basin, the underlying basement appears to be mostly continental rather than oceanic (Hirsch et al. 1995; Garfunkel 1998; Netzeband et al. 2006a; Roberts and Peace 2007; Segev and Rybakov 2010). The Eratosthenes High, now located at the northern edge of the African Plate (Fig. 3), remains linked to the less-deformed Levant Basin margin by a terrain of hyper-extended continental crust (Makris et al. 1983; Robertson et al. 2012; Klimke and Ehrhardt 2014; Welford et al. 2015; Inati et al. 2016), although one more recent geophysical study (Feld et al. 2017) has suggested the presence of incipient oceanic crust to the south of the high.

Convergence of the African and Eurasian plates began during the Cenomanian (Müller and Roest 1992; Robertson 1994; Bowman 2011; Klimke and Ehrhardt 2014). Relative plate motions accelerated during the Campanian, resulting in the closure of the Neotethys Ocean and ophiolite obduction at different times and locations onto the Eurasian and Arabian plates. These events also led to initiation of the Cyprus Arc and the Latakia thrust zone (Fig. 3) in Cyprus (Garfunkel 2004; Netzeband et al. 2006a; Bowman 2011; Montadert et al. 2014), and promoted inversion and crustal shortening, often influenced by basement structures, throughout the Levant region (e.g. Syrian Arc folds and the Palmyrides) (Garfunkel 1998; Walley 1998; Vidal et al. 2000; Brew et al. 2001; Roberts and Peace 2007; Frizon de Lamotte et al. 2011; Ghalayini et al. 2014). From the Maastrichtian to the mid-Eocene, the basin accumulated thick deep-water marls (May 1991; Roberts and Peace 2007; Welford et al. 2015; Steinberg et al. 2018); other late Cretaceous sediments are condensed or absent. Except for the recently discovered Zohr gas field (Bertello et al. 2014; Esestime et al. 2016), and several other Mesozoic carbonate discoveries since (Fig. 1), larger petroleum reservoirs in the basin reside in Oligo-Miocene siliciclastic sediments where they onlap and drape over older geomorphic highs (e.g. Leviathan field) or are incorporated into younger inversion structures (e.g. Tamar field).

The current eastern Mediterranean Basin is bound to the north by the Cyprus subduction zone in which ocean-floor and passive-margin carbonates of the Cyprus and Latakia basins have overridden the more mildly deformed Levant continental margin and its Mesozoic–Pliocene cover (Robertson 1998a, b; Plummer et al. 2013; Klimke and Ehrhardt 2014). The bathymetry surrounding the Eratosthenes High reflects this complex geological setting (Kokinou and Panagiotakis 2018).

The Paleogene and Neogene sections across the southern Levant Basin are widely correlatable and dated by reference to exploration wells in offshore Egypt, Israel and Cyprus, but the deeper basinal Mesozoic stratigraphy has not yet been penetrated. In the absence of calibration, determining the timing and equivalence of Mesozoic structural and stratigraphic events is difficult throughout the basin and relies on extrapolation from the more accessible basin margins and near-shore exploration wells. Seismic quality at depth within the basin is also only marginally adequate for such purposes due to the presence of overlying thick Messinian salt and the fact that seismic acquisition parameters have typically been optimized for shallower formations. This is a particular issue with respect to potential Mesozoic isolated carbonate prospects within the deeper Levant Basin. The Eratosthenes High has thus become central to understanding these questions as it offers: (a) the opportunity to sample the stratigraphy via shallow boreholes; and (b) access to high-quality 3D seismic data without the complications of a thick salt overburden.

The Eratosthenes High is a flat-topped submarine rise, c. 100 × 80 km, located between Cyprus and Egypt. Its northern margin is c. 80 km south of the Cyprus coast (Fig. 1). The water depth over the crest is c. 1000 m and deepens to c. 3000 m in the adjacent oceanic basins. Despite the prominence of this feature in the seascape of the eastern Mediterranean, its origin, geology and tectonic significance remain incompletely understood.

A step change towards determining the nature of the Eratosthenes High followed Leg 160 of the Ocean Drilling Program (ODP) in 1995 (Kempler 1998; Mart and Robertson 1998). This survey drilled three boreholes and several shallow pilot holes on its northern flank (Figs 1 and 4) and dredged seafloor samples from the crest. Results from these boreholes form data points in every subsequent study. While the ODP boreholes provided robust lithological, environmental and age data for the shallowest section, they offered little information on the deeper stratigraphy of Eratosthenes High, which has so far been interpreted only from poor-quality regional 2D seismic data (e.g. see Galindo-Zaldívar et al. 2001; Roberts and Peace 2007; Peace et al. 2012; Klimke and Ehrhardt 2014; Montadert et al. 2014; Papadimitriou et al. 2018).

The Eratosthenes High has been colliding with Cyprus since at least the late Miocene, causing updoming and pronounced east–west extensional faulting over its crest (Robertson 1998a, b; Galindo-Zaldívar et al. 2001). Significant magnetic and gravity anomalies exist beneath the structure and point to large volumes of volcanics at depth (Ben-Avraham et al. 1976; Erbek and Dolmaz 2014). The southern margin of the Eratosthenes High is fault-bounded, with downthrow on the basinward margin of c. 4 km at the Cretaceous level; in the basinward hanging wall, a series of downfaulted blocks exhibit growth sedimentation (Fig. 4).

The current seafloor extent of the high corresponds to neither its structural nor depositional footprints as its true margins are buried by up to 1200 m of Messinian salt on all sides and are obscured from surface view (e.g. Figs 4 and 5a).

Speculative sparse 2D seismic surveys acquired in 1975 by CGG-Spectrum provided the first regional seismic data across the Eratosthenes High, although quality and processing were limited (e.g. see Roberts and Peace 2007). Further regional 2D seismic surveys were carried out in 1993 and 2000 (e.g. see Peace et al. 2012). All have been utilized in numerous published studies in the intervening period, and selected lines and learnings from these datasets were incorporated into this study.

Our study utilized three seismic datasets: (1) a 3628 km² proprietary 3D survey acquired by Noble Energy in 2011 on behalf of the (then) Cyprus Block 12 lease holders (Fig. 5a, b); (2) a 9500 km proprietary dense 2D (2 × 2 km) survey acquired in 2007 by Noble Energy on behalf of the lease holders; and (3) three CGG-Spectrum speculative regional 2D lines, reprocessed and depth migrated in 2000. The first two datasets provided good images of the subsurface, while the latter is of only moderate to poor quality but gave essential regional context by allowing a tie-in of nearby ODP boreholes and extension of interpretation outside the 3D survey areas. All seismic datasets were depth migrated.

The 2D and 3D seismic datasets were combined into a single project. Seismic stratigraphic units were interpreted and mapped initially on the 3D survey and subsequently out to the limits of the data. All interpretation was carried out on depth-migrated data to ensure, as far as possible, accuracy with respect to thicknesses and geometries of both stratigraphic and structural features. The seismic character, the internal morphology, and the geometries of the interfaces with underlying and overlying units were all used to determine the seismic stratigraphic and structural significance of features based on industry-standard techniques (e.g. Bubb and Hatlelid 1977; Mitchum et al. 1977; Vail et al. 1977; Sheriff 1980; Skirius et al. 1999; Catuneanu et al. 2009). Principal horizons were then extrapolated where possible to regional tie points in the Levant Basin in order to constrain ages, although in this regard significant uncertainties result from abrupt differences in the structural elevation of equivalent units between basin and high.

Velocity contrasts in the highly deformed and lithologically diverse shallow section around the margin of the high create frequent columnar blank data zones, pull-ups and thickness artefacts in deeper intervals that can impair seismic interpretation. Imaging beneath the Messinian salt wedge, which is strongly disrupted by faulting and thrusting, is poorest, while away from the salt cover the resolution and data consistency in the shallow section are good and locally excellent. Some artefacts could not be completely eliminated, even on the depth-migrated data.

The stratigraphy of the southeastern Eratosthenes High has been divided into nine principal seismic stratigraphic units, which are summarized in Table 1. The deepest of these are poorly imaged, and, hence, are difficult to accurately characterize and interpret. Some of the stratigraphic units identified are broadly equivalent to previous interpretations made from older 2D data (e.g. Montadert et al. 2014; Papadimitriou et al. 2018) and this is indicated where appropriate. Only general accounts are given in this section; specific aspects are discussed and illustrated more fully later. Justifications for the age determinations given here and in Table 1 are discussed in a later section.

• Deepest section (Paleozoic basement?): the deepest section comprises low-reflectivity, often noise-blanked, panels of high-amplitude, discontinuous reflectors (Fig. 4). Flattening on certain shallow horizons creates geometries at this level that are suggestive of fault blocks or half-graben with a wedged infill. Due to poor imaging, over-interpretation of this interval is a hazard and it is not discussed further. However, its top in the centre of the Eratosthenes High, at around 5 km below sea level (BSL), is consistent with estimates based on modelling by Klimke and Ehrhardt (2014) and Feld et al. (2017) for crystalline basement.

• Unit 1 (Permo-Triassic?): Unit 1 consists of a lower, vaguely stratified interval (Subunit 1A) of variable thickness up to 1800 m, with discontinuous, parallel to divergent reflectors of predominantly higher amplitude (Fig. 4). The upper third (Subunit 1B) is of more uniform thickness (c. 700 m), displaying only discontinuous low-amplitude reflectors, which in places appear to onlap the top surface of Subunit 1A. Unit 1 is most clearly discerned on sections flattened on shallower picks and appears to infill underlying topography, implying a possible synrift origin where the footwalls lay on the northwestern sides of half-graben. Both subunits are obscured or absent on 2D data towards the middle of the high.

  Unit 1 is equivalent to the intervals designated ER-E and ER-F in Montadert et al. (2014) and ESP1 in Papadimitriou et al. (2018). Both authors allocated a mid-Jurassic age to this interval. Mounded geometries, interpreted by Papadimitriou et al. (2018) as transgressive isolated carbonate platforms, could not be identified in our dataset.

• Unit 2 (Triassic or Lower Jurassic): this stratified interval comprises high-amplitude, discontinuous, sub-parallel or slightly undulating reflectors that locally appear truncated at the upper surface. The geometry of Unit 2 is wedge-shaped, thinning markedly in the study area from around 1700 m in the centre of the Eratosthenes High to less than 200 m in the south, east and west over a distance of c. 50 km (Figs 4, 6 and 7a). Reflectors brighten and become more defined and continuous as the unit thins towards the margins of the Eratosthenes High; processing has enhanced this attribute in the 3D dataset. On strike lines, Unit 2 exhibits broadly parallel top and bottom boundaries and is abruptly truncated at the margin of the high (Fig. 8).

  Unit 2 is equivalent to ER-D of Montadert et al. (2014) and ESP2 of Papadimitriou et al. (2018); both allocated it a Jurassic age and illustrated the unit as a dome-like body nested within the periclinal morphology of the high.

• Unit 3 (Jurassic): this 1200–1800 m-thick, tabular succession exhibits consistent, low- to moderate-amplitude, parallel-continuous internal layering (Fig. 6). It thickens by c. 30% towards the SE, much of this in the upper two-thirds of the unit (Fig. 7b), and is truncated everywhere at the fault-controlled escarpment at the southern edge of the high (Figs 6 and 8). Noise, velocity anomalies and abundant faults around the margin generate frequent vertical seismic blank zones, spurious structural anomalies and pull-up or push-down artefacts. Variations in reflector character suggest that there may be four or more depositional cycles within Unit 3.

  The lower boundary is planar, and is characterized by a fairly continuous, moderate-amplitude reflector (Fig. 6). Internal reflectors are concordant with this surface, with no discernible onlap onto the top of Unit 2. The top boundary is a high-amplitude continuous reflector or reflector pair, which is strongly rugose.

  This unit is equivalent to the interval identified as ER-C on regional 2D lines by Montadert et al. (2014) and as ESP3 by Papadimitriou et al. (2018), and was interpreted by both as a shallow-water carbonate platform of Lower Cretaceous age.

• Unit 4 (Aptian–Cenomanian): Unit 4 is up to c. 900 m thick (Fig. 7c) and is separated from underlying Unit 3 by the high-amplitude rugose reflector described above, and from Unit 5 by a moderate- to locally high-amplitude continuous reflector. There is no discernible onlap onto the top surface of Unit 3.

  Unit 4 comprises two composite sequences, Subunits 4A and 4B (Fig. 6), and subordinate sequence sets. Subunit 4A has an aggradational ramp-like morphology and extends across the whole study area. It is thickest (up to 500 m) in a belt close to the margin of the Eratosthenes High but thins gradually northwestwards to around 100 m over a distance of 60 km (Fig. 6). Seismic facies consist of parallel-continuous to discontinuous reflectors of variable amplitude, thickening patchily, with a rather chaotic internal character in the NW of the study area. It is separated from Subunit 4B by a moderate- to high-amplitude continuous but rugose, reflector.

  Subunit 4B exhibits a complex internal stratigraphy of numerous shingled, sigmoidal, NW-directed clinothems (Fig. 6) grouped into at least three additional high-frequency sequence sets. Clinoform belts transition laterally into parallel-continuous topsets and bottomsets to the SE and NW, respectively. Seismic amplitudes are highly variable throughout this unit. Deposition was strongly influenced by synsedimentary NW–SE strike-slip fault corridors and sags. In the north of the area, large mounded bodies prograded into the sags from the inter-sag ridges (Fig. 6). Regional 2D seismic lines north of the study area show that progradation also occurred from the west and east during this stage to isolate an intrashelf basin.

  Unit 4 is equivalent to ER-B of Montadert et al. (2014) and appears to be equivalent to ESP4 of Papadimitriou et al. (2018); these authors extrapolated this interval across the whole of the Eratosthenes High and suggested a mid- to Late Cretaceous age.

• Unit 5 (Turonian–Miocene): Unit 5 exhibits a layered seismic facies characterized by continuous, high-amplitude parallel reflectors (Fig. 6); the lower portion is often less reflective than the upper. The lowest internal reflectors show local onlap onto the top of Unit 4 (Fig. 6), but the two units are otherwise concordant. The unit thickens gradually from c. 60 m around the southern edge of the intrashelf basin to local maxima of 250 m in sags and behind the Eratosthenes escarpment (Figs 6 and 7d). There is no seismically discernible progradation along the margin facing the Levant Basin.

  Unit 5 thickness is influenced by relief on top of Unit 4 and by deep channels incised into its top. In the intrashelf basin, Unit 5 thickens to over 600 m and has lower reflectivity than at other locations (Fig. 6). This expanded section reveals stratigraphic complexity that cannot be resolved in the thin, conformable ‘topset’ but which has proved to be an essential key to correlation at this stratigraphic level. Unit 5 has been subdivided into four subzones, 5A–5D, discussed in detail later.

  Unit 5 is broadly equivalent to ER-A of Montadert et al. (2014), and appears to be equivalent to both ESP5 and ESP6 of Papadimitriou et al. (2018). Both authors interpreted this interval as pelagic sediments, ranging in age from Paleogene to Miocene.

• Unit 6 (Messinian): Updip, this interval forms an impersistent veneer less than 80 m thick of high-amplitude, mostly parallel-discontinuous seismic facies that directly overlies Unit 5, but it is often absent or below seismic resolution. Unit 6 fills erosional channels along the routes of antecedent fault-controlled sags, comprising high-amplitude, concave-up layers or chaotic masses. Channels up to 250 m deep have locally incised through Unit 5 into the top of Unit 4.

  Downdip, this unit is more complex. The lower part, Subunit 6A (Fig. 6), forms a consistent apron or terrace up to 130 m thick along the whole southeastern margin of the Eratosthenes High, locally abutting erosionally truncated Unit 5. In dip sections it forms sheets of high-amplitude, parallel-discontinuous or inclined reflectors; slides and slumps are common. In strike view, the seismic facies consists of high-amplitude, parallel-discontinuous to continuous reflectors, locally expanding into pods of chaotic character. The upper part, Subunit 6B (Fig. 6), comprises several variable sequences of very low-amplitude, wedge-shaped clinothems, which in most places offlap and downstep basinwards (Fig. 6), but in some they prograde or aggrade such that the topsets onlap Subunit 6A. This unit thickens basinwards, in places exceeding 400 m.

• Unit 7 (Messinian): Unit 7 is restricted to the SE and represents basin-filling halite of Messinian age (Hsü et al. 1977; Bertoni and Cartwright 2006), ranging in thickness from 2500 m in the adjacent Levant Basin to 900 m in the disturbed zone above the Eratosthenes High margin (Fig. 9), overlapping this in places by 15 km. The base of Unit 6 in the Levant Basin is marked by a thin package of high-amplitude, parallel-continuous reflectors that are little-disturbed by faults, while the top is marked by a thin, high-amplitude double seismic reflector that is extremely disrupted due to faulting, thrusting and folding within the evaporites (Figs 6 and 8).

  The bulk of Unit 7 adjacent to the escarpment is seismically transparent or contains only weak, irregular or highly discontinuous, shingled to recumbent reflectors (Figs 6 and 8). In areas away from the high, two or three groups of discontinuous, high-amplitude reflectors are present (e.g. Fig. 4). Previous studies have subdivided the eastern Mediterranean Messinian evaporite succession into several seismic stratigraphic units based on intercalations of non-evaporitic sediments (Bertoni and Cartwright 2006, 2007; Netzeband et al. 2006b; Gvirtzman et al. 2013; Feng et al. 2016, 2017; Güneş et al. 2019); all appear to be present here.

  The leading edge of the Messinian salt is a terrain of stacked thrusts and folds that wedges out bluntly. Overlying Plio-Pleistocene sediments have been shortened by at least 5 km and elevated by c. 500 m (Fig. 6). Much of the deformation appears to have been caused by extrusion of basal salt layers, noted by Libby and Underhill (2015), a phenomenon well imaged in our data (Figs 6 and 10). Regional salt migration within the Levant Basin is related to the loading of basin-centre evaporites by the Nile cone, and by northwestward gravitational translation of the salt and overlying sediments (cf. Libby and Underhill 2015; Allen et al. 2016; Ben Zeev and Gvirtzman 2020; Zucker et al. 2020; Hamdani et al. 2021).

  Halokinesis created a pronounced circumferential depression around the periphery of the Eratosthenes High, sometimes characterized as a ‘moat’ (Robertson 1998a; Montadert et al. 2014; Libby and Underhill 2015), that has been partly infilled by Pliocene–Holocene sediments (units 8 and 9). This basin, triangular in cross-section (Figs 4 and 6), was termed the ‘South Eratosthenes Basin’ by Galindo-Zaldívar et al. (2001); a superficially similar northern basin, termed the ‘North Eratosthenes Basin’ by the same authors, corresponds to the Cyprus Trench. Similar relationships exist on the eastern and western sides of the high (Montadert et al. 2014; Papadimitriou et al. 2018), although there is asymmetry in the degree of salt superposition around its flanks (Klimke and Ehrhardt 2014, fig. 1; Kokinou and Panagiotakis 2018).

• Unit 8 (Pliocene): Unit 8 forms a wedge of sediment up to 200 m thick with low-amplitude, parallel-continuous seismic facies, and indistinct top and bottom boundaries, which covers Unit 6 within the moat (Fig. 6). Significant deformation of both Unit 6 and Unit 8 has occurred in front of the allochthonous salt wall and this has deformed their distal correlatives, complicating basinward correlation. Favourable lines show Unit 8 to be contiguous with the regional post-salt sediment blanket. Within the Levant Basin, Pliocene sediments have a fairly uniform thickness of c. 300 m and consist of parallel-continuous seismic facies in which reflector amplitudes increases upwards (e.g. Fig. 8); strata everywhere exhibit an undulating, faulted and folded character caused by deformation of the underlying salt.

• Unit 9 (Plio-Pleistocene–Holocene): this forms the final sedimentary drape in all areas, the top surface of which is the seafloor (Figs 6 and 8), and has an almost transparent parallel-continuous layered seismic facies throughout. On the Eratosthenes High, its lower boundary subtly onlaps residual topography in units 6 and 8, as well as the margins of some under-filled channels. The unit undulates minimally in thickness but is somewhat thicker (60–100 m) over the Eratosthenes High and in front of the allochthonous salt wall (up to 200 m) than in the adjacent Levant Basin (c. 60 m). ODP Leg 160 site surveys proved up to 100 m of Pleistocene–Holocene pelagic muds, sapropels and volcanic ash layers at the crest of the Eratosthenes High (Limonov et al. 1994). The Plio-Pleistocene cover in this part of the Levant Basin is relatively thin due to a limited contribution from regional siliciclastic sediment sources such as the Nile Delta at this time (e.g. Sagy et al. 2020).

The Eratosthenes High has been an elevated, isolated feature within the Levant Basin since the early Mesozoic, making it very unlikely that a siliciclastic succession at the scale of Unit 3 could be intrinsically or extrinsically sourced. The tabular geometry, layered internal architecture and seismic facies of this unit (Figs 6–8) are characteristics of many large-scale carbonate platforms (cf. Philip et al. 1995), but there are otherwise few unequivocal seismic stratigraphic indicators of depositional environment. Typically, such platforms are hundreds or thousands of metres thick and laterally persistent for tens to many hundreds of kilometres. The flat tops of shelves of this type comprise shallow-water environments, often with restricted circulation, at around sea level, which generate horizontally layered successions of stacked sedimentary cycles (Eberli et al. 2004; Bosence et al. 2009). This makes up volumetrically the largest portion of the platform, often extending over tens of thousands of square kilometres, typically with monotonous parallel-continuous seismic facies (e.g. Fontaine et al. 1987; Macurda 1997). Unit 3 represents the fault-truncated interior of such a platform and shows consistent horizontal, parallel-continuous seismic facies throughout the more than 4000 km² of the study area.

Where they face deep basins, such platforms often exhibit truncated escarpments rather than progradational margins, reflecting their origins in rift-margin settings (cf. Read 1982b; Schlager 2005). Remnants of similar carbonate platforms, ranging in age from Triassic to Cretaceous, are exposed in widespread locations around and within the eastern Mediterranean Basin (Bosellini 1989; Philip et al. 1995; Rusciadelli and Shiner 2018), including the peri-Adriatic regions of Italy (Eberli et al. 1993; Peace et al. 2012; Santantonio et al. 2013; Morsilli et al. 2017), Greece (Funk et al. 1993) and the Balkans (Obradović et al. 1993; Vlahović et al. 2005), the Levant region of Israel and Lebanon (Sass and Bein 1982; Ross 1992; Bromhead et al. 2022), Egypt (Tassy et al. 2015), and the Taurides in Turkey (Yilmaz et al. 2016) (Fig. 2). Malta and Sicily represent fragments of such platforms (e.g. Micallef et al. 2016). Platforms of this type also form the Mesozoic stratigraphy of the passive Atlantic margins of Africa and the Americas (e.g. Jansa and Wiedmann 1982; Wissman 1982; Schlager and Camber 1986; Schlager 1989; Davison 2005). An extant example, and the only large legacy fragment of the Tethyan Mesozoic carbonate platforms, is the Great Bahama Bank where seismic facies of the interior similarly consist of horizontal, parallel-continuous reflectors (cf. Sarg 1988; Eberli and Ginsburg 1987, 1989; Macurda 1997).

Our interpretation of the regional 2D seismic grid supports the conclusion that Unit 3 probably covered the whole of the Eratosthenes High to form a major carbonate platform nearly 150 km across (cf. Robertson et al. 1998; Klimke and Ehrhardt 2014; Papadimitriou et al. 2018), isolated within the eastern Mediterranean Basin and with a bathyal environment on all sides. This is an area much larger than the Bahamian Turks and Caicos platform and approaches the scale of the Great Bahama Bank.

Down-to-the-basin faults extend along the whole margin, locally creating partially detached slivers and blocks (e.g. Fig. 4). The platform-interior seismic facies of Unit 3 are truncated at the escarpment and the Unit 4 platform inherited this feature so that their structural margins are broadly coincident (Figs 6 and 10b–e). The escarpment has a markedly scalloped morphology (Fig. 10a), a phenomenon that is common in both modern and ancient carbonate escarpment margins (Mullins et al. 1986; Schlager and Camber 1986; Mullins and Hine 1989; Hine et al. 1992; Stewart et al. 1993; Rusciadelli et al. 2003; Eberli et al. 2004; Principaud et al. 2015; Etienne et al. 2021) and steep volcanic marine slopes (e.g. see Masson et al. 2002; Mitchell et al. 2002; Boulesteix et al. 2013; Crutchley et al. 2013); these features represent the arcuate scars generated by collapse and wasting of unstable slopes above listric faults.

Our study area forms a broad southeastwards palaeopromontory, the SW and NE margins of which are truncated by broader scallops (Figs 8 and 10a) that probably have a similar origin. It is notable that the footprint of the Eratosthenes High mapped from regional 2D data by Montadert et al. (2014, poster 2) and Esestime et al. (2016, fig. 4) exhibits numerous large arcuate re-entrants, some of which are tens of kilometres across, suggesting that margin collapse possibly occurred at a basin-significant scale (cf. Etienne et al. 2021). Such features are common on other steep-margined Cretaceous platforms in areas around the Mediterranean (e.g. Bosellini and Morsilli 1994; Bosellini et al. 1999; Tassy et al. 2015; Le Goff et al. 2019). While difficult to measure, the escarpment could also have experienced erosional retreat (cf. Schlager et al. 1984; Schlager and Camber 1986); the Florida and Bahamian escarpments, for example, have retreated by tens of kilometres in places (see e.g. Mullins et al. 1986, 1991; Mullins and Hine 1989). It could be speculated that the Zohr structure also represents a remnant of a once much larger Eratosthenes platform, the original margins of which have been excavated by collapse.

A substantial base-of-slope talus apron or megabreccias might be anticipated on the adjacent basin floor, as these are common on modern and ancient carbonate escarpment margins (e.g. Freeman-Lynde and Ryan 1985; Eberli et al. 2004; Janson et al. 2011; Jo et al. 2019; Lehrmann et al. 2020; Etienne et al. 2021), but it is difficult to identify the basinal correlatives of the shallow-water succession upon the high. Strata in the deep hanging wall (7000–12 000 m) show a range of geometries, including growth sedimentation (Fig. 4), and it seems possible that a foundered extension of the truncated platform may be present at this level. Basinward-wedging geometries do exist in the deeply buried hanging-wall strata but imaging is poor at depth and there is thus little unequivocal evidence for such features. Resedimented deposits of equivalent age to the shallow-water carbonate succession have been interpreted on older 2D data on the western margin of the Eratosthenes High (cf. Papadimitriou et al. 2018), although these are similarly poorly imaged.

An enigmatic upturned ‘lip’ is present in our data along the whole length of the escarpment (Figs 4, 6 and 10b–e), most notably in units 2 and 3. A pronounced elevated edge is illustrated on older 2D time sections and has generally been characterized as a pull-up velocity artefact (e.g. see Montadert et al. 2014; Papadimitriou et al. 2018). This mounded and often seismically blank feature superficially resembles a constructional platform rim but, where resolvable, the internal seismic facies is layered and similar to Unit 3 as a whole (e.g. Fig. 10). Given the lithologically diverse, structurally complex nature of this margin with abrupt lateral juxtaposition of carbonates, salt and shales, this is considered to be in large part a processing artefact. Internal reflectors within Unit 3 also diverge gently towards the SE, and the succession thickens by c. 30% over a distance of around 30 km (Figs 6 and 10). This does not appear to be a velocity artefact but may be the consequence of flexural subsidence during accumulation.

The Unit 3–Unit 4 transition represents a relative sea-level rise and back-step of a platform margin. Whether there was an intervening episode of subaerial exposure and karstification or of deeper-water sedimentation is unclear. The pervasively rugose, high-amplitude character of the top-Unit 3 reflector may signify karstification, although there is no resolvable epikarstic relief, erosion or collapse. The interior and north of the high were initially more deeply submerged than the southern and western margins.

The southern Eratosthenes High is traversed by prominent NW–SE-trending, vertical or steeply inclined normal faults spaced at between 2 and 10 km. The faults offset Unit 1 and Unit 2 by several hundred metres and at the Unit 3–Unit 4 boundary splay upwards into younger strata to form intensely faulted, linear synsedimentary ‘sags’ up to 10 km wide (Figs 8 and 11a, b). Active fault movement commenced during deposition of Unit 4B and residual subsidence persisted until the end of deposition of Unit 5. Unit 3 thickness and seismic facies are unaffected by either faulting or sag development.

The linear, upward-splaying character of the faults hints at a strike-slip character, but there is little evidence for lateral offset, shear or en echelon segments when viewed on depth or horizon coherency displays, mainly due to the lack of seismic reference features in the deeper section. The deep-seated nature of the master faults, and their orientation parallel to much older fault trends that traverse the Levant and Herodotus basins (cf. Aal et al. 2000; Longacre et al. 2007; Segev and Rybakov 2010), suggests that they may reflect reactivation of basement lineaments older than the Mesozoic rifting events. The widths and close spacing of the sags means that little of the area of the platforms in units 4 and 5 was spared the impact of synsedimentary subsidence, and they had a profound influence on the evolution of the Unit 4 carbonate platform.

Syndepositional faulting in the Unit 4 platform reflects a dynamic tectonic setting, but it is difficult to identify a causative event of late Lower or mid-Cretaceous age. The principle tectonic phenomena in the eastern Mediterranean Basin during this period were driven by convergence of the African–Arabian and Eurasian plates; this culminated in a northward-facing subduction zone, and subsequent southward obduction of the Troodos and other ophiolite complexes onto this margin during the Late Cretaceous, as well as the ‘Syrian Arc’ compression (Robertson 1998d; Sagy et al. 2015). Although the Eratosthenes High crustal fragment was further from the Cyprus trench at this time, it is possible that the faulting represents an early ‘far-field’ response to this event, during which older lineaments in the Arabian Plate, possibly of Triassic or Early Jurassic age, were reactivated as strike-slip faults (cf. Le Pichon et al. 2019; Van Hinsbergen et al. 2020).

Unit 4 platform stratigraphy is complex and evolved in four stages (Fig. 12), here informally termed ‘sequence sets’:

• Sequence Set 1 (SS1): the new carbonate platform seeded at numerous locations along the elevated margin of the Eratosthenes High during or following flooding of Unit 3. It evolved as an aggradational, tabular bank (Subunit 4A) up to 500 m thick and up to 10 km wide in the dip direction along the whole of the southeastern edge of the Eratosthenes High (Fig. 6). Regional 2D data show that this complex extends beyond the study area to the west and east and probably rimmed at least the southern portion of the high (Fig. 13a). The seismic facies in this zone is parallel and continuous, consisting of up to three moderate-amplitude reflectors. Except in advantageous windows, details in the critical area adjacent to the escarpment are obscured by seismic blank zones beneath the salt and on many lines there is little resolvable internal structure. The basinward edge is truncated where it coincides with the Unit 4 escarpment but is otherwise set back by up to 5 km; reflector terminations in such settings are locally basinward-dipping (e.g. see Fig. 6). The aggradational character and lack of progradation during this period suggests that the platform was challenged by rising relative sea level. While growth was mostly of keep-up style, the initial phase was likely to have been in catch-up mode.

  Along the internal margin, this subunit remained morphologically ramp-like, extending northwestwards into the interior of the high for at least 60 km, thinning gently to c. 100 m (Fig. 6). In the NW of the study area, 35–40 km from the escarpment margin, local thickening occurs at this stratigraphic level, representing the first indications of larger mounded bodies and progradational lobes that developed in later sequence sets.

  Context, seismic facies and geometry suggest that the interior of the Eratosthenes High at this stage was a sheltered, open-water internal sea (Fig. 13a), possibly dominated by neritic carbonate or chalk deposition in up to 200 m of water, based on the height of the ramp-like internal margin.

• Sequence Set 2 (SS2): an event, tentatively interpreted as a relative sea-level fall, promoted strong progradation of the interior margin of the platform towards the NW (Figs 6 and 13b). At least five high-frequency sequences are bundled in this initial phase of progradation. Shelf-slope breaks exhibit a nearly flat trajectory over a distance of 10 km, with topset aggradation of about 150 m (Fig. 12). Each sequence consists of a well-defined belt of sigmoidal clinoforms, often syn- or post-depositionally distorted by faults and slumps, which grades towards the SE into parallel-continuous topsets. These onlapped and covered the marginal bank in SS1 but are also truncated at the escarpment so that the nature of the basinward margin remains uncertain. To the NW, the initial clinothems downlap directly onto the top of SS1 (Fig. 12), while later clinothems pass into parallel-continuous bottomsets that persist for tens of kilometres into the centre of the high and concordantly overlie those of SS1 (Figs 6 and 12). Referenced to clinoform height, the water depth in the interior of the high during this phase of growth was of the order of 200 m.

• Sequence Set 3 (SS3): progradation continued during deposition of SS3 for a further 10 km. The shelf-margin trajectory in this sequence set steepens, suggesting that the rate of relative sea-level rise accelerated (Fig. 12). This was accompanied by a change in clinoform steepness from c. 3° in SS2 to c. 11° in SS3 and an increase in height to c. 250 m so that the clinothems are narrower than in SS2. Linear depressions continued to develop in the areas affected by the fault corridors (Figs 11 and 13c) and these were rimmed by prograding carbonate margins; in the NW, platform growth from the south, east and west enclosed an intrashelf basin.

  The transition from ramp-like to steeper shelf-margin morphology between SS1 and SS2 may signify evolution from shoal-dominated to more generally bioconstructed, possibly rudist-dominated, platform margins. Bottomset deposits towards the centre of the high also thicken in this interval, reflecting a higher basinal sedimentation rate, possibly as a consequence of reduced basin area or an increase in the size or productivity of, or export from, the surrounding carbonate factory. There is considerable variability in the seismic character of these high-frequency sequences depending on seismic line orientation and where syndepositional subsidence modified primary geometries.

  Reflectivity of the boundaries between the correlative parallel-continuous topsets in this bundle is generally weaker than in SS2 (Fig. 12), suggesting relatively more consistent lithologies within the platform interior. Reflective surfaces in the slope and lower slope have higher amplitudes but are sparser and less continuous than in SS1 or SS2 and difficult to track over large distances; this indicates greater impedance variations, perhaps where muddier sediments from the interior basin lapped onto the slopes or where condensed sections and hardgrounds developed during flood backs, or possibly due to thin-bed tuning. As in SS2, clinoforms are often distorted, possibly due to contemporaneous seismicity, slumping or, in places, by post-depositional subsidence and faulting within the sags.

• Sequence Set 4 (SS4): in this final lowstand growth phase (Figs 12 and 13d), depressions within the sags and between prograding margins were infilled to a level close to the top of the Unit 4 platform. The dominant seismic facies in these locations is trough-filling, in several stages, with concave-up reflector-defined packages that thin and lap onto the margins of the depressions (Fig. 14). Infill appears to have occurred in a rolling fashion from south to north as lows filled up and were internalized within the platform. In the intrashelf basin, this final phase of progradation is resolvable as a lowstand wedge. During this period, the surface of the Unit 4 platform must have been subaerially exposed.

The contrast in depositional style between carbonate platforms in units 3 and 4 is striking. The latter exhibits a catch-up and keep-up growth style, with initial aggradation in Subunit 4A and then strong progradation towards the NW for more than 40 km towards a deeper internal sea which formed an ‘empty bucket’ (sensu Schlager 1981) within the centre and north of the Eratosthenes High (Fig. 13f). A map of the thickest portions (topset and slope) of SS1, SS2 and SS3 illustrates the progressive northwestward offset of clinothems across the study area (Fig. 15). Regional data show that progradation in Unit 4 also occurred from both the east and west over at least the southern portion of the Eratosthenes High (see also Montadert et al. 2014; Papadimitriou et al. 2018, fig. 8), suggesting that a complex pattern of peripheral shallow-water platforms and deeper internal basins and troughs existed at this time. The shallow-water platform in Subunit 4B shows comparatively thin topset aggradation compared with the distance it prograded (Fig. 12), reflecting a relatively slow rate of accommodation space growth, with much of the initial height increase having taken place during deposition of Subunit 4A. The total thickness of Unit 4 suggests that this was a largely subsidence-driven process.

The dominant progradation direction in Unit 4 was towards the NW, but carbonate platforms also seeded on the inter-sag highs and prograded along and outwards from these elevations, limited in the deeper depressions by up to 300 m of relief (Fig. 14a, b). The prograding margin of the platform consequently became strongly lobate, controlled by the positions of troughs which were accentuated by carbonate growth as the platform aggraded. Fixed inlines transect mounds and platform lobes obliquely in places so that individual seismic lines sometimes show multiple examples of the northwestward-directed margin asymmetry (e.g. Fig. 14a). Clinothems in Unit 4 commonly collapsed into the troughs or were rotated and steepened by contemporaneous subsidence, possibly in response to seismicity, a common trigger for such events (Hine et al. 1992; Mullins et al. 1992; Spence and Tucker 1997; Rusciadelli et al. 2003). Broader, shallower sags or groups of less pronounced sags were annealed by carbonate growth, becoming progressively infilled from both sides (Fig. 11a, b).

The persistent northwestward progradation points to a strong oceanographic control on platform growth, possibly leewardness to dominant southeasterly winds or currents. Such stratigraphic architecture is highly reminiscent of the interior of the Cenozoic Great Bahama Bank, where the deep escarpment at the Atlantic eastern margin also hindered basinward progradation. Here the dominant northeasterly trade winds drove strong asymmetrical westward progradation of the leeward margins in several internal linear basins and re-entrants that were ultimately completely infilled and eliminated (see Eberli and Ginsburg 1987, 1989). There are remarkable similarities too, both in terms of platform scale and stratigraphic style, to the Eocene–Holocene Maldives ‘mega-atoll’ in the Indian Ocean (Purdy and Bertram 1993). This exhibits a comparable centripetal deposition style, particularly in the Miocene. In the Maldives, peripheral flat-topped carbonate banks, which seeded upon a much more extensive Oligocene carbonate platform, initially aggraded and then prograded for distances of up to 25 km into a 200–500 m-deep internal shelf-sea with only a limited increase in accommodation space (Belopolsky and Droxler 2004a, b). Ocean-facing margins to this complex were escarpments up to 3000 m high that inhibited basinward progradation.

The presence of several prominent lowstands within Unit 4 (Figs 12b and 13) points to an additional shorter-term cyclic control on sedimentation, the signal of which is difficult to deconvolve from other depositional or tectonic processes. Lowstand-mimicking seismic stratigraphic relationships can be generated through the collapse of carbonate margins due to over-steepening or seismicity, or to sediment bypass of a depositional slope. However, recent data show that short-term climatic variations, and thus eustatic sea-level fluctuations, during the Cretaceous were also driven by minor events of polar and continental glaciation (Maurer et al. 2012; Sames et al. 2015; Ray et al. 2019). The magnitudes of the relative sea-level falls measured from depth-migrated seismic data in the examples discussed here, acknowledging the uncertainties this involves, suggests that these were mostly of the order of 20–50 m. Sea-level variations on this scale are broadly consistent with the magnitudes of Cretaceous glaciation-driven eustatic sea-level falls (Ray et al. 2019).

Inward progradation from the peripheries of the Eratosthenes High in Unit 4 (SS2 and SS3) isolated an internal basin or an enclosed southward-projecting re-entrant from the internal shelf sea (Figs 13c and 16). The depression originated in an area of amalgamated sags, most clearly illustrated by strike lines (Fig. 11). The location of the basin was thus structurally determined, but its evolution and isolation were the result of carbonate platform progradation and aggradation; this reduced the dimensions of the basin from initially 50–60 km across to c. 10 km at termination. On its southern side, connection to the Levant Basin was restricted to a narrow seaway c. 10 km long and 3 km wide, also sag-controlled and narrowed by progradation from the sag margins (Fig. 16).

Intrashelf basins are common in the geological record, and have developed at a range of scales and degrees of isolation within carbonate platforms since the Pre-Cambrian (Markello and Read 1981; Read 1982a, b; Vernhet and Reijmer 2010; Bourget et al. 2013). Their origins are polygenetic, related individually or in combination to platform aggradation and progradation, build-up amalgamation, long-term subsidence, or halokinesis. They are particularly characteristic of the broad Mesozoic shelves and epeiric seas that rimmed Tethyan passive margins (Koop and Stoneley 1982; Gerdes et al. 2010). Larger intrashelf basins occur in the Jurassic and Lower–Middle Cretaceous carbonate platforms of Mesopotamia and the Arabian Gulf (e.g. see Burchette 1993; Van Buchem et al. 2002; Droste 2010), and in Lower Cretaceous shelves surrounding the Gulf of Mexico (Wilson 1975; Scott 1993, 2010; Kerans 2002; Osleger et al. 2004) where the stratigraphies of the host platforms resemble those described here. Fringing, aggrading carbonate platforms almost always play a role in their formation and perpetuation. Intrashelf basins also typically experience periodic restriction during their development as free access to the open-marine environment becomes throttled, resulting in temperature or salinity stratification of the water column, which in turn may promote episodes of anoxia and deposition of organic-rich sediments or changes in the carbonate-producing biota (e.g. see Droste 1990; Burchette and Wright 1992; Van Buchem et al. 2002; Razin et al. 2017).

Over most of the area, Unit 5 exhibits many of the seismic stratigraphic characteristics of a terminal or drowning succession (cf. Schlager 1989; Ehrlich et al. 1993) on top of the Unit 4 platform (Figs 6, 11 and 12). Many carbonate platforms in the geological record exhibit substantial caps of either pelagic chalks or, when submerged but still viable as a neritic carbonate factory, open-marine, wave-, storm- and current-reworked packstone or grainstone sediments, or smaller more ecologically challenged build-ups. Isolated platforms may drown in a stepwise fashion, accompanied by a reduction in size of the shallow platform top (Grötsch and Flügel 1992; Saller et al. 1993; Bachtel et al. 2004; Courgeon et al. 2016). The character of drowning successions varies widely, and they often evolve as a deepening-upward succession, from storm- and wave-reworked grainstone and packstone to condensed pelagic chalk, depending on factors such as substrate morphology, the rate of increase in accommodation space and the nature of the carbonate factory, as well as a range of other oceanographic variables (see relevant discussions in Hine and Steinmetz 1984; Schlager 1989; Sattler et al. 2009; Purkis et al. 2014). Drowning is often preceded by subaerial exposure that, after reflooding, hinders immediate re-establishment of a healthy carbonate factory (Schlager 1981, 1989; Ehrlich et al. 1990, 1993; Blomeier and Reijmer 1999; Graziano 1999; Marino and Santantonio 2010).

The northernmost ODP borehole 160-967, low on the flank of the Eratosthenes High (Figs 4 and 17f), proved a deepening-upward Mesozoic section at least 320 m thick, comprising (a) ‘shallow-water’ Aptian limestones (c. 150 m), overlain by (b) Cenomanian–Maastrichtian pelagic marls (c. 240 m) and (c) Paleogene chalks (c. 60 m) (Flecker et al. 1998; Major et al. 1998). The updip wells 965 and 966, c. 20 and 30 km further south, respectively (Figs 4 and 17f), proved shallow-water carbonate sections 180 and 220 m thick, encompassing the whole of the Miocene (Coletti et al. 2019); this overlies at least 50 m of glauconitic late Eocene chalk (Major et al. 1998; Robertson 1998c).

Data quality on legacy 2D seismic lines that criss-cross the Eratosthenes High is poor and published sections are generally presented in two-way time; artefacts and multiples abound. Nevertheless, on these and on our reprocessed and depth-migrated lines, the distinctive parallel-continuous seismic facies of Unit 5 and the transparent seismic facies of Unit 9 are particularly persistent and are visually correlatable from the north of our study area to the ODP borehole locations (cf. Major et al. 1998, fig. 21), a distance of around 30 km, making them laterally equivalent to the sections cored in the crestal ODP borehole 160-965 and 160-966. The thicknesses of units 5 and 8 (Table 1) are also comparable to the drilled Plio-Pleistocene and Miocene sections. These holes, therefore, are very likely to have sampled our Unit 9 and Unit 5 but did not penetrate Unit 4. Both Unit 4 and Unit 5 are probably currently exposed in the north of the Eratosthenes High by the largest east–west faults (see also Mitchell et al. 2013).

The c. 50 m of condensed, glauconitic Late Eocene chalk at the bottom of ODP 160-966 is separated from the Miocene section by a hiatus of around 10 myr (Staerker 1998; Coletti et al. 2019). The boundary between the two is sharp, but whether this is due to omission or submarine erosion is unknown. If the latter, this could point to a previously more complete Paleogene succession, equivalent to the Oligocene pelagic interval identified in ODP 160-967. Erosion of pelagic sediments over oceanic submarine highs and guyots during relative sea-level falls, with consequent changes in energy levels and currents, is a well-documented phenomenon (Karig et al. 1970; Mitchell et al. 2015; Clark et al. 2018). Whatever the case, the presence of Eocene chalk below shallower-water Miocene sediments in the ODP boreholes indicates that the Unit 4 carbonate platform was drowned long before deposition of the Miocene section. Despite a convincing seismic stratigraphic presentation over most of the area, Unit 5 as a whole cannot therefore properly be characterized as a drowning succession over the Unit 4 carbonate platform.

Following demise of the Unit 4 platform, the intrashelf basin and the deepest structural sags remained as the only significant depressions in an otherwise exposed, flat-topped, shallow-water shelf (Figs 13c and 16). The expanded stratigraphy in the intrashelf basin provides critical insight into the complexity of Unit 5 and the events at this transition (Fig. 17). Unit 5 has been divided into four subunits:

• Subunit 5A: the lowstand that terminated the Unit 4 platform was followed in Subunit 5A by a relative sea-level rise of the order of 40–50 m, which deposited a thin, aggradational succession that onlapped residual topography on top of Unit 4 and ultimately blanketed the whole shelf to the south. At the southern edge of the intrashelf basin, this unit expanded into a thin aggradational margin in which clinothems are markedly sigmoidal (Fig. 17). Coeval basin-floor sediments exhibit parallel-continuous seismic facies with local mounds, probably signifying small basin-floor fans (Fig. 17). A predominantly shallow-water depositional system is indicated for Subunit 5A where it overlies the Unit 4 platform top, while in the adjacent basin, based on slope height, water depth was around 270 m.

• Subunit 5B: this subunit, up to 300 m thick and deposited as several packages, infilled most of the accommodation space in the intrashelf basin (Fig. 17). It exhibits low-amplitude parallel, continuous seismic facies throughout and, although it mimics a lowstand succession, each discrete package increases in thickness towards the basin centre and laps onto the margin, but not over it, suggesting relatively passive accumulation of fine-grained sediment, possibly hemipelagic chalks. Deposition of the oldest internal package was restricted to the south of the basin as a slope wedge, suggesting that sediment supply might conceivably have been influenced by the location of the tectonic sags. Tangential toplap at the upper boundary near the southern margin is most likely to have resulted through erosional truncation.

• Subunit 5C: this filled all remaining accommodation space in the basin and exhibits poorly reflective, parallel-continuous seismic facies. In the basin it is c. 120 m thick, possibly reflecting differential compaction in Subunit 5B, and thins at the margin without marked onlap terminations to overlie Subunit 5A everywhere as a veneer up to 60 m thick (Fig. 17). The transparent to weakly layered seismic facies of Subunit 5C points to fine-grained probably pelagic sediment. Context suggests that that this subunit probably represents the Eocene pelagic interval tagged below Miocene carbonates in ODP 160-967.

• Subunit 5D: away from the deeper depressions, in the thin succession overlying the Unit 4 platform, all subunits are difficult to resolve, and Subunit 5B is absent. Subunit 5C is overlain concordantly by the high-amplitude, layered ‘topset’ seismic facies of Subunit 5D (Figs 14 and 17), established above as equivalent to the Miocene succession in the ODP boreholes.

The cored interval in ODP 160-966, the closest borehole to our study area, shows a punctuated, aggrading carbonate succession that coarsens and shallows upwards in three almost equal 60 m stages (Coletti et al. 2019), from (1) an initial somewhat condensed interval with a ‘cool-water’ biota dominated by echinoderms and large benthic and pelagic foraminifera, through (2) a neritic interval dominated by rhodoliths and larger benthic foraminifera, to (3) a fine-grained, coralline algal-benthic foraminiferan-coral ‘reefal’ facies containing miliolids (Emeis et al. 1995; Major et al. 1998; Robertson 1998c; Coletti et al. 2019). The top two of these intervals were demonstrably deposited in the photic zone, while the lower was probably deposited in water depths greater than 60 m (Coletti and Basso 2020). This trend is illustrated by gamma-ray logs from both of the crestal ODP 160 boreholes, which show aggradational profiles in the lower two units, only cleaning and shallowing upward for the top 20 m in ODP 160-966 (e.g. see Major et al. 1998). This catch-up progression of carbonate facies (sensu Schlager 1981) encompasses the whole Miocene (Coletti et al. 2019) and is interpreted to have been enabled by a relative sea-level fall (Major et al. 1998; Robertson 1998c; Whiting 1998), although the underlying mechanism behind this event and its magnitude remain uncertain.

Subunit 5D thickens locally in the NW of our study area to around 120 m, forming a mounded, internally layered seismic facies (Fig. 18). Mound location was controlled by antecedent relief over the Unit 4 inter-sag ridges. Individual mounds are up to 3 km across and all display aggradational or slightly back-stepping geometries, attributes that suggest a response to rising relative sea level. Up to four cycles of deposition are present, each 40–50 m thick, separated by high-amplitude, continuous seismic reflectors. Over most of the study area, however, the mounded seismic facies is absent and the whole of Subunit 5D consists of parallel-continuous striped seismic facies. Where contiguous with the mounds, this facies correlates with the lower mound cycles (Fig. 18).

Over an isolated, flat-topped submarine plateau, a relatively low-productivity Miocene neritic carbonate factory would generate a layered, topography-blanketing, open-shelfal sedimentary succession consistent with the seismic expression of the Miocene succession tagged in the ODP boreholes (cf. Sangree and Widmier 1979; Halfar and Mutti 2005; Merino-Tomé et al. 2012). The Miocene ‘reefal’ facies identified in ODP 160-966 is thus more likely to represent a variable mosaic of coralline-algal biostromes and rhodolith pavements (cf. Bosence and Pedley 1982; Quaranta et al. 2012), only in later stages forming local low-relief mounds over subtle antecedent topographical elevations. Challenged possibly by rising relative sea level, these would have been unable to expand and amalgamate to form a large platform (e.g. see Robertson 1998c). Messinian sea-level draw-down in the Mediterranean Basin terminated any preceding trend.

Because information has been limited, researchers have been unable to add robust detail and chronology to the stratigraphy of the Eratosthenes High, beyond the generic model that it is a Mesozoic–Cenozoic carbonate platform over a fragment of continental basement. The result is a legacy of diverse stratigraphic and structural interpretations of the same data that have sometimes been speculative and difficult to reconcile. Most discussion has centred on the ages of the principal seismic stratigraphic units: i.e. our units 3 and 4, as these are less constrained by data than Unit 5. This is an important question as these intervals probably form the principal gas reservoirs in discoveries around the Eratosthenes High. Our seismic dataset provides a contribution to this discussion but also has limitations in the deeper section as data acquisition and processing parameters were optimized for intermediate depths. Here, we outline our rational for allocating ages to seismic stratigraphic units 3, 4 and 5; these will undoubtedly be properly resolved if results from exploration wells are eventually published.

The seismic stratigraphy of the Eratosthenes High has typically been interpreted by comparison with regions surrounding the Levant Basin, principally Israel, Lebanon and Egypt, on the assumption that these shared a similar structural history. A degree of similarity is to be anticipated, particularly for the early Mesozoic when the Eratosthenes High and onshore margins were more closely positioned. However, the Eratosthenes High has been a buoyant, isolated feature probably since the Triassic, so divergence from Levant stratigraphy is also possible. The large carbonate provinces of Apulia and the Adriatic, while also broadly belonging to the southern Tethyan realm, were never closely juxtaposed to the Levant Basin and rifted away from this margin in the Permian (Stampfli 2005) and show divergence, particularly in the Cretaceous.

In both the Levant Basin and northern Egypt, the younger Mesozoic successions exhibit expansive, aggradational Jurassic carbonate platforms followed by overall back-stepping Cretaceous platforms (Gardosh et al. 2008; Hawie et al. 2013; Tassy et al. 2015; Garfunkel and Gardosh 2023), the succession being punctuated by unconformities, volcanic episodes and siliciclastic intercalations. In particular, there is a marked hiatus between the Upper Jurassic and the late Lower Cretaceous (Hirsch and Picard 1998). Underlying these formations is a thick, varied succession of Triassic synrift and marginal-marine evaporite, carbonate and clastic sediments (Gardosh et al. 2008; Korngreen and Benjamini 2010). Clastic input is, with reasonable certainty, absent from any Jurassic and Cretaceous stratigraphy on the Eratosthenes High, but the presence of extrusive and intrusive volcanics and volcaniclastic sediments cannot be excluded, given the widespread magmatic nature of Mesozoic tectonism in the Levant Basin (Lang and Mimran 1985; Garfunkel 1991; Laws and Wilson 1997; Wilson et al. 1998; Abdel-Rahman 2002; Segev 2009; Segev and Rybakov 2010).

A recent critical control point has been provided by the Zohr-1 exploration well, c. 40 km south of our study area (Fig. 1). This penetrated a carbonate succession upon a semi-detached promontory projecting southwards from the Eratosthenes High (Roberts and Peace 2007; Esestime et al. 2016; Cozzi et al. 2021). Given the proximity and tectonic context of Zohr and the rest of the high, it seems likely that the subsidence histories of the two were coordinated.

The limited information published on this borehole points to a continuous carbonate succession at least 680 m thick consisting, from the top downwards, of:

• c. 30 m of Miocene deep-water marl;

• c. 150 m of Cenomanian–Turonian restricted carbonate platform-interior facies (‘tidal flat facies association’);

• c. 200 m of Albian reefal or carbonate margin platform facies (‘reef-related facies association’); and

• c. 300 m of Aptian restricted carbonate platform-interior facies (‘tidal flat facies association’).

Sketch logs in Bertello et al. (2016), Cozzi et al. (2021) and Bromhead et al. (2022) imply several sequences and unconformities. An abrupt upward change from platform-interior to ‘reefal’ facies over a karstified surface at the Aptian–Albian boundary implies recovery of a carbonate margin following a period of subaerial exposure. The Cenomanian–Turonian platform-interior section sits disconformably upon this (Bertello et al. 2016) and records a similar event. Both of these ‘type-1’ unconformities appear to have been of sufficient duration to have promoted karstification and the section may have been overprinted by a longer karst event at the terminal upper boundary.

On thickness alone, it is tempting to equate the Zohr-1 succession with Unit 4 in the study area. A convincing comparison can also be drawn between the seismic facies succession in Unit 4 and the upper cored portion of the Zohr well. A section near the escarpment (e.g. Fig. 6), for example, would show an initial aggrading, possibly reefal or biostromal succession (Subunit 4A, c. 400 m thick), overlain by horizontally layered platform-interior (Subunit 4B, c. 350 m thick). Seismically resolvable lowstands, probably involving subaerial exposure and karstification of the platform tops, occurred between Subunit 4A and Subunit 4B, and between sequence sets within Subunit 4B. Given the continuity of stratigraphic units across the high, this suggests that Unit 4 may have an Aptian–Cenomanian or Turonian age in common with the Zohr section. The post-Unit 4 succession differs between the two locations, suggesting that Zohr may have remained elevated longer than the main high, because the Eocene is missing, possibly allowing a longer period for karstification, and subsided more rapidly after termination of the Unit 4 platform, because the shallow-water Miocene section is missing.

Albian sediments were not identified in ODP 160-967, but recovery was just 1.0% in the lower cored section (Flecker et al. 1998; Major et al. 1998), so that sample locations, ages and facies interpretations are likely to be uncertain. On image logs, the Aptian carbonates are thinly bedded and tentatively characterized as ‘lagoonal’ (Premoli Silva et al. 1998), and on gamma-ray logs exhibit a punctuated aggradational profile (Major et al. 1998). The lowest 94 m of the hole was tectonically brecciated and diagenetically altered, and therefore undated. The lower 20 m of Cenomanian pelagic or hemipelagic sediments in the ODP borehole contain intercalations of shallow-water grainstones (Major et al. 1998; Premoli Silva et al. 1998), possibly indicating the persistence laterally of a carbonate platform.

Papadimitriou et al. (2018) correlated the whole of Unit ESP4 (=Unit 4) to the ODP 160-967 borehole location. If this is correct, then the whole of Unit 4 would necessarily be of Aptian age, or older. It is difficult to reconcile this interpretation with the presence of karstified Albian and Cenomanian shallow-water carbonates in Zohr and probably in our study area, albeit some distance removed, and the widespread occurrence of shallow-water Albian–Turonian carbonates in onshore Levant Basin successions. Moreover, the hemipelagic succession overlying the Aptian in ODP 160-967 is absent everywhere from the Unit 4 platform top. An equivalent to this interval does, however, probably exist in subunits 5B and 5C in the intrashelf basin. This depression therefore provides a window into events occurring in deeper water following the demise of the shallow-water platform. In the intrashelf basin, probable pelagic sediments overlie c. 450 m of basin-floor carbonates (subunits 4A–5A) redeposited from the adjacent carbonate platform (Fig. 17). The intrashelf basin is also surrounded by platform margins and so conceivably presents a more proximal aspect than the ODP 160-967 location.

In view of the extremely poor core recovery in the ODP 160-967 borehole and a biota of limited diagnostic value (Flecker et al. 1998; Premoli Silva et al. 1998), Albian and Cenomanian strata could potentially be present here, but overlooked, and this succession might resemble that of the intrashelf basin. A more likely alternative is that the Aptian section equates to Subunit 4A in our study area. Away from the southern margins of the high, this forms a 100–150 m-thick sediment blanket that thins somewhat northwards. It cannot be distinguished on old 2D seismic lines but is present at the limit of our survey and probably covers most of the Eratosthenes High (cf. Fig. 6). Subunit 4B becomes less continuous north of our study area, suggesting that it may not entirely cover the high. Hemipelagic sediment, perhaps as drifts, may therefore have accumulated directly on top of (Aptian) Subunit 4A at the ODP 160-966 location. The apparent absence of Cretaceous pelagic sediments on top of Unit 4 suggests that either the platform was exposed during the Upper Cretaceous or that fine-grained sediment of this age was completely removed by currents; both would be difficult to demonstrate unequivocally.

The interpretation above would tie Subunit 4A clearly to the Aptian interval in Zohr-1 and would suggest an Albian–Cenomanian age for Subunit 4B. It is more difficult to allocate ages to the basinal subunits 5A–5C. Subunit 5A might represent a short-lived, possibly depth-challenged late Cenomanian or Turonian shelf that briefly revived the Unit 4 platform (Fig. 17). On seismic-facies grounds, Subunit 5B could represent chalks of Late Cretaceous or even Paleocene age. Truncation of Subunit 5B at the basin margin (Fig. 17b) could be the consequence of a minor relative sea-level fall or a change in oceanographic currents.

Time-equivalent depositional analogues for Unit 4 (and Zohr) are probably the well-documented Cenomanian–Turonian Mishrif Formation and equivalents (Burchette 1993; Droste and Van Steenwinkel 2004; Razin et al. 2010; Mahdi and Aqrawi 2014) or the lower Aptian Shuaiba Formation (Droste 2010; Maurer et al. 2010; Yose et al. 2010) in the Arabian Gulf; both formations, prolific petroleum reservoirs, exhibit a strongly progradational, intrashelf basin-filling character. Complex progradational margins of mid-Cretaceous age are present in areas of the Levant Basin, in northern Israel, Lebanon and southern Syria (e.g. Buchbinder et al. 2000; Grosheny et al. 2017). In all these examples, carbonate platforms exhibit a comparable strongly progradational, intrashelf basin-filling character and were impacted by periods of subaerial exposure.

From the above argument, Unit 3 must have, at youngest, an earliest Cretaceous or Jurassic age. The lowest Cretaceous (Berriasian–early Hauterivian) was a period of global sea-level lowstand, following a Tithonian maximum (Davies et al. 2002; Haq 2014). Deposits of this age in Israel, Lebanon and Egypt consist largely of terrestrial and shallow-marine sandstones (Fig. 2) (Gardosh et al. 2008; Hawie et al. 2013; Tassy et al. 2015; Ghalayini et al. 2018). A marked unconformity between the Jurassic and the Cretaceous in onshore areas, and the absence of large early Cretaceous carbonate platforms suggests that these may also be missing from the stratigraphy of the Eratosthenes High, a period that would be represented by subaerial exposure at the Unit 3–Unit 4 boundary. If so, then a Jurassic age for Unit 3 seems most likely. A continuous 1–2 km succession of shallow-water carbonates spanning the whole of the Jurassic is present throughout the onshore and immediate offshore Levant (e.g. see Gardosh et al. 2008, 2011; Hawie et al. 2013; Tassy et al. 2015).

The deep stratigraphy of the Eratosthenes High (units 1, 2 and older) is seismically poorly imaged. Some recent studies appear to have regarded the present inclined flanks of the high as primary palaeodips and interpreted seismic stratigraphic geometries accordingly. Papadimitriou et al. (2018), for example, suggested that Unit ESP2 (=Unit 2, this study) might represent a back-stepping Jurassic carbonate platform. Segev et al. (2018) interpreted the same unit as a domed palaeovolcanic centre. However, a contradiction arises if Unit 2 is interpreted as a positive feature of any kind, because: (a) relief would have been up to 2000 m (minimum of c. 1400 m); (b) Unit 3 does not clearly onlap Unit 2; (c) Unit 3 consists mostly of platform-interior sediments; and (d) Unit 3 exhibits no platform margin and therefore no slope facies that might have infilled significant accommodation space. The Unit 3 platform could not have developed and prograded from a shallow-water locus near the crest of a large positive feature.

The combined envelope of units 4 and 5 defines a tabular carbonate body of relatively consistent 1000 m thickness over most of the high, with top and bottom boundaries roughly parallel in both dip and strike directions (e.g. Figs 6 and 8). Regardless of any water-depth variations or exposure at the termination of Unit 4, aggradation and topographical infilling in Unit 5 continued to be controlled by the same sea-level datum under a perpetuated tectonic regime. A fundamental characteristic of shallow-water carbonate platforms throughout geological history is that a mature platform over-produces sediment compared with any accommodation space increase, develops a flat top at or close to sea level and expands through progradation (Kendall and Schlager 1981; Schlager 1981; Tucker and Wright 1990; Wright and Burchette 1996). Kendall and Schlager (1981) devised the terms ‘start-up’, ‘catch-up’ and ‘keep-up’ to denote the successive stages in this progression, and these are seen to good effect in the Miocene carbonate platform in Unit 5. The three carbonate platforms we have described were no exceptions to this rule and all were without doubt geologically flat-topped features at sea level prior to their demise.

In a structurally complex setting, the uneroded top surface of a carbonate platform therefore provides a horizontal sea-level datum against which to interpret underlying stratigraphy and palaeostructure. This important consideration has been under-utilized in the literature on the Eratosthenes High and has significant implications for reconstructing its deeper stratigraphy.

On regional 2D lines, Unit 2 is correlatable across the Eratosthenes High, thinning to the east, south and west, from an average of around 1500 m in the centre (Fig. 4) and has typically been interpreted in the literature as a Jurassic carbonate platform (cf. Roberts et al. 2010, fig. 5; Papadimitriou et al. 2018). It extends to, and is truncated at, the escarpment (Fig. 11b–e). Applying the simple back-stripping technique described above to either of the Unit 3 or Unit 4 carbonate platform tops resolves the structure of deeper formations at the time of platform demise. From our dataset and regional 2D lines, this shows not only that the whole of the Eratosthenes High was geologically flat topped during deposition of Units 3, 4 and 5 but that Unit 2 also possessed a relatively flat-topped geometry (Fig. 19a, b). The overlying Pliocene–Holocene chalk (Unit 9), by comparison, forms a topography-independent drape of fairly uniform thickness across the irregularities of the high, as anticipated for passively deposited pelagic sediments, and so was deposited after deformation of the high.

On lines flattened on top Unit 3, reflectors in Unit 2 lap onto the lower boundary and appear to be truncated at the top boundary (Fig. 19), suggesting that both surfaces represent unconformities. The high-amplitude seismic character of Unit 2 points to a strong impedance contrast and ‘faster’ lithologies compared with the overlying Jurassic carbonate platform and underlying Unit 1. A nominal P-wave velocity of c. 4300 m s⁻¹ calculated for this interval is relatively undiagnostic but, for a perhaps thin-bedded succession with mixed composition, this could indicate anhydritic, dolomitic or cherty carbonates, or conceivably even lithologies with a volcanic component.

Detailed as our seismic dataset is, it covers just a segment of the Eratosthenes High and the character and age of this enigmatic body in Unit 2 are difficult to determine from seismic data alone. However, neither its geometry nor seismic facies are relatable to a Jurassic platform in the sense portrayed in the literature, even on depth-migrated seismic lines, nor does it show much similarity to the seismic character of Jurassic and Cretaceous formations of the Levant margin. We see several possibilities for the interpretation of Unit 2:

1. Conceptually, it could represent onlap of platform-interior facies onto the trailing edge of a basinward- prograding, aggrading carbonate margin in Unit 1; the strong lower boundary of Unit 2 would then necessarily be a facies-boundary reflector and a Jurassic age might be possible.

2. The geometry of this body appears consistent with a northwestward-facing depositional system, onlapping and pinching out southeastwards, possibly deposited during post-rift thermal subsidence, or it could represent infill of a basin within the Eratosthenes High. This would speak for a lowest Jurassic or Triassic age.

3. It could be an evaporitic or dolomitic carbonate succession, possibly of Triassic age, equivalent to synrift deposits in the Levant Basin.

4. It could represent a Lower Jurassic or Triassic volcanic or volcaniclastic interval.

The gentle periclinal flexure of the Eratosthenes High with quaquaversal flanks inclined at around 3° (Fig. 3) is a geologically young phenomenon coincident with the Late Miocene arrival of the crustal block at the Cyprus subduction zone (e.g. Kempler 1998; Robertson 1998a, b; Whiting 1998; Galindo-Zaldívar et al. 2001; Kokinou and Panagiotakis 2018). The crest of the high has been elevated by nearly 3000 m relative to the escarpment lip beneath the Messinian salt (Fig. 4) from a uniform starting condition at the top of Unit 5 at around sea level, and there is no evidence for significant submergence prior to this. Faulting and downdip rotation and steepening within clinoformed intervals in Unit 4 as far as 20 km from the escarpment also suggest the possibility of some detachment at the Unit 3–Unit 4 boundary in response to post-depositional tilting. This deformation has created a swarm of WSW–ENE normal extensional faults sub-parallel to the Cyprus subduction zone, with throws of up to 400 m. These form horst-and-graben structures in the north and centre of the high (Fig. 4) that entirely post-date the carbonate platforms but have served to dissect them. Correlation is facilitated by back-stripping the displacements on the largest faults. The current submarine flat top of the high is a feature often emphasized in images of the modern seafloor (e.g. see Mascle et al. 2006; Mitchell et al. 2013; Ballard et al. 2018; Kokinou and Panagiotakis 2018) but, in our view, this simply represents the tilted legacy of the broad, plateau-like top of the superposed Cretaceous and Miocene carbonate platforms.

The erosional channels incised into Unit 5 (Figs 11 and 18) have persisted as seafloor features that continue to influence downslope movement of deep-water sediment (Fig. 20). The channels represent fluvial incisions of Messinian age, markedly enhanced by updoming of the high, and presumably correspond broadly to the ‘Messinian Erosional Surface’ recognized elsewhere in the Mediterranean region (e.g. Camselle et al. 2014; Gorini et al. 2015; Cornée et al. 2016; Ben-Moshe et al. 2020). Karst dissolution is widespread at this surface (Audra et al. 2004; Mocochain et al. 2006; Bakalowicz 2014) and it is possible that Unit 6 is also karst-related. The dolomitized, matrix-rich carbonate breccias present above the Miocene section in both ODP 160-965 and 160-966 are of Messinian age (Major et al. 1998) and conceivably represent fills to similar depressions. Palaeosols and thin limestones with a brackish-water biota are also present above the Miocene limestones in both ODP boreholes (Major et al. 1998; Robertson 1998c; Coletti et al. 2019) and potentially point to a latest Messinian or earliest Pliocene pluvial regime, perhaps reflecting the end-Messinian ‘Lago Mare’ period (see Orszag-Sperber 2006).

The downdip, offlapping terraces in Subunit 6A are inclined with the flank of the high and internal listric faults; slide scars and deformation are common (Fig. 21a), probably indicating that they were deposited early during deformation. Seismic facies and amplitude in this subunit are reminiscent of Unit 6 near the crest of the high, which suggests that it also consists of downslope-transported material. These deposits may represent alluvial fans, tufas or shorelines of dolomitized carbonate conglomerates or even marginal anhydritic evaporites. It is difficult to locate the original onlap edge of the salt as this is obscured beneath deformed allochthonous material, but it is assumed to have been at around the level of, or a short distance downdip from, interpretable strata in Unit 6; this would not necessarily indicate maximum water levels within the basin during this period, although there is no seismic evidence that the high was further onlapped or that the flanks were significantly tectonically depressed during initial deformation.

Subunit 6B continued the offlapping trend established in Subunit 6A and was deposited following deformation but at an overall lower base level. It exhibits much lower seismic amplitude, consisting of sequence sets with horizontal top surfaces, toplapped tangentially by internal clinoforms (Fig. 21a, b). It comprises sediments of different composition and probably had a fully subaqueous origin. In some areas, evaporite deformation has forced considerable structural and stratigraphic complexity on this thinly layered unit. The terraces in Subunit 6B may represent Messinian palaeoshorelines or conceivably fringing carbonate bodies or evaporite wedges that tracked base-level falls during the Stage 2 sea-level drawdown of the Messinian Salinity Crisis (cf. Meilijson et al. 2019; Heida et al. 2022). Depositional events surrounding this period were regionally diverse but comparable ‘falling-stage’ forced regressive depositional systems have been described from basins in southern Spain (Franseen and Mankiewicz 1991; Franseen et al. 1998; Warrlich et al. 2005), and convincing examples exist elsewhere in the region (Gorini et al. 2015).

Mounded features up to 200 m high, with transparent seismic facies and no reflective outline, occur sparsely along this trend and appear to originate at the base of Subunit 6A (Fig. 21b). They are associated with normal faults, suggesting a possible genetic relationship, although they are circular or elongated in plan rather than linear. Elevated strata above the mounds in the lower portion of Subunit 6B exhibit updip sediment ponding and differential compaction, indicating that they formed primary seafloor features. The impact of the relief over the mounds dies out within the first few clinothems of Subunit 6B, pinning their relative age to before deformation of the Messinian salt. While the origin of these enigmatic features remains unclear, it is possible that they represent fault-associated mud or salt diapirs. Or, if the mounds are constructional in nature, then a microbial or tufa origin seems likely in this setting, possibly promoted by hydrological head consequent on uplift of the Eratosthenes High.

As there is no evidence for deformation or uplift during deposition of Subunit 5D, updoming of the Eratosthenes High to its present morphology must have been a rapid early Messinian event, a timing that is consistent with some previous interpretations (e.g. Robertson 1998a, b). Unit 8 is contiguous with the post-salt section in the Levant Basin and represents onlap of the regional Pliocene cover onto the flank of the Eratosthenes High, conceivably during refilling of the eastern Mediterranean Basin. The onlapping Messinian salt would thus have been translated onto a structural ramp formed by the dip slope of the high, namely the top of Unit 5 (Figs 6 and 10). It seems unlikely that loading of the flanks of the Eratosthenes High by Messinian salt contributed to their depression (cf. Major et al. 1998) as deformation was mostly complete prior to salt migration. It is consequently erroneous to talk about a substantial Mesozoic, Paleogene or earlier Miocene ‘crest’ (Robertson 1998c, d; Coletti et al. 2019; Coletti and Basso 2020) for the Eratosthenes High. Moreover, it follows, for reasons discussed, that it is unlikely that a large Miocene ‘palaeo-atoll’ existed at its ‘crest’: differential relief had not developed at that time.

The significant volumes of lean biogenic gas discovered over the last 15 years in the deep-water eastern Mediterranean Sea have become a critical resource for surrounding countries, with implications too for natural gas supply to Europe. Exploration activity is being pursued in earnest throughout the region, as reflected, for example, in offshore bid rounds being held this year (2023) in Israel, Lebanon and Egypt, or in the number of high-profile exploration wells being or expected to be drilled. Potential Mesozoic-age (‘deep’) carbonate plays are of particular interest elsewhere in the Levant Basin, as potential targets are coming into reach with the application of modern drilling technology, and younger plays are generally thought to have matured. The Eratosthenes High is a massive feature that is likely to have retained unique evidence of the geological evolution of the region and therefore attracts both exploration and academic interest. Our study clarifies the nature, timing and rate of events that occurred during the Mesozoic and Cenozoic evolution of the high and may serve as structural and stratigraphic model for the interpretation of carbonate reservoirs of this age in the region.

Reservoir heterogeneity is a consideration even in the development of gas reservoirs, and our study, albeit at a seismic scale, suggests that substantial stratigraphic complexity is likely to be encountered in the Mesozoic carbonate interval, particularly in the principal Early–mid-Cretaceous reservoir units. While precise predictions of reservoir quality from seismic data are difficult, certain seismic facies can serve as broad proxies for reservoir potential. In the Zohr reservoir, for example, best reservoir quality in terms of porosity and permeability was encountered in the Albian interval (e.g. Cozzi et al. 2021), a zone of grainy rudist sediments and patch reefs sandwiched between much poorer-quality Aptian and Cenomanian platform-interior successions. While parallels of this sort with Zohr can definitely be stretched too far, platform-margin, platform-interior and basinal seismic facies can generally be distinguished in our dataset. Facies with most reservoir potential might, therefore, be represented by the prograding shelf margins of Unit 4. If this unit does indeed range in age from Albian to Cenomanian, it suggests that, depending on the location around the high, Cenomanian rather than Albian rudist facies might form the principal reservoir interval.

All intervals in Zohr appear to exhibit extensive karstification, which implies that this may also be the case in equivalents on the Eratosthenes High. Syn- and post-depositional faulting and fracturing also seem to be pervasive in the Mesozoic section, at least in the study area, and some of the major faults are deeply rooted. The combination of high matrix porosity and permeability and widespread conduit networks of this kind could point to potential future problems in reservoir development, during the drilling phase, and/or by early water breakthrough at wells or completion collapse during the production phase. Fluid overpressure in sediments beneath the Messinian evaporites is widespread in the eastern Mediterranean Basin, and in places results in breaches in the form of mud volcanoes. Such over-pressured intervals may pose a drilling risk.

Using modern depth-migrated 3D seismic data, we have shown how three superposed carbonate platforms on the Eratosthenes High, with different carbonate factories, developed under contrasting tectonic regimes, leading to a very different stratigraphic evolution of each. The oldest platform, probably of Jurassic age, has an aggradational depositional style and a tabular ‘Bahamian’ morphology and covered all of the high. The margin of this platform was truncated by faulting and collapse, resulting in a scalloped palaeo-escarpment.

The Jurassic platform was submerged, probably after a period of subaerial exposure, and a new platform, of likely Aptian age, seeded at its margin as a linear trend of aggrading shoals or reefs that rimmed much of the southern half of the high. A minor relative sea-level fall forced rapid northwestward progradation of the proximal margin of this complex for up to 50 km, probably during the Albian and Cenomanian, as clinoformed high-frequency sequence sets that infilled nearly all accommodation space. The Cretaceous platform was dissected by synsedimentary NW–SE-trending, fault-controlled sags that were infilled by carbonate progradation. Platform growth isolated an intrashelf basin that remained connected to the Levant Basin by a narrow channel. At platform termination the basin was just 10 km across.

A final carbonate succession, Unit 5, formed a thin blanket over the antecedent platform top. The intrashelf basin fill consists of probable Upper Cretaceous–Eocene hemipelagic sediments equivalent to those in ODP 160-967 in the north of the high. The uppermost portion of Unit 5 represents a shallow-water neritic shelve equivalent to the Miocene interval in updip ODP boreholes and locally contains discrete mounds, the locations of which were influenced by antecedent topography. At termination, Unit 4 and Unit 5 combined had generated a flat-topped, tabular body around 1 km thick and up to 150 km across that covered much of the high.

Deformation of the Eratosthenes High to its current periclinal morphology occurred after deposition of Unit 5, the surface of which was incised by channels along residual sag trends during the Messinian sea-level drawdown, depositing a fringe of debris or shoreline sediments around the periphery of the high. The oldest of these are inclined, while the youngest are offlapping, downstepping, flat-topped sedimentary bodies that tracked the falling Messinian base level. Post-salt Pliocene sediments subsequently onlapped and covered these downdip deposits prior to deformation by the allochthonous Messinian salt wall.

Flattening seismic profiles on the tops of any of the carbonate platforms described here back-strips the current structure of the Eratosthenes High to reveal stratigraphic geometries in deeper formations. This shows that Unit 2 was probably a ‘topography-filling’ body and not a positive structure as previously interpretations have suggested. This enigmatic unit may represent anhydritic or dolomitic carbonates, possibly of Triassic age.

NewMed LP (previously Delek Drilling LP), for which this work was carried out as part of a broader study, is thanked for permission to publish this paper. We are grateful for helpful comments on the manuscript prior to submission from Dan Bosence, and for excellent suggestions during review by Fabio Lottaroli, Mike Simmons and an anonymous reviewer; these very much helped improve the content and structure of our manuscript.

TB: conceptualization (lead), formal analysis (lead), investigation (equal), methodology (lead), software (supporting), writing – original draft (lead), writing – review & editing (lead); GG-G: conceptualization (supporting), data curation (lead), formal analysis (supporting), investigation (equal), software (lead), visualization (lead), writing – original draft (supporting), writing – review & editing (supporting); KK: conceptualization (supporting), funding acquisition (lead), validation (equal), writing – original draft (supporting), writing – review & editing (supporting).

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The seismic dataset analysed during this study is not publicly available and is proprietary to the licence partners/Cyprus government.