A terminal Messinian flooding of the Mediterranean evidenced by contouritic deposits on Sicily

The evolution of marine gateways and sea straits exerts major control on bottom current depositional systems. A well‐known interval in geological history characterized by frequent changes in marine connectivity is the Messinian Salinity Crisis (5.97 to 5.33 Ma) when the Mediterranean allegedly experienced major (>1 km) sea‐level drawdown followed by a catastrophic marine replenishment at the base of the Zanclean. Controversy exists around the timing and mode of this event as unambiguous flood deposits have so far never been drilled or recognized in outcrops. In the Sicilian Caltanissetta Basin (Italy), the Messinian/Zanclean boundary is directly underlain by the Arenazzolo Formation. This 5 to 7 m thick sandy sedimentary interval may reveal a genetic link with the abrupt refilling of the Mediterranean, but at present a detailed study to understand its origin is lacking. In this work, the Arenazzolo Formation at Eraclea Minoa has been studied by a multi‐method approach, employing detailed facies description, grain‐size analyses, petrographic analyses and palaeocurrent analyses. Palaeogeographical reconstructions and facies associations show that the Arenazzolo Formation sands were deposited on the northern flank of the Gela thrust front by persistent bottom currents, flowing parallel to the regional slope physiography, during a transgression. It is hypothesized that these currents are associated with the active circulation of surface and intermediate water masses coeval with a terminal Messinian flood, when basin margins overtopped and a reconnection between western and eastern Mediterranean was created. The Arenazzolo Formation is a unique example of a contouritic deposit formed by bottom currents that establish during the reconnection of major isolated water bodies.


INTRODUCTION
During the late Messinian, major changes in intrabasinal and extrabasinal connectivity in the Mediterranean Sea led to a major drawdown of Mediterranean base-level and highly fluctuating salinities (Selli, 1960;Hs€ u, 1972;Hs€ u et al., 1973;Krijgsman et al., 1999;Roveri et al., 2014). The so-called Messinian Salinity Crisis (MSC) was a short-lived ecological and environmental crisis, which caused the accumulation of more than a million cubic metres of salt on the Mediterranean sea floor (Ryan, 2009;Haq et al., 2020). The MSC is largely thought to have evolved in three stages (Roveri et al., 2014, Fig. 1A): (i) Primary Lower Gypsum (PLG) deposition concentrated in shallow marginal basins (Lugli et al., 2010) and possibly intermediate basins (Ochoa et al., 2015;Raad et al., 2021) during Stage 1 (5.97 to 5.59 Ma); (ii) maximum sea-level drawdown with halite precipitation in the deepest basins (5.59 to 5.55 Ma); and (iii) gypsum deposition alternating with brackish water marls in intermediate and deep basins (5.55 to 5.33 Ma). For this third phase (also termed Stage 3 or Lago-Mare) three competing palaeoenvironmental scenarios exist.
3 The Mediterranean was oscillating with high-amplitude precession-controlled base-level fluctuations (Rouchy & Caruso, 2006;Andreetto et al., 2022a). Scenarios 1 and 3 imply a significant bathymetric contrast between the Messinian and Zanclean and suggest a catastrophic end to the MSC with the collapse of the Gibraltar Sill, and the abrupt re-establishment of open marine conditions all over the Mediterranean at the beginning of the Zanclean (Hs€ u et al., 1973;Garcia-Castellanos et al., 2009;Amarathunga et al., 2022).
The concept of a major bathymetric change at the end of the Messinian is supported by the observation of erosional and depositional morphologies, recorded in seismic data-sets from the deep Mediterranean (Garcia-Castellanos et al., 2009Micallef et al., 2018;Spatola et al., 2020). Incisions and chaotic flood deposits recognized in seismic profiles are interpreted as the expression of a Zanclean reflooding of the Mediterranean, reaching discharges possibly above 100 sverdrup (Sv) (Garcia-Castellanos et al., 2009). Sandstones straddling the Messinian/Zanclean boundary at Ocean Drilling Program (ODP) sites 974 and 975 (Tyrrhenian and South Balearic Sea) are thought to be the result of suspension transport during this Zanclean Flood (Zahn et al., 1999). Still, unambiguous observations of flood deposits in outcrops on land have not been documented and linked to the Mediterranean reflooding. Micallef et al. (2018) provide photographs of Zanclean breccias deposited on the Hyblean Plateau and presumes these are flood deposits, but description and interpretation of these facies is limited.
One of the best studied records of the Messinian to Zanclean transition in the Mediterranean is that of the Sicilian Caltanissetta Basin ( Fig. 2A and B). It comprises the Messinian Upper Gypsum unit (UG), the uppermost Messinian Arenazzolo Formation and lowermost Zanclean Trubi Formation. The UG unit (corresponding to the third MSC phase) bears seven rhythmic couplets of primary gypsum beds, coarse terrigenous deposits and marls (Manzi et al., 2009;Andreetto et al., 2022a), of which the upper two are shown in Fig. 2C. The uppermost clays just above UG cycle 7 are overlain by a 5 to 7 m thick sandy to silty unit called the Arenazzolo Formation (Fig. 2C), located right below the basal Zanclean (Ogniben, 1957;Brolsma, 1978;Cita & Colombo, 1979). Sedimentological facies descriptions of the Arenazzolo Formation have never been dealt with in depth and a genetic link between these deposits and the Zanclean reflooding has never been proven. This study aims to construct a sedimentary model for the uppermost part of the UG and the Arenazzolo Formation at the Eraclea Minoa and Capo Rosello sections on Sicily and explores whether the Arenazzolo Formation could be the expression of the terminal flood that ended the MSC. Further recognition of the imprint of the Mediterranean reflooding will lead to fundamental insights into the nature of environmental upheavals, in particular large outburst floods and associated re-establishment of gateways between basins.

GEOLOGICAL SETTING
The study area is situated in the Sicilian Maghrebides, an east-west trending, arcuate, fold-andthrust belt, linking the African Maghrebides to the Calabrian arc and the Apennines (Butler et al., 1995(Butler et al., , 2015Catalano et al., 1996;Lickorish et al., 1999). This broader area is characterized by a tectonic history of continued accretion of thrust sheets and clockwise rotation of allochthons, since the latest Oligocene (Catalano et al., 1996(Catalano et al., , 2013. One of the main thrust sheets is the Gela Nappe carrying the wedge-top Caltanissetta Basin as the main depocentre between smaller wedgetop basins in the north (Belice Basin, Ciminna Basin) and the Gela Foredeep and Hyblean Foreland in the south (Fig. 2). According to palaeomagnetic analysis, the Caltanissetta Basin experienced a ca 30°post-Messinian clockwise rotation (Duermeijer & Langereis, 1998).
The Messinian sedimentary successions on Sicily are generally subdivided into various lithostratigraphic units that can be attributed to the three MSC stages described earlier (Figs 1A and 2B). According to Roveri et al. (2008bRoveri et al. ( , 2014, Stage 1 is characterized by Primary Lower Gypsum (PLG) deposition in the shallower wedge-top basins, interfingering with marine shales and limestones in the deeper areas (5.97 to 5.59 Ma), Stage 2 is the acme of the MSC and is comprised of Resedimented Lower Gypsum (RLG), halite in the basinal areas and limestones in the marginal areas (5.59 to 5.55 Ma), and Stage 3 is characterized by cyclic alternations of primary evaporites and coarse terrigenous deposits and marls. In the Caltanissetta Basin, the stratigraphic units of Stage 3 consist of the Gessi di Pasquasi Formation (Selli, 1960) or Upper Gypsum (UG) and Arenazzolo Formation (Cita & Colombo, 1979;Manzi et al., 2009). The UG successions generally consist of six couplets of marls and gypsum (cycles 1 to 6) followed by two couplets of marls and sandstones (cycles 6 0 to 6″), followed by a final couplet of marls and gypsum (cycle 7) and topped by the Arenazzolo Formation (Fig. 2C). A comprehensive facies and sequence stratigraphic model of the entire UG of the Caltanissetta Basin was established by Manzi et al. (2009) and later refined by Andreetto et al. (2022a), built on a model that envisages recurring precession-controlled depositional cycles (van der Laan et al., 2006;Hilgen et al., 2007). The stratigraphic position of the Arenazzolo Formation is interpreted inconsistently in literature (see Grothe et al., 2018, for different scenarios). In most cases, the Arenazzolo Formation is depicted as a bounding unit between the UG and the Trubi Formation. Alternatively, it figures as a lateral equivalent of the UG (see figs 2 and 13 in Manzi et al., 2021), although evidence for interfingering between the two units is not observed.
The Arenazzolo Formation is a sandy to silty, locally cross-laminated formation of a few metres thick. It is often genetically linked to a small deltaic system that is also interpreted to have formed the thin sandstone lobes of UG cycles 6 0 to 6″ (Schreiber et al., 1976;Roveri et al., 2008a;Manzi et al., 2009). Some have suggested that the Arenazzolo Formation is a distal equivalent of a fanglomerate at the foot of the Peloritan Mountains (Decima & Wezel, 1973), or that it formed in a non-marine littoral setting at the edge of a lake or delta lobe (Cita & Colombo, 1979). Other studies propose that the Arenazzolo Formation is a transgressive sequence implying a significant pre-Zanclean deepening of the Mediterranean (Brolsma, 1978;Londeix et al., 2007;Bache et al., 2012;Popescu et al., 2021). Brolsma (1978) interpreted the Arenazzolo Formation as an initial phase of Trubi deposition rather than a terminal phase of the Messinian evaporation cycles, suggesting a gradual environmental transition between the Arenazzolo Formation and the Trubi Formation. Popescu et al. (2009) andBache et al. (2012) consider the Arenazzolo Formation as the expression of an earlier (5.46 Ma) Mediterranean reflooding by Atlantic waters, following a prolonged period of nondeposition represented by the Messinian Erosional Surface (MES) at the base of the Arenazzolo Formation. Their interpretation requires 6.5 precessional cycles and implies that the UG is a lateral equivalent of the fully marine MSC Stage 1. This diverse range of interpretations illustrates the need for a more focused sedimentological study of the Arenazzolo Formation to elucidate its depositional environment.

Fieldwork
A total of five sections were analysed ( Fig. 3A to C). Sections ERA, ERB and ERC cover the interval between the top of UG cycle 7 and the Trubi Formation, and are exposed along the strip of beach at Eraclea Minoa (37°23 0 24.18″N, 13°16 0 36.09″E, Fig. 3B). Section ER6 is located directly north of Eraclea Minoa and comprises UG cycle 6″. Section CR covers part of the Arenazzolo Formation and the boundary with the Trubi Formation at Capo Rosello (37°17 0 35.40″N, 13°28 0 10.88″E, Fig. 3A), which is also located along a strip of beach 20 km to the south-east of Eraclea Minoa. Section CR has been studied extensively before (Brolsma, 1978;Cita & Colombo, 1979), but the main part of the Arenazzolo Formation is now covered by a recent landslide. All sections were logged at a 1 cm resolution. Several distinct facies and facies associations were recognized and categorized in all five sedimentary logs (summarized in Tables 1 and 2).

Grain-size analysis
Grain-size analysis was performed on 77 samples in log ERA (11 samples were taken in the mudstones and sandstones below the Arenazzolo Formation, 64 samples in the Arenazzolo Formation). The aim of this analysis was to: (i) test the facies distinctions made in the field and improve characterization of depositional processes of these facies; and (ii) to generate a record of variation in grain-size distribution throughout the Arenazzolo Formation. The measurements were performed by a Mastersizer 2000 laser particle sizer (Malvern Panalytical, Malvern, UK; for complete methods see Supplementary material S1). Grain size was recorded between 0.2 lm and 2000 lm, subdivided into 100 class intervals. Both sample and subsample duplication were performed and no significant difference in results was found. Grain-size distribution curves and cumulative distribution curves were plotted using Gradistat (Blott & Pye, 2001). Derivation of grain-size parameters was done based on Krumbein & Pettijohn (1938). Calculation of statistical parameters and their physical description is based on the Folk & Ward (1957) graphical method.

Petrographic analysis
A total of 22 thin sections were prepared, covering most of the representative facies.
Petrographic analysis was aimed at making qualitative observations of composition, texture and sedimentary structures. Thin sections were scanned using a Leica M165C stereomicroscope (Leica, Wetzlar, Germany). All thin sections were studied in both plane polarized and crosspolarized light using a Leica DM750 optical microscope, and several representative photographs were made, using a Bertrand lens and camera.

Palaeocurrent analysis
Palaeocurrent analysis was performed on trough cross-laminated sandstones of the Arenazzolo Formation at sections ERB and ERA, using an approach based on theory outlined in Slingerland & Williams (1979) and DeCelles et al. (1983). High resolution photographs of several two-dimensional exposures of trough crosslamination were made in the field. The orientation (strike-dip) of the exposures was measured with a geological compass. All laminae sets were interpreted and assigned an estimated trough axis orientation representing the palaeocurrent direction. The procedure of estimating the orientation of the trough axes is explained in more detail in Supplementary material S2.

Description of sections and outcrops
Log ER6 covers ca 7 m of UG cycle 6″ and consists of interbedded sandstones, siltstones and mudstones (Fig. 4). Log ERA covers ca 3 m of UG cycle 7, followed by ca 7 m of the Arenazzolo Formation. On top of the Arenazzolo Formation a ca 50 cm thick clay interval is present and in turn overlain by the Zanclean Trubi Formation (Fig. 5). Log ERB is located 269.5 m to the west of ERA and bears a similar succession, with comparable thicknesses of lithostratigraphic units. ERC is located 142.5 m to the west of ERB and bears the same stratigraphic interval as ERA and ERB but with reduced thickness and without a mud interval right below the Trubi Formation.
Between ERB and ERC, an asymmetrical fold pair developed in the interbedded muds and selenitic gypsums of UG cycle 7 (Fig. 5A). This fold has a fold hinge with a trend of 314.8 and a plunge of 10 S. The axial plane is 160.1/22.3 W (Fig. 5B). Log CR only covers the uppermost part of the Arenazzolo Formation, followed by the characteristic top clay layer and the Zanclean Trubi Formation ( Fig. 6).

Facies
Eleven facies (F1 to F11, see Figs 7 and 8 for field photographs and Figs 9 and 10 for thin section images) were recognized in the studied sections, and summarized in Table 1. Facies F1 consists of light beige unconsolidated lithic to bioclastic arenite occurring in single beds Relatively proximal, prograding fluvio-deltaic depositional system ranging in thickness between 10 cm and 100 cm ( Fig. 7A to C). Beds have normal grading and sharp erosive bases and often decimetre-scale to metre-scale low-angle cross-stratification. Petrographically, F1 is dominated by moderately rounded quartz, lithic clasts (including carbonate extraclasts and glauconite), mixed with a heterogeneous assemblage of bioclasts (bryozoans, echinoids and foraminifera) that are relatively large (some are up to 0.16 cm) and fragmented ( Fig. 9A to D). Larger bioclasts are concentrated at the bases of beds. F1 is matrix-supported and clear variation in composition and grain size across laminae is rare. In terms of grain-size distribution, F1 is bimodal, very poorly sorted and with a very fine to fine skew (Fig. 7M). F1 has a platykurtic peak between 1 lm and 7 lm (clay to very fine silt) and a major leptokurtic peak between 80 lm and 200 lm (fine sand). Facies F2 consists of massive orange to greyish bioturbated siltstones, with bed thickness ranging between 10 cm and 40 cm (Fig. 7D). Facies F3 consists of dark greyish mudstone which commonly varies in degree of darkness and intensity of bioturbation (Fig. 7E). Some intervals have mottled, bioturbated portions and are relatively light grey in colour. Other intervals are darker and show no evidence of bioturbation. F3 often bears subtle lamination due to the alternation of relatively muddy intervals and siltier intervals. In some places, broken shell fragments are found floating within the mud, occasionally concentrated in single laminae that have a sharp lower contact and normal grading. In terms of grain size, F3 has a mesokurtic, unimodal, poorly sorted grain-size distribution with a symmetrical skew, and a mean grain size ranging between 3.7 lm and 9.0 lm (very fine silt; Fig. 7N).
Facies F4 consists of lithic arenite to gypsarenite. The framework of this facies is composed predominantly of quartz grains and reworked immature gypsum clasts. Single beds are normally graded and 1 to 10 cm in thickness. They have a sharp undulatory base, often conformable with underlying strata and characterized in some places by flute marks (Fig. 7F to H). Locally,  cross-lamination and wavy lamination is observed, and the bases of single beds often have a brownish to orange stain. In some places, layering is offset by small reverse faulting (Fig. 7G).
Facies F5a is a gypsarenite or gypsirudite. Space between grains is filled by matrix and a whitish cement and clasts are exclusively gypsum in the form of predominantly sand-sized grains. Beds are typically 1 to 2 cm in thickness (Fig. 7I). F5b consists of gypsarenites similar to F5a, but thinner (typically 1 to 3 mm), and interbedded with mudstone. F5b is usually 5 to 10 cm thick (Fig. 7J). Gypsarenites or gypsirudites of F5a and F5b have a framework that is entirely dominated by grains with characteristic swallow-tail crystal habit, evident from thin section images ( Fig. 9E and F). In thin section, this facies shows a clear fining-upward trend from grains that are roughly 0.15 cm in diameter to grains that are 0.03 cm in diameter. Texture can be both clast-supported as well as matrix-supported with a matrix that is a beige brownish very fine silt or mud.
Facies F6 is a rudstone (consisting solely of bivalve shells) with a muddy matrix (Fig. 7K). Shells are dominantly intact, and sometimes valves are still connected. In between dominantly intact shells smaller broken fragments occur. Single beds of F6 are between 3 cm and 4 cm thick and usually underlie mudstone intervals that contain minor bivalve shell debris. This facies corresponds to the 'lumachella' horizons, rich in Dreissena sp. and Melanopsis sp. (Decima & Wezel, 1973;Manzi et al., 2009).
Facies F7 is composed of thin (0.3 to 1.0 cm) layers of detrital elongated diatomaceous aggregates distributed and closely packed in a mud matrix, with both a sharp upper and lower contact (Fig. 7L). These layers have been termed speckled beds by Chang & Grimm (1999). F7 typically interbeds in regular intervals with relatively dark mudstones.
Facies F8 consists of a very fine planar laminated sand, organized in centimetre-thick laminae, grading in colour from beige to grey ( Fig. 8A and B). Single beds are between 5 cm and 30 cm in thickness. Colour variations within beds reflect subtle variation in grain size. Although bedding planes are not exposed properly, some fallen blocks in the vicinity of the outcrop show parting lineation (Fig. 8C). Spacing between individual ridges of this lineation is in the order of 1 to 2 cm. The grain-size distribution of F8 is generally bimodal, with very fine to fine skew and a platykurtic to mesokurtic first peak and a second leptokurtic peak. The first peak is at 10 lm and the second peak is between 40 lm and 70 lm  8F). Petrographically, F8 has a relatively uniform composition ( Fig. 10A and B). The siliciclastic fraction contains predominantly monocrystalline quartz, black lithic grains (mostly carbonate mudstone intraclasts), and elongated phyllosilicate grains. Besides a siliciclastic fraction, there is a bioclastic fraction that is dominantly composed of small (around 0.25 mm) foraminifera, usually relatively intact, in contrast to the more diverse and coarser bioclast assemblage of sandstones of F1. On the scale of polished thin sections, F8 is clearly laminated and shows compositional and textural differentiation across laminae. Texture varies across laminae between clast-supported, quartzdominated towards more matrix-supported, composed of phyllosilicates, foraminifera, and a very fine matrix. Both planar lamination as well as low angle downlapping cross-lamination are observed in thin section. Where cross-lamination is observed, the topsets of laminae are wellpreserved, and bottom sets show curved downlapping onto underlying erosional surfaces ( Fig. 10A and B). In some places, laminae have undergone soft sediment deformation, and show flame structures that are oriented in the same direction as the dip of laminae foresets (Fig. 10A). Intervals with soft sediment deformation are truncated at the top by erosional surfaces.
Facies F9 is a beige to grey very fine trough cross-laminated sand (Fig. 8D). Single beds are between 20 cm and 60 cm. Wavelength of single troughs is between 8 cm and 22 cm, and amplitude of troughs is typically between 2 cm and 4 cm. Variation in trough dimensions occurs throughout the interval and troughs are usually higher in wavelength and lower in amplitude close to intervals of planar laminated sandstones of F8 (Fig. 11). Laminae are mostly angular but become slightly tangential or sigmoidal when the face of exposure is parallel to the trough axis. Grain-size variation is reflected in colour variations across laminae sets. In terms of grainsize distribution and petrography, this facies is similar to F8 (Fig. 10C).
Intervals composed of F8 and F9 show a distinctive vertical grain-size trend throughout sections, coincident with variations in bedform dimensions. For the interval between 300 cm and 600 cm stratigraphic height in ERA, this variation is plotted in Fig. 11A. Generally, planar laminated sandstones (F8) are characterized by coarser grain size than trough crosslaminated sandstones (F9). Between 380 cm and 470 cm, a clear coarsening-upward sequence is observed, where grain size increases gradually from roughly 10 lm to roughly 35 lm. This coarsening sequence is coeval with a gradual change from high amplitude, low wavelength troughs, to low amplitude, high wavelength troughs, and finally to planar laminae (as illustrated in the interpreted photograph, Fig. 11B).
Facies F10 is a light beige very fine sand with massive bedding (Fig. 8E). F10 shows minor bioturbation in some intervals. Bed thickness is typically between 10 cm and 50 cm. Grain-size distribution of F10 is unimodal, poorly sorted and has a symmetrical to fine skew. Mean grain size ranges between 12 lm and 18 lm (medium to coarse silt). Petrographically, F10 is different and stands out from the other Arenazzolo Formation facies because it lacks evident traction structures (Fig. 10D).
Facies F11 is a brownish to beige very fine sandstone with decimetre-scale convolute lamination that is truncated by overlying massively bedded sandstones of F10 (Fig. 8E). Bed thickness varies between 5 cm and 60 cm. Colour variation between blue and beige/brown across laminae and different intervals is similar to the colour variation observed in F8 and F9.

Facies interpretations
F1 was formed by rapid deposition from gravitydriven density currents, evident from their structureless and poorly sorted texture on micro-scale, and normal grading and sharp base on bed scale. Compositional heterogeneity indicates a relatively proximal sediment source. The absence of erosional structures within single beds is evidence that beds resulted from single phases of deposition. Variations in thickness and grain size among F1 beds probably reflect different degrees of distality with respect to the source area. F2 was also formed by gravitydriven density currents decelerating and depositing silt, probably as a (distal) equivalent of F1. F3 was formed by background settling and periodic burrowing in a relatively quiet environment. The dark colour of this facies is indicative of the periodic development of oxygen-poor conditions. F4 was formed by density currents with a relatively large contribution of immature gypsum fragments, indicating the proximity of a gypsum source. The undulatory base, wavy lamination and small-scale reverse faulting of many of these layers relates to syn-sedimentary or post-sedimentary deformation after rapid deposition on top of water saturated muds. Facies 5a and 5b originated from erosion of a nearby gypsum source, and subsequent transport and deposition, without any mixing with grains from other sediment sources prior to deposition. Shell beds of F6 were deposited by the accumulation of bivalve shells and pelagic settling in a well-ventilated water column, which was occasionally colonized by bivalves and other burrowing organisms, in contrast to the more restricted and stagnant water column in which F3 muds were deposited. Deposition of F7 results from the upslope entrainment and downslope transport of diatomite aggregates (Chang & Grimm, 1999).
F8 and F9 were deposited by a fluctuating but overall persistent bottom current, evident from several observations. First of all, F8 and F9 are the result of continued accretion of sand of rather uniform grain size and composition, combined with the continued influence of traction and suspension (expressed in trough crosslamination and planar lamination). Trough cross-laminated sandstones (F9) represent the relatively weak currents. Planar laminated sandstones (F8), with a slightly higher mean grain size and parting lineation on bedding planes represent the stronger currents, probably in an upper stage plane bed flow regime (hence the parting lineation). The common variation in size and geometry of laminae and cross-sets of the trough cross-laminated sandstones suggest that also within single facies intervals, current speeds were fluctuating. Second, coarseningupward grain-size trends of several tens of centimetres, coeval with changes in bedforms, are a strong indicator of a gradually increasing current strength (Fig. 11). Third, thin section images of F8 and F9 show compositional and textural difference across laminae that reflect the repeated change between bedload deposition and deposition out of suspension. Local reactivation surfaces illustrate short episodes of erosion, followed by resumption of deposition. The presence of millimetre-scale mud-clasts within a matrix of very fine sand, suggests that currents had erosional capability to erode nearby muddy or silty seabed. In addition, the preservation of topsets in cross-laminated intervals and the presence of soft sediment deformation indicates high sedimentation rates. Besides evidence for fluctuating current action, deposition of F8 and F9 was also relatively distal. The lower variety of lithic clasts observed in thin section, the fine grain sizes and the absence of coarse shallow marine benthic bioclasts suggests that deposition of the Arenazzolo Formation (F8 and F9) was more distal than sandstones of the UG (F1). In addition, the sudden appearance of elongated phyllosilicates implies the influence of an additional source. The fine platykurtic to mesokurtic peak around 10 lm for F8 and F9 could be caused by an additional (metamorphic) source that delivered these phyllosilicates. Finally, the relative absence of the finest grain-size fraction in F8 and F9 could be the result of stronger winnowing action during deposition, which seems plausible in a scenario with strong persistent currents.
F10 formed during rapid instantaneous deposition of sand (while the formation of traction structures was suppressed), in contrast to F8 and F9 that resulted from alternating phases of erosion and deposition creating centimetrescale traction structures. F11 is characterized by decimetre-scale convolute laminations which typically form right after or during deposition when a sediment bed is in a temporarily weakened state (Kuenen, 1953;Middleton, 2003;Bjørlykke, 2015). Triggers for such lamination to form can be various: water waves, unusually rapid sedimentation, or seismic triggers are the most likely ones (Owen & Moretti, 2011).

Facies Associations
Sedimentary facies are linked to form five distinct Facies Associations (FA) based on regular recurrence of vertically related facies. Facies Association 1 (UG) occurs in section ER6
Facies Association 3 (Arenazzolo Formation) is composed of alternating planar laminated (F8) and trough cross-laminated (F9) sandstones. Boundaries between facies are generally sharp, although in some intervals more gradual transitions occur between high amplitude trough cross-laminated sandstones and planar laminated sandstones. FA3 is followed by a sequence of massively bedded (F10) and convolute laminated sandstones (F11) comprising FA4 (Fig. 4).
Facies Association 4 is 75 cm thick in both ERA and ERB but thins outs and disappears Facies Association 5 is an interval dominated by mudstones (F3) that change in colour from light grey to dark grey upward in the stratigraphy (Fig. 6). Locally, wood fragments and siltier intervals are found within the mud. The transition between FA5 and the whitish Trubi marls coincides with the Messinian/Zanclean boundary and is characterized by abundant trace fossils. Centimetre-scale burrows penetrate the uppermost clays of FA5 and are filled by whitish marls of the Trubi Formation.

Palaeocurrent directions
The approximate directions of trough axes of interpreted laminae sets in FA3 are presented in Figs 12 and 13, based on two exposures of the lowermost Arenazzolo Formation (see Fig. 5 for the stratigraphic position, and Appendix S1 and S2 for detailed measurement technique). The results of exposure ERA (with a strike orientation of 02°) and exposure ERB (with a strike orientation of 80°) were combined. This yielded a palaeocurrent direction that is east to southeast (Fig. 13), without correction for tectonic rotation.

Reconstruction of the palaeoslope
Using the geometry of strata confining the deformation of UG cycle 7, the palaeoslope of the study area can be reconstructed. Deformation of UG cycle 7 is typical for subaqueous gravitational sliding of a layered sequence on an unstable slope, in which sliding is directed perpendicular to the strike of the slope (Strachan & Alsop, 2006). The mechanical stratigraphy of selenite gypsum interbedded in softer marls facilitated the short wavelength folding in the gypsum (see Butler et al., 2015). Field measurements indicate that the fold hinge is directed roughly north-west (strike of 314.8, see Fig. 5). Correcting for the ca 30°clockwise post-Messinian rotation of the Caltanissetta Basin (Duermeijer & Langereis, 1998), this means that the Messinian palaeoslope at Eraclea Minoa was dipping towards the north/north-east (Fig. 14A), which is in accordance with earlier structural and palaeogeographical reconstructions of the Messinian Caltanissetta Basin (Lickorish et al., 1999;Manzi et al., 2021). It also implies that the dominant palaeocurrent direction (see Figs 13 and 14) was directed roughly along the regional slope (i.e. towards the east/south-east).

Sedimentological interpretation
Facies Association 1 records the progradation of a fluvio-deltaic system, characterized by a transition from distal prodeltaic mudstones (F3) and siltstones (F2) deposited by flood-related density currents to deltaic channel sandstones or interdistributary bay sandstones (F1), which were deposited down-slope in a more proximal deltafront environment (Fig. 15A). This interpretation is consistent with earlier work on transport directions of resedimented gypsum in the Caltanissetta Basin (Manzi et al., 2009) and also fits general facies models of fluvio-deltaic systems in tectonically active foreland basin systems (Mutti et al., 2000;Bhattacharya, 2006). Sandstones (F1) in this facies association generally hold a wide assemblage of redeposited shallow heterogeneous marine bioclasts and immature lithic clasts (Fig. 9), indicative for a deltaic system entering a shallow marine/lacustrine environment, eroding and redepositing sediment fed by a nearby source. FA2 was deposited in a similar fluvio-deltaic system as FA1 but with a different source area. The frequent occurrence of redeposited selenites (F4, F5a and F5b) occurring in FA2 and not in FA1, suggests that erosion and redeposition occurred while primary selenites were not significantly buried yet, and could thus be reworked.

Palaeogeographical reconstruction
Fluvio-deltaic deposition of FA1 and FA2 (Fig. 15A) took place in the main Caltanissetta wedge-top basin (Roveri et al., 2008b). Variability in distality of facies was controlled by precessional climatic changes and sediment supply (Manzi et al., 2009;Andreetto et al., 2022a). The degree to which the Caltanissetta Basin was connected to other basins is uncertain. Some studies suggest that at least a partial connection between western and eastern Mediterranean subbasins must have prevailed during UG deposition, highstands (Andreetto et al., 2021). The main argument for this is that high water-level conditions are necessary to explain the presence of Paratethyan ostracods in marginal Mediterranean subbasins (Gliozzi et al., 2007;Stoica et al., 2016) and their 87 Sr/ 86 Sr signature, which reflects the mixing between local river water and an anomalohaline water mass filling the open Mediterranean Basin (Andreetto et al., 2021(Andreetto et al., , 2022a(Andreetto et al., , 2022b. In addition, many authors interpret a general transgressive trend throughout the upper part of the UG which could be a precondition for Atlantic floods to periodically cause incursions (Butler et al., 1995). Still, water level must have been able to get low enough to disconnect the basin and oversaturate the brine to such a degree that gypsum could precipitate, favouring a scenario where the water level was, at least periodically, relatively low in the Caltanissetta Basin (as illustrated in Fig. 15A).
The Caltanissetta Basin follows a general proximal to distal trend from north to south, reflected in fractionation of grain size, and distality of facies (Maniscalco et al., 2019), as well as thickness variation of the UG throughout the basin (Manzi et al., 2009). In the current study's analyses, the sandstones of FA1 and FA2, are considered as distal equivalents of coarser UG facies in the north. The thrust nappes bordering the basin in the north are a logical sediment source for these sandstones. In addition, the Gela thrust front, located south from Eraclea Minoa may have provided an additional sediment source, subaerial or submarine (Manzi et al., 2021). This is also supported by the observation that the regional slope was dipping north-eastward during later deposition of the Arenazzolo Formation (see Figs 13 and 14). Reconstruction of the regional slope. (A) Shows the envisioned deformation of Upper Gypsum cycle 7 and the interpreted relationship with the palaeoslope and measured palaeocurrents. The fold hinge has a strike of 284°(after correction for 30°clockwise rotation) which means that the palaeoslope was dipping to the north to north-east. Note that rose diagrams indicate along slope transport direction. (B) Field photograph of the folded Upper Gypsum, used for the palaeoslope reconstruction. Fig. 15. Schematic palaeogeographical scenarios for deposition of the uppermost Upper Gypsum (UG) sandstones and the Arenazzolo Formation. Palaeogeography is based on tectonic information (Lickorish et al., 1999;Roveri et al., 2008a,b;Catalano et al., 2013;Manzi et al., 2021). (A) Scenario for Upper Gypsum deposition with disconnected sub-basins and proximally sourced sediment from the north and possibly from the Gela thrust front. (B) Arenazzolo Formation scenario after transgression and reconnected western and eastern Mediterranean, with along-slope currents forming a sandy contouritic drift. RLG, Resedimented Lower Gypsum; UG, Upper Gypsum.

Transgression and establishment of a bottom current (FA3, FA4 and FA5)
Sedimentological interpretation Onset of the contour parallel current of the Arenazzolo Formation (FA3). FA3 represents a clear change in depositional style, compared to underlying UG deposits of FA1 and FA2. FA3 deposition was submarine, given the abundant marine microfauna in the sediment (Londeix et al., 2007). Whereas deposition of FA1 and FA2 was governed by episodic density currents and background settling, deposition of FA3 was governed by strongly fluctuating but persistent along-slope currents. The Arenazzolo Formation is a relatively shallow contouritic deposit that was formed along the local physiography of the Gela thrust front (Fig. 15B).
Key results of the analysis of the Arenazzolo Formation presented here are in line with established facies models of contourites. First of all, traction structures like cross-lamination and planar lamination (as observed in FA3) have been recognized in many contourites (Shanmugam et al., 1993;Mutti & Carminati, 2012), and ripples, lineations and other bedforms have been recognized in modern-day bottom current systems (Stow et al., 2013;Rebesco et al., 2014;Esentia et al., 2018). Continuous accumulation of sand, uninterrupted by phases of sediment starvation, and the relatively hostile late Messinian environments likely prevented bioturbation (which is often used to interpret contourites) from destroying traction structures and explains why they are well-preserved. Second, palaeocurrent reconstruction and slope orientation (Fig. 14) suggest a contour-parallel current direction, which is a crucial diagnostic feature in many contourite studies. Also, the lateral correlation of individual cross-laminated and planar laminated facies over ca 400 m (see FA3 in Fig. 5) suggests that these facies are likely part of a larger sandy contourite drift system. In addition, the shift from relatively fine (ca 10 lm) to coarser (ca 35 lm) grain size, coeval to the change from trough cross-lamination to planar lamination (as documented in ERA, Fig. 11) fits with the interpretation of a contourite (Gonthier et al., 1984). In theory, the contourite model includes a shift from weak to strong currents and a shift back to weak currents, reflected in a bi-gradational sedimentary sequence. In the case of the Arenazzolo Formation, at least, this shift from weak to strong currents is present (Fig. 11).
Slope instability during deposition of the Arenazzolo Formation (FA4). Massively bedded sandstones and decimetre-scale convolute laminated sandstones of FA4 overlie FA3 and are syndepositional or slightly post-depositional to the deformation of UG cycle 7, evidenced by the western pinch-out of FA4 (see Figs 5 and 14). The formation of FA4 was probably the result of a combination of several factors, illustrated in three schematic drawings in Fig. 16. Following deposition of FA3 (Fig. 16A), the onset of sedimentation of FA4 was caused by gravitational sliding of UG cycle 7 ( Fig. 16B and C). Steep local slopes and newly created topography caused rapid sedimentation into the topographic lows, sourced from nearby FA3 sediments on topographic highs. Due to the pace of sedimentation and the slope instability, the sediment was in a weakened state and under the influence of continued increasing overburden pressure; some of these weak sediment beds formed convolute lamination. Both the initial gravitational sliding of gypsum on the regional slope as well as the deformation of weakened sediment beds were the result of increased overburden pressure due to a rise in sea level.
The transition to the marine Trubi Formation (FA5). The change to mud-dominated FA5 indicates the demise of the sediment source that fed the Arenazzolo Formation. The abrupt boundary with the Trubi Formation, characterized by pervasive burrowing, indicates a further transgression, strong change in sediment source and the rapid invasion of burrowing organisms. This specific transition likely resulted from the sudden change to deep (open) marine environments caused by concurring palaeoceanographic changes at Gibraltar that installed the modern two-way exchange pattern of Mediterranean-Atlantic connectivity at the base of the Zanclean. Any evidence for a major erosional unconformity between the Arenazzolo Formation and Trubi Formation as was previously postulated by Clauzon et al. (1996) was not observed. Moreover, an erosional unconformity at the base of the Arenazzolo Formation that Clauzon et al. (2005) correlate to the MES also was not observed.
Current results are also difficult to reconcile with the interpretation of Bache et al. (2012) that the Arenazzolo Formation comprises 6.5 precession cycles reflected by dark-light alternations. In the first place, such dark-light alternations have not been observed in the field. Second, there is no independent calibration for such a cyclostratigraphic interpretation based on other dating techniques. Third, and most importantly, present observations imply recurring phases of deposition and erosion (FA3) interrupted by sudden gypsum deformation affecting sedimentation (FA4), which is in conflict with the stable and quiet depositional conditions required to record precession cycles.

Palaeogeographical reconstruction of the Arenazzolo Formation
All observations indicate that the Arenazzolo Formation formed in a transgressive submarine environment exposed to a strong contour parallel current. A rise in sea level potentially connected previously disconnected fold and thrust belt basins generating a strong current flowing east to south-eastward, along the slope. Two options for the origin of such a current are hypothesized in Fig. 15B. Firstly, a current could have originated in the west (Adventure Bank), overtopping a local high formed by the Gela thrust front, directed towards the southeast and enhanced by the local physiography formed by the curved Gela thrust front. Alternatively, the current might have been generated by the establishment of a connection between the formerly disconnected Belice Basin (wedge-top) and Caltanissetta Basin creating a strong bottom current, flowing through a Belice -Caltanissetta gateway, parallel to the Gela thrust front. In both scenarios, previously nearby deltas that delivered UG sandstones were drowned due to the increase in base level.

THE 'TERMINAL MESSINIAN FLOOD' OF THE MEDITERRANEAN
A genetic link between the Arenazzolo Formation and Mediterranean flooding?
The theory of a Zanclean Flood generally implies a significant drawdown of the Mediterranean (1750 to 1900 m lower compared to modern-day levels), with a relatively abrupt end of the crisis generating a Mediterranean-wide flood of unprecedented discharge (Garcia-Castellanos et al., 2009;Micallef et al., 2018;Spatola et al., 2020;Amarathunga et al., 2022). In this scenario, the Sicily Sill is the easternmost barrier of the terminal flood coming in from the Atlantic. After overcoming the barrier of the Gibraltar strait, the western Mediterranean refills until it reaches the level of the Sicily Sill (see Figs 15B and 17). Next, it overtops this sill and establishes a strong eastward current that links the western and eastern Mediterranean, terminating on the eastern border of the Malta Escarpment (Fig. 17). It is exactly at this escarpment that a thick chaotic depositional body is present in the seismic profiles, which was earlier linked to the flooding event .
The Sicilian Caltanissetta Basin and the Gela Foredeep are the best candidates for gateways connecting the eastern and western Mediterranean. The present-day structural low around the Catania Plain only experienced significant subsidence during the late Pliocene , and does not bear any evidence for being a depocentre at the end of the Messinian. In addition, the main activity of the Sicily Rift Zone (Fig. 17), south of the Messinian foredeep basins also post-dates the Messinian (Civile et al., 2010). Thus, the present hypothesis of a current flowing contour-parallel to the basin physiography of the Sicilian fold-and-thrust belt and continuing its way south of the Hyblean Plateau (where Messinian evaporites are relatively thick; Micallef et al., 2018), eventually terminating in the Noto Canyon, seems plausible.

Age and duration of the flooding event
The preferred scenario presented here implies that just prior to Atlantic-Mediterranean reflooding, the western and eastern Mediterranean were disconnected Amarathunga et al., 2022;Andreetto et al., 2022a). One remaining question concerns the timing and duration of the actual flooding event. Two age constraints are available: The top of UG cycle 7 which has an astronomically tuned age of ca 5.345 Ma (Andreetto et al., 2022a) and the base of the Trubi Formation which is dated at 5.333 Ma (Hilgen et al., 2007), leaving ca 12 kyr between the top of UG cycle 7 and the Trubi Formation. Model calculations of the Zanclean Flood by Garcia-Castellanos et al. (2009) suggest a long first period of very little incision at Gibraltar (possibly several thousands of years) and an exponential increase in incision rate and water flow thereafter. Before the western Mediterranean water level reached the Sicily Sill, flow rates slowed down again due to the reduction in hydrological gradient. It is somewhere between this slowdown of the infill of the western Mediterranean and the full reestablishment of marine conditions of the entire Mediterranean that the Arenazzolo Formation was deposited. Following the model calculations of Garcia-Castellanos et al. (2009), a time window for Arenazzolo Formation deposition of a few hundreds of days is assumed (see fig. 3 in Garcia-Castellanos et al., 2009). Given the uninterrupted accretion of sand throughout the Arenazzolo Formation, and the evidence for high sedimentation rate and high flow rate this seems plausible. It leaves ca 12 kyr for deposition of the clays between the top UG cycle 7 and the base Arenazzolo Formation, meaning an approximate sedimentation rate of a few tens of centimetres per thousand years, which is reasonable for a mud-dominated prodelta environment, affected by erosion.
Some studies have concluded that there is a need for multiple Mediterranean-wide events of oscillating high-amplitude base level change during the latest Messinian (Rouchy & Caruso, 2006;Andreetto et al., 2022a). This is mainly based on the similarity of the UG on Sicily and Cyprus in the eastern Mediterranean, which also comprise six to seven gypsum-marl couplets (Rouchy et al., 2001). If the scenario of multiple events of connecting and disconnecting the western and eastern Mediterranean basins is correct, an open question remains as to how earlier incursions of western Mediterranean waters into the eastern basin had not triggered a similar depositional unit as the Arenazzolo Formation in the Caltanissetta Basin.
The official base of the Zanclean, the Global Stratotype Section and Point (GSSP) of the Zanclean Stage, that is the Messinian/Zanclean boundary, has been defined at the base of the Trubi Formation at the Eraclea Minoa section (Van Couvering et al., 2000). Following this definition and following this study's palaeoenvironmental interpretation of the Arenazzolo Formation, the Mediterranean flooding did exclusively take place in the latest Messinian and the entire Mediterranean was filled at the base of the Zanclean. In that respect, the authors propose to term the Mediterranean flooding 'Terminal Messinian Flood' to better reflect the temporal evolution of this remarkable event.

CONCLUSIONS
Present work offers new insights into the nature of the flood ending the Messinian Salinity Crisis. Prior to the flooding, sandstones in the Sicilian Upper Gypsum unit (Facies Association 1 and Facies Association 2) were deposited between the delta front and the distal prodelta of a local delta sourced from the basin margin. These Upper Gypsum deposits contrast with the overlying Arenazzolo Formation, which is notably different in terms of lithofacies, microfacies, grainsize distribution and overall composition. Observations suggest that the Arenazzolo Formation is the result of a persistent current, fluctuating in strength and directed east to south-eastward, parallel to the regional slope, outside the reach of the local deltas that fed the Upper Gypsum sandstones. This current was interrupted by the sudden development of gravitational sliding of Upper Gypsum cycle 7, creating a topography that focussed rapid deposition into a low. Deformation of gypsum and sand during this final phase of the Arenazzolo Formation deposition resulted from an increased overburden pressure caused by a sudden rise in sea level.
This study proposes a palaeogeographical model that combines the depositional nature of the Arenazzolo Formation, the expected physiography of the Sicilian fold-and-thrust belt during the late Miocene and reported seismic evidence of the impact of a Mediterranean reflooding on the Sicily area. In this model, the Arenazzolo Formation was part of a contouritic drift in between an overtopping western Mediterranean and a rapidly refilling eastern Mediterranean. For the first time, potential onshore evidence is provided for a terminal Messinian flood. This can be used as ground-truth data to calibrate current models of its impact and hydrodynamics. Furthermore, it can serve to improve understanding of contourite deposition in gateways that reconnect formerly isolated basins.

Supporting Information
Additional information may be found in the online version of this article: Supplementary material S1. Excel sheet 1 contains grain-size data used for Figs 7, 8, and 11.