Introduction

Today, more freshwater is extracted from the Mediterranean Sea by evaporation than is added by continental runoff1. Blocking the net import of Atlantic Ocean waters through the Gibraltar sill would result in a Mediterranean sea level fall at a rate of about 0.5 m yr−1 (ref. 2). During the Messinian Salinity Crisis (MSC), the deposition of nearly one million km3 of evaporite salts3,4 (mainly gypsum, CaSO4.2H2O, and halite, NaCl) is a testimony of highly restricted hydrological exchange with the Atlantic Ocean. Whether or not evaporite deposition was accompanied by a major (≥1 km) sea level drawdown, however, is still debated. Estimates of the magnitude of the base-level fall during the MSC rely on geological indicators such as deeply incised canyons onland5,6,7,8,9, widespread erosional surfaces on the margins10,11,12,13 and in the basins14,15,16 and evaporite facies from drilled cores17,18,19,20. However, these indicators are not interpreted unequivocally21,22,23,24 and drawdown estimates range from ~ 200 m to ~ 2 km9,11,14,21,25,26,27. Existing hydro-chemical models cannot solve the controversy because they show that the complete MSC halite body can be deposited in scenarios that include periods of Mediterranean isolation and sea level drawdown28, but also in scenarios involving exclusively a partially restricted, brine-filled Mediterranean9,10,29.

We use the chlorine stable isotope composition of halite to estimate the volumetric halite precipitation rate in the Mediterranean basins during the MSC. Compared to other geochemical tracers such as strontium isotopes30,31, chloride has the advantage of being a major constituent of halite and, for all practical purposes, is derived solely from inflowing seawater. Laboratory contamination issues are also minimal for this reason. At the basis of our approach is the chloride isotope reservoir effect which results from the preferential incorporation of the heavy isotope 37Cl (as opposed to the lighter isotope 35Cl) into precipitating halite32. The extent of the isotope reservoir effect depends on the balance between the rate of halite precipitation, which tends to decrease the 37Cl/35Cl ratio of dissolved chloride, and the rate of addition of dissolved chloride from the Atlantic Ocean, that tends to push Mediterranean 37Cl/35Cl towards heavier global ocean values. The faster the precipitation, the stronger the 37Cl-depletion of the precipitating halite. Evaporative drawdown of a halite-saturated water column33 results in very high rates of halite precipitation, and in the most extreme dissolved Cl, and thus halite, 37Cl-depletion.

Results and discussion

Chloride isotope composition of halite

We analyzed 60 drill cuttings from the Aphrodite well (Levantine basin, eastern Mediterranean) starting from the onset of halite precipitation and spanning a thickness of 1393 m upcore (Fig. 1). All analytical results are presented in Supplementary Tables 1–3. In the Levantine basin, the Messinian evaporite package is divided into the following seven seismic subunits34. Four transparent units are composed mainly of halite. Two intercalated, highly reflective units are composed of halite with a higher, albeit laterally variable, clay content. A seventh unit is composed of a mixture of anhydrite, clays and sands, separated from the underlying units by the Intra-Messinian Truncation Surface (IMTS, interpreted as a dissolution surface)34 (Fig. 1d). Bulk XRD analyses of the Aphrodite cuttings (Fig. 2) show that, within seismo-stratigraphic units 1 to 4 and the bottom 360 m of seismo-stratigraphic unit 5 (3943–2700 m depth interval), halite dominates the mineral assemblage (>80 wt-%), with minor (<20 wt-%) amounts of anhydrite (CaSO4) and only trace amounts of CaCO3 and quartz. In the topmost 140 m of seismo-stratigraphic unit 5, the abundance of halite decreases sharply, and that of all the other minerals increases. No samples are available from the well above 2550 m depth, hence no samples are available from unit 6. The drill cuttings also contain sylvite (KCl) and barite (BaSO4) (Supplementary Fig. 1) which were artificially added to the drilling fluids to control their salinity and density, respectively.

Fig. 1: The Messinian Salt Giant of the Mediterranean sea.
figure 1

a Study area showing data and on-land and offshore Messinian evaporite thickness distribution4 over a Mediterranean Basin map (background relief and bathymetry maps taken from https://www.geomapapp.org/). Chloride isotope data included in this work comes from the hydrocarbon industry well Aphrodite (AP2) located in the Levantine basin (eastern Mediterranean) and from Deep Sea Drilling Project (DSDP) sites 134 (western Mediterranean), 374 (Ionian Basin, eastern Mediterranean) and 376 (Florence rise, eastern Mediterranean). b location of the seismic section (red dashed line) shown in d. c Petrophysics and associated lithologic log of the Aphrodite well. GR-Gamma Ray (graph ranges between 0 – left - and 100 API) and RES-Resistivity (graph ranges between 0 – left - and 1000 Ωm; logarithmic scale) logs; d Composite seismic section (d-d’) in the Levantine basin showing the Messinian evaporites composed of seven seismo-stratigraphic units34. Notes: in a, symbols are as follows: RB, Rhodes Basin; FB, Finike Basin; AB, Antalya Basin; CB, Cilicia Basin; LRB, Larnaca Basin; LTB, Latakia Basin; LEB, Levant Basin; HB, Herodotus Basin; CYB, Cyrenaica Basin; SB, Sirte Basin; IB Ionian Basin; TB, Tyrrhenian Basin; L-PB, Liguro-Provençal Basin; ALB, Algerian Basin; in the lithological log of c, orange layers represent clastic inclusions. Black horizon in seismic section of d is a high-amplitude continuous reflection corresponding to a layer of clastics encased in halite; in d, vertical distance is in km below sea level.

Full size image
Fig. 2: Composite log and isotopic results.
figure 2

X-Ray Diffraction (XRD, colour legend as in Fig. 1c) (a), mineralogical (c–f) and chloride isotopic (g) composition of cuttings from the Aphrodite 2 hydrocarbon exploration well from the deep Levant Basin, alongside a seismic line crossing the same well (b), and chloride isotopic composition (h) of halite samples from DSDP sites 134, 374 and 376. Notes: Quartz, CaCO3 (calcite), CaSO4 (anhydrite) and NaCl (halite) are in weight-% of dry sediment, discarding the presence of barite (BaSO4) and sylvite (KCl) which were added to the drilling fluids. FBI stands for Foraminifer Barren Interval36. The red dashed line in g and h represents the equilibrium chlorine stable isotope composition of halite precipitating from seawater. In c–h, yellow layers (units 1, 2, 4 and 6) are nearly pure halite, gray layers (units 3 and 5) are halite mixed with terrigenous material. In g, the red circle and red star represent the δ37Clhalite data points used to fit the numerical model providing a range of drawdown estimates (Fig. 3). The horizontal black line in b is the high-amplitude continuous seismic reflection corresponding to a layer of clastics encased in halite, also shown in Fig. 1d; the dotted black line in c through g represents the vertical position of this reflector relative to the cuttings of core Aphrodite 2. Vertical distance is in meters below sea level. Source data are provided as a Source Data file.

Full size image

Halite grains for Cl-isotope analyses were selected from the drill cuttings under a binocular microscope, excluding grains that contained evident signs of impurities. We cannot exclude, however, that minor amounts of Cl-bearing sylvite (KCl) contributed to the Cl-isotope signature we measured. Mass-balance calculations, however, suggest that the contribution is negligible (see supplementary materials). The δ37Cl values at the base of the halite layer (0.4 ‰ vs standard mean ocean chloride, hereafter SMOC) are in isotopic equilibrium with oceanic dissolved chloride (0 ‰ vs SMOC) (Fig. 2g). Subsequent halite, forming Levantine basin seismo-stratigraphic units 1 to 4 and the bottom 360 m of seismo-stratigraphic unit 5 (3943–2700 m depth interval), is slightly depleted in 37Cl (average δ37Cl = 0.31 ‰ vs SMOC) compared to equilibrium with global ocean dissolved chloride. We interpret this as reflecting a modest isotope reservoir effect, in which the rate of dissolved chloride sequestration into halite becomes important relative to the influx of chloride from the Atlantic. In the topmost 140 m of seismo-stratigraphic unit 5, δ37Cl values decrease abruptly, reaching a minimum of −0.30 ‰ vs SMOC at 2630 m depth (Fig. 2g). This is interpreted to reflect a sudden increase in the rate of halite precipitation relative to seawater chloride replenishment.

We also analyzed 9 halite samples from drill cores obtained by drilling the topmost few tens of meters of the deep-basin Mediterranean halite layer during Deep Sea Drilling Project (DSDP) Legs 13 and 42 A from the Florence Rise (Site 376, eastern Mediterranean), Ionian Abyssal Plain (Site 374, eastern Mediterranean) and west Sardinian margin (Site 134, western Mediterranean). A stratigraphic correlation between these sites and the Aphrodite section is not possible, although the position of the DSDP halite samples at the top of the halite body implies that they represent the final stages of the high-salinity evaporite phase of the MSC (Fig. 2). Halite from DSDP samples shows a range of 37Cl depletions from near-equilibrium with oceanic Cl at Site 376 (0.33 <δ37Cl ‰ vs SMOC < 0.37), to a strong 37Cl-depletion at Site 374 (−0.14 <δ37Cl ‰ vs SMOC < 0.04). Halite from DSDP Site 134 has intermediate δ37Cl (0.08 ‰ vs SMOC) (Fig. 2). The 37Cl-depleted isotope signature of halite from DSDP sites cannot be influenced by the composition of drilling fluids, because the halite sub-samples were obtained by mechanically breaking solid halite layers in the DSDP cores, rather than extracted from an admixture of drill cuttings and drilling fluids like for the Aphrodite samples.

Two-phase deposition of the Mediterranean halite layer

We used a numerical model to quantify halite precipitation rates and the extent of sea-level drawdown during the MSC. The model is based on mass balances of water and dissolved ions (35Cl−, 37Cl−, Na+, Mg2+, SO42−, Ca2+, K+ and HCO3−) in the western and eastern Mediterranean basins and is forced by the evaporative flux, continental runoff and hydrological restriction at the Gibraltar sill. Evaporation and continental runoff fluxes are based on the Miocene General Circulation Model results of Simon et al.35. Starting from initial chemical and hydrologic conditions corresponding to the modern Mediterranean, the model is integrated forwards in time for 20 kyr, after which the hydrological connectivity at Gibraltar is impeded, salinity rises, and evaporitic minerals precipitate. With a salinity-corrected evaporation rate starting at the average value (1 m yr−1) reported in Simon et al.35 values of continental runoff and hydrological restriction at Gibraltar are found to obtain a good match between simulated and observed4,25 halite volumes and halite δ37Cl values. A full description of the numerical model, model forcing and model constraints is given in the supplementary material.

The best fit of the model to the volume and δ37Cl values of halite (Fig. 3) is obtained if a two-phase hydrological restriction-isolation scenario at Gibraltar is considered. First, the deep-water outflow at Gibraltar is restricted but not blocked, leading to the precipitation of gypsum, followed by the precipitation of halite. This halite is characterized by a modest chloride isotope reservoir effect, and decreasing halite δ37Cl from the value in equilibrium with oceanic dissolved chloride (0.4 ‰ vs SMOC) to the average value measured from seismo-stratigraphic unit 1 to the lower 360 m of unit 5 (0.31 ‰ vs SMOC) (Fig. 3h). In this first phase, lasting about 35 kyr, halite precipitates exclusively in the brine-filled eastern Mediterranean; halite does not precipitate in the western Mediterranean (Fig. 3e) because the brine transferred from the eastern to the western Mediterranean basins across the Sicily sill is in thermodynamic equilibrium with halite, but not supersaturated.

Fig. 3: Results of the standard model run in the western (orange dashed line) and eastern (solid blue line) Mediterranean basins.
figure 3

a Salinity in g/kg; b water level in km; c volume (km3) of precipitated calcite (CaCO3); d volume (km3) of precipitated gypsum (CaSO4.2H2O); e volume (km3) of precipitated halite (NaCl) and volume (km3) of total precipitated salts (faint black line); f volume (km3) of precipitated epsomite (MgSO4.7H2O); g volume (km3) of precipitated kainite (KClMgSO4.3H2O); h δ37Cl of halite and kainite calculated by the model compared to the measured δ37Cl values of halite form Units 1 to the bottom 2/3 of Unit 5 in the Aphrodite-2 core (AP-phase 1), halite from the top 1/3 of Unit 5 in the Aphrodite-2 core (AP-phase 2) and the halite samples at DSDP site 134 (west Sardinia margin). The blue shaded areas represent the ranges of water level (b), halite volume (e) and δ37Clhalite (h) in the eastern Mediterranean that result if the model is fitted to the lowest measured δ37Clhalite value (red circle in h) or the 2nd lowest δ37Clhalite value (red star in h), by increasing the freshwater continental input in the eastern basin during the drawdown phase (see text); these two δ37Clhalite data points are highlighted in Fig. 2; the green lines in b, e and h represent model results corresponding to this scenario of increased continental water input.

Full size image

After this first phase, the deep outflow at Gibraltar is blocked and the inflow of Atlantic water into the Mediterranean is stopped, leading to the complete isolation of the Mediterranean Sea from the Atlantic Ocean. As a result, the water level in the entire Mediterranean basin drops until the depth of the Sicily sill is reached, and then continues to fall independently in the western and eastern Mediterranean basins, forced by their respective freshwater budgets (Fig. 3b). The rate of water level fall is very rapid, with 50% of the total drawdown taking place in less than 3.5 kyr, 80% in less than 6 kyr, and the total drawdown attained in less than 10 kyr. Evaporative drawdown causes an abrupt increase in the rate of halite precipitation in the eastern Mediterranean, and to the initiation of halite precipitation in the western Mediterranean (Fig. 3e). The increased halite precipitation rate results in a strong Cl-isotope reservoir effect and an abrupt drop in the δ37Cl of precipitating halite (Fig. 3h).

In the eastern Mediterranean basin, ~ 40% of the total halite volume formed in the standard model run accumulates in 35 kyr, while the total halite volume accumulates in ~ 50 kyr, in agreement with a “rapid36” ( ~ 50–60 kyr), rather than a “slow37” ( ~ 600 kyr), mode of halite accumulation. This time frame is consistent with the widely accepted stratigraphic position of the halite deposit being accumulated during phase 2 of the MSC across the Mediterranean basins (5.60–5.55 Ma; e.g. CIESM38; Roveri et al.3; Ryan et al.39) and, more specifically, in the Levant Basin (Manzi et al.36).

After roughly 10 kyr from the beginning of the second phase, the calculated volume of halite in the western (74,900 km3) and eastern (492,300 km3) Mediterranean basins approaches the estimates of Heida et al.25 (110,000–115,000 km3) and Haq et al.4 (585,000–657,000 km3), respectively, while the δ37Cl of halite in the eastern (−0.30 ‰ vs SMOC) and the western (0.08 ‰ vs SMOC) Mediterranean matches the minimum values measured from the upper 140 m of unit 5 in the Aphrodite well and the top of the salt layer at DSDP Site 374 (both in the eastern Mediterranean) and the value measured in the western Mediterranean at DSDP Site 134, respectively (Fig. 3h). Sensitivity tests (see supplementary material section) show that only a small range of continental runoff values and hydrological restriction at Gibraltar results in a good match of the model to the halite δ37Cl data and halite volume estimates4,25, and that within this range, the calculated time to accumulate the halite body and the extent of the drawdown change little.

Our results solve a controversy regarding the mode of halite precipitation – deep basin, deep water40 vs deep basin shallow water17 - that is outstanding since the discovery of the deep basin evaporites during DSDP Leg 1317 (Fig. 4). We show that in the eastern Mediterranean approximately 40% of the halite deposit was formed in a water-filled Mediterranean basin, connected to the Atlantic Ocean by a two-way hydrological exchange, corresponding to the deep-basin, deep-water model40. The remaining 60% of the eastern halite deposit was formed by evaporative drawdown of a halite-saturated water body, which is reminiscent of the deep-basin, shallow water model17, except that halite precipitation takes place during the drawdown, rather than in a deep, desiccated basin after the drawdown took place.

Fig. 4: Modes of halite accumulation during the Messinian Salinity Crisis deduced from chlorine isotopes of halite and geochemical modeling presented in this work.
figure 4

a Two phases of halite accumulation. During phase 1, halite accumulates in a brine-filled Mediterranean basin that is connected to the Atlantic Ocean but has a hydrologically restricted outflow; during phase 2, the Mediterranean is fully isolated from the Atlantic Ocean and halite accumulates during a major evaporative drawdown event; b tentative comparison of the phases of halite accumulation (this work) with the Messinian Salinity Crisis (MSC) chrono-stratigraphic stages after CIESM38, Roveri et al. (ref. 3) and Ryan39. Notes: (i) while we can estimate the duration of the two phases of halite accumulation (a), we cannot date them. The tentative correlation with MSC chrono-stratigraphic stages (b) is based on the assumption that halite accumulation is restricted to phase 2 of the MSC3,38,39; (ii) due to lateral erosion-transport processes, at the Aphrodite 2 site the relative thickness of halite layers is different from the halite volumetric ratios at the scale of the eastern Mediterranean basin.

Full size image

Drawdown of the Mediterranean sea

The water level drawdown we constrain is significant in the western Mediterranean ( ~ 850 m), and extreme in the eastern Mediterranean ( ~ 2100 m), implying that western and eastern sub-basins lost 43% and 83% of their water volume, respectively. In the eastern Mediterranean, large parts of the continental margins would have been completely desiccated, though the deep Ionian, Herodotus and Antalya basins were still covered by a > 1 km-thick brine layer. Our model also predicts that the drawdown in the eastern Mediterranean resulted in the precipitation of the Mg-K-salts epsomite (MgSO4.7H2O; 46,700 km3, Fig. 3f) and kainite (KMgClSO4.3H2O; 23,200 km3, Fig. 3g), although Mg-K salts were not detected by XRD in our samples, possibly due to a lateral heterogeneity of brine composition in the eastern Mediterranean, or their dissolution in the drilling fluids or during the washing of drill cuttings. The sequence of precipitation halite-epsomite-kainite is consistent with that observed in seawater evaporation experiments where the minerals are separated from the evaporating brine as they precipitated (“fractional precipitation)41.

In an attempt to obtain a more conservative estimate of the drawdown extent, we increased the continental freshwater input to the eastern Mediterranean basin in order for the model to reproduce the second-most negative halite δ37Cl data point measured from cuttings of the Aphrodite well (δ37Cl = −0.19 ‰ vs SMOC; Fig. 2e, star). In these conditions, a more limited drawdown of ~1700 m is calculated (Fig. 3b, green line and star). Taken together these model runs imply that the eastern Mediterranean experienced a large-magnitude drawdown of 1.7–2.1 km during the MSC.

Independent evidence for the occurrence of a sea-level drop and for the subaerial exposure of the basin margins come from the decrease in halite concentration and the peak in Quartz abundance (up to 58 wt-%) in the top 140 m of Unit 5, suggesting significant terrigenous input into the basin during drawdown. The erosion that took place concomitant with the short-lived sea-level drawdown resulted in only a limited accumulation of halite during drawdown at the Aphrodite site (upper 1/3 of Unit 5 above the black horizon traced in the seismic line displayed in Fig. 1d and Fig. 2), compared to the underlying, thick halite layer deposited under a full Mediterranean water column (Units 1–4 and lower 2/3 of Unit 5), leading to a non-linear relation between thickness and volume.

The drawdown of 1.7–2.1 km estimated in this work contrasts with the smaller-magnitude (600 m) drawdown estimated from the restoration of the geomorphological base level of the buried Messinian Nile canyon9, although the canyon development cannot be precisely dated. If the two estimates refer to the same drawdown event, this implies that the large drawdown was too short-lived to be recorded in the Nile valley geomorphological profile. However, our work shows a complete disconnection of the eastern Mediterranean from the western Mediterranean, implying that, following the first, short-lived, large-scale drawdown, water levels in the eastern Mediterranean would have varied under the influence of an oscillating freshwater budget35, and of the hydrological connection to the Paratethys drainage system42. This hydrological mode, which lasted until the reconnection with the Atlantic via the western Mediterranean, possibly as late as the Zanclean (5.33 Ma)43,44, leaves ample time for intermediate water levels such as those deduced from the study of topographic profile of the Messinian Nile Canyon9 to develop.

Consequences of Mediterranean Sea drawdown

Our results have broader implications for the biological, geologic and climatic evolution of the Mediterranean realm, and beyond. In the western Mediterranean sub-basin, the 0.85 km sea-level drawdown we constrain would have resulted in the formation of a land-bridge connecting Africa and Europe through the Betic region and the Balearic Islands. This land bridge would have allowed continental-terrestrial vertebrate fauna to colonize the Balearic islands, a migration that is known to have taken place during the Messinian based on paleontological evidence45. A drawdown of 1 km in the western and eastern Mediterranean may have resulted in enough lithospheric unloading to trigger a magmatic pulse in the form of volcanism and dykes in the pan-Mediterranean area at the end of the MSC46. Our evidence for drawdowns of ~0.85 km and 1.7–2.1 km in the western and eastern Mediterranean, respectively, make this scenario even more likely. Alongside lithospheric unloading, the huge size of the Mediterranean depression created by MSC water level drawdown - corresponding to a volume loss of 69% of the Mediterranean water body - would have generated planetary-scale climate effects47, inducing changes in precipitation patterns, a scenario suggested by rainfall proxy data47.

Methods

X-Ray diffraction (XRD)

The XRD measurements were carried out on an Empyrean diffractometer (@Malvern-Panalytical), equipped with a copper tube and a PIXcel multi-channel detector. The configuration is Bragg-Brentano θ - θ and the sample holder is a spinner. The mineral phases are identified using Highscore plus (@Malvern-Panalytical) software and the ICSD-Panalytical database. Semi-quantitative evaluation of mineral abundances is obtained by scaling the intensities of the principal mineral diffraction peaks to those of reference materials applying reference intensity ratio (RIR) coefficients.

Cl-isotope measurements

We measured the chloride isotope composition of halite grains hand-picked under a binocular microscope, which overwhelmingly reflects that of halite. First, the halite grains were dissolved in distilled water. Then, dissolved Cl ions were precipitated into AgCl and the AgCl converted into CH3Cl48,49,50. The δ37Cl of CH3Cl was measured with a Delta plus XP or Delta V ThermoFisher dual-inlet gas source mass spectrometer. The δ37Cl compositions are reported in ‰ vs SMOC. The external reproducibility of the internal IPGP standard seawater dissolved chloride (Atlantique 2) was ± 0.025 ‰ (1σ, n = 48); note that this S.D. value is smaller than the size of the δ37Cl symbols in Fig. 2g.

Numerical model

A full description of the numerical model is provided in the supplementary material.