The Messinian Salinity Crisis: A Kilometre of Salt, and an Argument About How Dry It Got

The same cores carried something the salt alone does not say. Alongside the gypsum, anhydrite and rock salt, Leg 13 recovered arroyo gravels and red and green floodplain silts. An arroyo gravel is a desert stream deposit: it forms in a dry channel that runs in flood and then stands empty. Pulling one out from under two kilometres of Mediterranean water is the most concrete single statement anyone can make about that spot having once been open air. Our own research file does not mention these deposits at all, and if this article could keep only one fact, it would keep this one.

The total volume of Messinian evaporites is estimated at roughly 1 million cubic kilometres. That figure carries an implication that belongs beside it, because the implication is what makes the number make sense: precipitating that much salt out of seawater takes more than one evaporation. The basin was refilled and evaporated repeatedly across the crisis, not once.

That total is softer than a round million sounds. The published framing that carries the figure also states that the estimate may be reduced by 50 to 75 percent as more information becomes available, and the mass figure sometimes quoted next to it is the same estimate expressed a different way, not a second and independent confirmation of it. The honest form is on the order of a million cubic kilometres, not a total.

Where the buried evaporites are thickest they reach as much as 3 km. That is a maximum, not a basin-wide average, and it should never be restated as three kilometres of salt lying evenly under the Mediterranean. This article's title says a kilometre of salt because a kilometre is the conservative end of what the deposit actually reaches, and the conservative end is the one worth putting in a headline.

Rivers flowing into the Mediterranean carved deep canyons during the crisis, matching a base level far below the one they had been graded to. The Rhone canyon extends more than 1,000 m deep beneath the present Camargue.

The Nile cut deeper still, and here this article has to correct its own source. The buried Messinian Nile canyon reaches roughly 2,500 m below present sea level just north of Cairo. Our research file's quick summary attaches that same 2,500 m figure to Aswan, roughly 1,000 km upstream, and that is wrong: at Aswan the canyon floor is around 200 m below sea level. The body of the file gives the depth without naming a city and is defensible; the summary line is not, and nobody should carry it forward. The two-city split is standard in the Nile canyon literature, including in the papers that dispute what the canyon means. Honest limit on our end: the ultimate primary source for both figures is older Egyptian subsurface work that this research pass did not fetch directly.

The third line of erosional evidence is not a landform but a surface. Seismic reflection profiles show a prominent erosional horizon, the M-reflector, running across the Mediterranean margins, and an erosional surface of that kind records subaerial exposure: air, weather and running water where there is now a water column. Taken with the canyons, these features are not explicable without a dramatic drop in Mediterranean water level.

The crisis is dated astronomically rather than by estimate. Krijgsman and colleagues, publishing in Nature in 1999, calibrated Mediterranean sedimentary cycles against the astronomical record and placed the onset at 5.96 million years ago. The end, the Zanclean flood, is dated to 5.33 million years ago. One note on precision: the current major review of the crisis, Krijgsman and colleagues again in Nature Reviews Earth & Environment in 2024, gives the onset as 5.97 million years. That is a refinement of the same astronomical calibration rather than a rival date, and this article uses 5.96 throughout instead of blending the two into a range neither source states.

The cause was tectonic. The African and Eurasian plates converged, progressively restricting and then closing the seaway at the Gibraltar arc that connected the Mediterranean to the Atlantic. Our file's own hedge on this is load-bearing and is kept here intact: the exact geometry and timing of that closure are reconstructed from structural geology and paleogeographic models, not directly observed. That hedge is also precisely where the 2025 challenge below lands.

The rest is a water budget. Evaporation from the Mediterranean greatly exceeds the freshwater its rivers deliver. Cut off the Atlantic inflow and the basin, roughly 3.7 million cubic kilometres of water, would dry up in scarcely more than 1,000 years. Notice what kind of statement that is. It is a hydrological calculation about a modern basin, not a measurement of anything that happened. It tells you that drying the Mediterranean is physically easy. It does not tell you how far the drying went.

Kenneth Hsu and colleagues proposed the desiccation model in 1973, in Nature and in the Initial Reports of the Deep Sea Drilling Project: that the Mediterranean had largely or completely dried, leaving a vast salt desert thousands of metres below present sea level. That the proposal was made, by whom, and when, is settled. Whether it is correct is the rest of this section, and it is not settled at all. The Tier 1 status of the proposal must not be allowed to leak into its conclusion.

The deep desiccation model, which our file attributes to Hsu, Clauzon and Garcia-Castellanos, holds that the Mediterranean was reduced to one or more hypersaline lakes, something like the modern Dead Sea, lying at the bottom of a vast depression roughly 3 to 5 km below modern sea level. One caveat on that attribution. Daniel Garcia-Castellanos, named there as a proponent, is a co-author of the 2022 paper below that argues for a far shallower drawdown. The attribution is fair for the position as it stood around 2009 and stale for the position now. Hsu owns the model unambiguously; the others should be dated to their era.

The shallow-water model is not a fringe objection to that. It is the position of a formal international workshop consensus, the CIESM workshop of 2008, and of the field's most-cited modern review, Roveri and colleagues in Marine Geology in 2014. On this reading the evaporites formed in shallow marginal basins while the deep Mediterranean retained water, and the deep basins were never fully desiccated. The same salt, deposited in a different place, under a different sea.

Our own file leans, and the hedge inside its lean is not decoration. It says that seismic evidence supports deep-basin evaporite deposition in at least some areas, favouring substantial, if perhaps not total, desiccation. Those four words, if perhaps not total, are the file declining to say the thing its own quick summary and this article's original working title both wanted it to say.

In 2022 the Nile canyon changed sides. Gvirtzman, Heida, Garcia-Castellanos, Bar, Zucker and Enzel, publishing in Communications Earth & Environment, restored the canyon's original topography and the vertical position of the Messinian coastline by unloading the post-Messinian sediment and accounting for flexural isostasy and compaction. Their estimate for the original depth of the Nile's geomorphological base level is about 600 m below present sea level: a drawdown 2 to 4 times smaller than the raw canyon had been read to imply, with the salt precipitating under 1 to 3 km of water rather than on a dry floor. Their abstract states that this conclusion is at odds with the nearly-desiccated basin model that has dominated the scientific literature for 50 years. Their own closing hedge belongs with it: a 600 m drawdown is still about five times larger than any glacial sea-level fall, so this is a downgrade of the drying, not a denial of it.

The case against went further in 2025. Roveri, Lugli and Manzi, in the Annual Review of Marine Science, published a fifty-year retrospective whose title names part of the standard story as myth. Their abstract states that the megaflood hypothesis rests on an erosional surface cutting upper Miocene rocks and sealed by lower Pliocene sediments, and that recent investigations suggest the connection between the Atlantic and the Mediterranean through Gibraltar may have remained active throughout the crisis, indicating more gradual physical processes and possibly a Mediterranean that was never desiccated. Two limits on how far this article can use that paper. The publisher's page refused our request, so we have read the abstract as deposited with CrossRef and not the paper itself; and an erratum record exists for it in the same series whose content we could not retrieve. Nothing is attributed to it here beyond what that abstract supports, and in particular no specific claim is named as one of the myths of its title.

The other camp has not stopped publishing either. Garcia-Castellanos, Heida, Palcu, Bulian and Sierro, in Science Advances in 2025, present a landscape evolution model in which erosional waves propagating into the surrounding continents added a gradual sea level rise on top of climatic oscillations inside the Mediterranean, offered as an explanation for the transition to fresher-water conditions. It is explicitly a model and it is carried here as one. Its value for this article is that the dispute is not one side advancing on a static opponent. Both camps have gained ground in the same few years.

One more disagreement belongs here, and it is ours. Our sister file on catastrophic flood geomorphology states flatly that the Mediterranean had evaporated almost completely, reduced to disconnected hypersaline lakes, with no hedge and no rival named. The file this article is built on does not do that, and is right not to. Where our own library disagrees with itself, this article follows the file that carries the dispute rather than the one that resolves it for free.

Garcia-Castellanos and colleagues, publishing in Nature in 2009, modelled the reopening of the Atlantic connection at the end of the crisis. The simulation produces a catastrophic refill: peak flow rates of up to 100 sverdrups, where one sverdrup is a million cubic metres per second, which puts the peak at roughly 1,000 times the discharge of the Amazon. In the published framing, ninety percent of the basin's flooding happens in an estimated window of somewhere between several months and two years, with complete refilling possibly taking about a decade. Every one of those figures is an output of that model. This is the single most important provenance flag in the article: none of them is a reading taken off anything.

The rest of the flood's superlatives come from the same place and deserve the same label. Sea level inside the basin rising at times by more than 10 metres per day: modelled. Erosion of the Camarinal Sill at 0.4 to 0.7 metres per day: modelled. A water drop of more than 1,000 metres at the breach: modelled. An earlier one-dimensional model assumed a range of more than 10 to 100 sverdrups, which is where the familiar figure entered the literature in the first place. These are not survey data, and the people who produced them never presented them as such.

One caution about apparent corroboration, and it is ours again. Our sister file gives the discharge as up to 100 million cubic metres per second, a thousand times the Amazon. That is arithmetically the same figure as 100 sverdrups, because a sverdrup is a million cubic metres per second. Two of our files carrying one number in different units is not two independent estimates, and it should never be read as agreement between sources.

Now the measured side. Garcia-Castellanos and colleagues published a review in Earth-Science Reviews in 2020 asking a pointed question: what evidence for this flood exists independently of the model that proposed it? The main answer they identify is a channel. It runs nearly 390 km from the Gulf of Cadiz in the Atlantic to the Algerian Basin in the western Mediterranean, is several hundred metres deep, and implies the excavation of roughly 1,000 cubic kilometres of Miocene sediment and bedrock. The channel is mapped. What cut it is the argument.

That review's own language is worth carrying rather than smoothing. It calls the kilometre-scale drawdown alleged, and the megaflood one of the competing scenarios proposed for the termination of this environmental crisis. The lead author of the paper that first proposed the flood is the lead author of the review that asks whether the flood holds up independently of it. That is the register this article is trying to match.

The first physical deposit arrived in 2018. Micallef and colleagues, in Scientific Reports, used seismic profiles and borehole data to identify an extensive, buried, chaotic sedimentary body in the western Ionian Basin, sitting above the Messinian salts and below the Plio-Quaternary open-marine sequence, consistent with a megaflood passing from the western into the eastern Mediterranean through a south-eastern Sicilian gateway. Measured off that data: a volume of 1,430 to 1,620 cubic kilometres, an extent of 160 km by 95 km, and a maximum thickness of 760 to 860 m. The paper's abstract argues that the body provides evidence for a large amplitude drawdown in the Ionian Basin, supports a Mediterranean-wide catastrophic flood at the end of the crisis, and suggests this is the largest known megaflood deposit on Earth.

The same paper also carries a peak discharge of roughly 100 million cubic metres per second and a flow velocity of up to 45 metres per second. Those were computed, not measured. This is the teaching case for the whole article: one paper, one team, one set of data, and two entirely different kinds of number inside it. The deposit's size was read off seismic reflection data and boreholes. The discharge and the velocity came out of a calculation. Both are legitimate science. They are not the same kind of fact, and a reader who takes the second for the first has been misled by nothing but the retelling.

In December 2024 the same lead author published land-to-sea evidence in Communications Earth & Environment, for the flood spilling over a shallow-water marine corridor in south-east Sicily. In the field: more than 300 asymmetric, streamlined erosional ridges aligned with the flood direction; poorly sorted breccia deposited between the Messinian and the Lower Zanclean Trubi Formation; soft-sediment deformation structures and clastic injections in the breccia and the units beneath it; and a 20 km wide erosional shelf channel connecting the ridges to the Noto Canyon. Those are observations of rock in outcrop and on the shelf.

The discharge associated with that route, 68 to 100 sverdrups, is again a modelled range and not a measurement. One limit on our own reporting of this paper: we have its findings through institutional publication records and press coverage rather than through its full text, so the detail above should be read as accurately relayed rather than independently checked against the paper's own pages.

There is a named rival for the flood as well as for the drying, and it is a different argument from the one above. Roveri and colleagues in 2014 propose a more gradual refilling through multiple phases, with the most catastrophic flooding confined to the final stage. That is a real third position between a single instantaneous breach and no flood at all.

And a fourth position, on the mechanism rather than the pace. Blanc, in Geodinamica Acta in 2002, argues from the geomorphology of the strait together with budget modelling and hydraulic calculation that the modern Strait of Gibraltar was opened by retrogressive erosion cutting back along an eastward-flowing Messinian stream, and that the Plio-Quaternary strait shares no geographical feature with the Miocene portals, which had been continentalised before the crisis ended. He also reads the widespread Lago-Mare facies as showing that the Mediterranean was fully isolated from the world ocean at the end of the crisis. This article cites that paper for the strait, which is what it is about, and for nothing else.

The submarine channel at the Gibraltar sill is real mapped topography, hundreds of metres deep, and a refilling flood would have eroded exactly there. The modern geomorphology of the strait is consistent with that. Reading the channel as the scar of one catastrophic event is the modelled part, and the 2025 review above argues that the same erosional surface admits a gradual explanation instead.

The claim, first. Removing a large fraction of the world ocean's dissolved salt into the Mediterranean evaporites would have lowered global ocean salinity by roughly 2 to 3 practical salinity units, with possible effects on thermohaline circulation and global climate. Now the two corrections. The first is arithmetic: our file pairs that salinity range with a figure of about 5 percent of the ocean's dissolved salt, and those two numbers do not reconcile, because 5 percent of a roughly 35 psu ocean is about 1.75 psu, not 2 to 3. This article does not restate the pair as though it worked. The second is a citation: the file attaches this claim to Blanc (2002), a paper about the geomorphology of the Gibraltar strait opening that says nothing about global ocean salinity or thermohaline circulation. We do not cite it here for this, and the next claim gives the source we do use.

The current major review frames the same effect more defensibly. Krijgsman and colleagues, in Nature Reviews Earth & Environment in 2024, define salt giants as massive salt deposits at least hundreds of metres thick formed during the evaporation of semi-enclosed seas, and put their effect at roughly 7 to 10 percent net extraction of evaporite ions from ocean water, persisting over million-year timescales, which is their argument for including salt giants in long-term carbon cycle models. They also identify calcium removal during salt precipitation as a mechanism that altered ocean chemistry and through it affected atmospheric carbon dioxide and global temperature. Note that an extraction of that size would give roughly 2.5 to 3.5 psu, close to the range our file states, so it may well be the percentage in our file that is stale rather than the salinity figure.

Closer to home, the crisis may have contributed to late Miocene climate cooling and to aridification in the Mediterranean region, with cascading effects on North African and Southern European ecosystems. The may have is our file's own wording and it stays exactly where it is.

With the sea down, there was a temporary land bridge between Africa and Europe, and the episode is sometimes called the Messinian terrestrial crisis. Some evidence suggests mammalian dispersals between the two continents in this window, but the record is sparse, which is why it sits here rather than higher. One thing needs saying plainly, because the phrase land bridge pulls English toward a picture it does not license. The animals crossing were late Miocene mammals; camelids and gerbils are the creatures named in the standard illustrated reconstruction of the crossing. There were no people. There would be no people anywhere near this region for millions of years.

Inside the basin, the extreme environments the crisis produced, hypersaline lakes, salt flats and deep canyons, may have driven speciation and extinction among Mediterranean marine organisms. The fossil evidence is limited and the claim stays speculative on our file's own assessment.

No, the Zanclean flood cannot be connected to any human cultural memory or flood mythology. It was a genuine geological catastrophe with extensive physical evidence behind it, and it happened roughly 5.33 million years ago, millions of years before any hominins inhabited the Mediterranean region. There was no witness, no survivor, and no consequence for any person. Connecting an event of that age to human flood traditions is anachronistic, and our file labels it debunked without qualification. Every superlative in this article describes a planet with nobody on it.