Between roughly 5.97 and 5.33 million years ago, the Mediterranean Sea was cut off from the Atlantic and lost most of its water to evaporation. The event is called the Messinian Salinity Crisis, and it left behind something hard to argue with: a layer of salt on the Mediterranean floor, in places up to three kilometres thick, amounting to close to a million cubic kilometres of evaporite. That salt is the physical record of a sea that was cut off, concentrated, and in places drawn down on a staggering scale.
The end of the crisis is the part that draws the attention. When Atlantic water found its way back through the Strait of Gibraltar around 5.33 million years ago, one widely cited model suggests the basin refilled fast enough that sea level inside it rose by more than ten metres a day at the peak. That figure is a model result, not a measurement, and it is worth being clear about which parts of this story are solid and which are inference.
What the salt actually shows
The crisis was identified in the 1970s, after deep-sea drilling and seismic surveys turned up enormous buried evaporite deposits beneath the Mediterranean seabed. Kenneth Hsü, William Ryan and Maria Cita published the desiccation interpretation in Nature in 1973, and the salt giant has anchored the subject ever since.
What the salt shows clearly is restriction and evaporation. The Mediterranean loses far more water to evaporation than its rivers and rainfall return, so it depends on inflow from the Atlantic to stay full. Close that connection, or narrow it enough, and the basin concentrates, then precipitates its dissolved salts. The thickness and extent of the deposits leave little doubt that this happened.
How far the sea level actually fell is a separate question, and a contested one.
Dry basin, or deep brine
There are two broad readings of the evidence, and they have coexisted for decades. One holds that large parts of the Mediterranean floor were exposed, a desert basin more than a kilometre below the global ocean, with rivers cutting deep canyons down toward it. The other holds that the basin stayed largely filled with dense brine, with the salt precipitating out of a deep hypersaline body of water rather than a dry pan.
The competing lines of evidence are real. Deep canyons cut by Messinian rivers, such as the buried gorge beneath the Nile and similar features under the Rhône, point to a low base level and support a major drawdown. Shallow-water fossils found in what should be the deep abyss also suggest exposure. Against that, the so-called Lago-Mare deposits found around the coastline look like fresher-water sediments and read more like a filled basin.
A 2024 study led by Giovanni Aloisi, published in Nature Communications, used chlorine isotopes in the salt to argue for two phases: an initial 35,000-year period of salt deposition confined to the eastern Mediterranean while the basin was still brine-filled, followed by a much shorter drawdown of less than 10,000 years during which level fell on the order of 1.7 to 2.1 kilometres in places. That work supports a substantial fall, while showing the picture is staged rather than a single simple drying.
A 2025 paper in Science Advances, from a group including Daniel Garcia-Castellanos, complicates it further by arguing the level changes through the crisis were themselves oscillatory, driven by river erosion and climate. The surface the eventual flood met may not have been a uniformly dry floor but a partially water-filled system at varying depths.
Where the ten metres a day comes from
The refill figure traces to a specific paper: Garcia-Castellanos and colleagues, in Nature in 2009. They modelled the Zanclean flood, the refilling event named for the Zanclean age that opens the Pliocene, using a 200-kilometre erosion channel identified in borehole and seismic data across the Gibraltar Strait. The model treats the flood as a feedback loop, where flowing water cuts the sill deeper, which lets more water through, which cuts faster again.
The numbers that come out are large. Peak discharge on the order of 108 cubic metres per second, roughly a thousand times the present Amazon. Incision of the channel floor above 0.4 metres per day. And, as a consequence of that discharge filling the basin, a sea level rise inside the Mediterranean exceeding ten metres a day at the peak. In the same model, as much as 90 per cent of the water transferred in a window of a few months to two years, after a slower trickle that may have run for thousands of years beforehand.
Every one of those figures is a model output constrained by the geometry of the eroded channel, not a directly observed rate. The 2009 paper says as much, noting that the flood’s nature and abruptness were poorly constrained. The erosion channel is real and measurable. The discharge and the daily rise are what the incision model implies given that channel. A 2020 review in Earth-Science Reviews, led by Garcia-Castellanos, gathered the independent evidence and largely held to a single-stage flood with a multi-month main pulse, but the fastest end of the timing range remains a model preference rather than a measured fact.
What the figure is good for
Stated plainly, the strongest claims here rest on different footings. The salt is observed. A large sea-level fall is well supported, with its exact depth and staging still under active study. The catastrophic refill is a model result, internally consistent and grounded in a real eroded channel, but an inference about rates that no instrument recorded.
The reason the flood scenario has held up is that it explains a feature that is otherwise awkward: a continuous 200-kilometre incision running across the strait and well out into the Atlantic, deeper than ordinary river erosion during the dry phase seems able to account for. A single overspilling flood that cut its own channel as it ran is a coherent explanation for that geometry. That is a different kind of support than a measurement, and it is the kind most of deep-time geology has to work with.
What to watch
The open questions are specific rather than vague. How dry the basin actually got, and whether the desiccation and the Lago-Mare evidence can be reconciled into one timeline, is the live argument, and the chlorine-isotope and erosion-climate work of the last two years is where it is moving. The refill rate is unlikely to be pinned down more tightly than a model can manage, since the event left a channel and a salt layer but no record of how fast the water rose on any given day. The next advances are likelier to come from the drawdown side, where new geochemical proxies can still be brought to bear on rock that is sitting in cores.
