Mediterranean RO
Map of the dried-out Mediterranean basin, salt lakes in the deep pits, filling from Gibraltar

The interactive map cannot run on this device; the picture shows the basin before the flood.

The day the Mediterranean filled: the Atlantic breaks through Gibraltar into a dried-out sea

5.33 million years ago, according to the best-known hypothesis, the Mediterranean had dried down to a few salt lakes, up to 2.7 km below the ocean. The flood model published in Nature in 2009 by Daniel García-Castellanos and colleagues, one of the competing explanations, re-run here on today's sea floor in place of the old one: the gate opens, most of the water comes in within about a year and a half, and the basins fill one after another up to the coast you choose.

Scroll: the gate at Gibraltar opens.

A sea that evaporated

Between 5.97 and 5.33 million years ago, the Mediterranean's links to the Atlantic all but closed. The sun took more water than the rivers brought, and salt settled on the basin floors: about a million cubic kilometres, in layers up to 1.5 km thick in the west and 3.5 km in the east.

How far the water fell is still argued over: estimates run from a nearly full sea to a floor almost entirely dry, and the flood needs a drop of at least a few hundred metres. Rivers draining into the basin cut canyons: the Nile's reaches 2,500 m deep beneath its delta, the Rhône's about 1,000 m. The map starts from the levels of the 2009 model: lakes 2,500 m below the ocean in the west and 2,700 m in the east. White is salt; red is the brine left behind.

289 thousand km³ of brine stayed in the pits; the rest of the basin, 3.56 million km³, was empty.

The gate at Gibraltar

If the flood happened, today's strait is its scar. Before it, a bridge of dry land joined Spain and Morocco. The model starts from a brook: water one metre deep running over the sill, about 6 m³ a second.

The mechanism is a loop: the water wears the sill down, a lower sill lets more water through, and more water wears it faster. The equations are those of Daniel García-Castellanos and colleagues (Nature, 2009): erosion grows with the drag of the water on the bed, and the flow follows Manning's formula for rivers.

Day one: a brook falling towards a lake 2.5 km below.

The years when nothing showed

In the reference run the water trickles for 12.8 years before the flow passes a million m³ a second. After ten years about 7,000 m³ a second pass; then the cut deepens faster and faster.

This is the least certain part of the story. In the paper's simpler two-basin setup the western lake reaches the Sicily sill after 14 years; if the first trickle was 10 cm deep, after 47; with another erosion law, after 4.4 or 3,017. The erosion law and the first trickle change the fast part far less: in the paper, 90% of the water still enters within 510–790 days.

The first lake forms in these very years: the Alborán pit, next to the strait, takes Atlantic water from year 10, day 51 on.

Year 13: the flow passes a million m³ a second.

The western basin fills

The water runs down a ramp of 1–4% towards the Alborán pit between Spain and Morocco. Once full, the pit spills towards the Algerian lake over a sill now 1,302 m deep.

At about 1,920 m below today's sea level, the Algerian lake tops the sill between Sardinia and Tunisia and joins the Tyrrhenian lake. The western basin becomes a single lake.

The day the Tyrrhenian lake joins the Algerian one: year 14, day 244.

The Sicilian cascade

When the western lake reaches the sill between Sicily and Tunisia, in year 15, day 64, water starts falling into the eastern basin, nearly 2.4 km lower. Over the following days the eastern lakes rise by up to 18.0 m a day.

Here, on the Malta Escarpment, Aaron Micallef and colleagues (2018) found traces they attribute to the flood: the Noto Canyon, about 27 km long and 1.5 km deep, with an amphitheatre head 6 km wide and 700 m high, and a chaotic body of 1,430–1,620 km³ of sediment at the foot of the escarpment, compatible with a flood deposit.

From west to east over the Malta Escarpment: a drop of nearly 2.4 km, with the 2009 levels.

The east, the Adriatic, the Aegean

The eastern basin has almost twice the area of the western one. The water fills the Ionian and Levantine pits first, then enters the Sea of Crete and the southern Adriatic, crossing the Otranto sill in year 15, day 272. Water first enters the northern Adriatic in year 16, day 55 and comes within 20 km of Venice only in year 16, day 170.

The flow peaks when the eastern basin also reaches the Sicily sill. From then on the two lakes rise together towards the ocean.

The day the east reaches the Sicily sill: year 15, day 328.

The peak

Near the end of year 15, the Gibraltar sill had been cut by more than 200 m. Through a gap 13 km wide ran 116 million m³ a second; about 11 months earlier, as the western lake neared the Sicily sill, the water had reached 55 m/s. The Amazon pours on average 206,000 m³ a second into the ocean: the flood carried about 560 times as much.

The western lake had risen by up to 14.9 m a day. In the model, the cut at Gibraltar reaches 226 m; from the Gulf of Cádiz to the Algerian basin, geophysicists have mapped an erosion channel about 390 km long.

116 million m³/s: about 560 Amazons.

The last metres

As the lake nears the ocean's level the slope vanishes and the water comes in ever more slowly. 90% of the water entered in 539 days; the last few metres took months.

From the first trickle to a full sea: 16.0 years in the reference run. The 2009 paper says 90% of the water was transferred in a period "ranging from a few months to two years".

539 days for 90% of the water.

Your coast

Tap the map near any shore or pick a city. See how far and how deep the salt lake lay when the flood began, the day Atlantic water first stood within 20 km of today's shore, and how fast it was rising then.

Pick a city · Tap a shore on the map

The towns, in the order the water reached them (2009 levels)
TownSalt lake at the startWater within 20 kmRise per day
Algiers27 km · −2,500 myear 14, day 3018.1 m
Nice56 km · −2,500 myear 14, day 3228.9 m
Palermo73 km · −2,500 myear 14, day 33610.7 m
Oran35 km · −2,500 myear 14, day 34310.7 m
Genoa75 km · −2,500 myear 14, day 35913.6 m
Ajaccio44 km · −2,500 myear 14, day 36113.9 m
Barcelona139 km · −2,500 myear 15, day 567.0 m
Rhodes26 km · −2,700 myear 15, day 1129.6 m
Beirut252 km · −2,700 myear 15, day 1848.0 m
Catania77 km · −2,700 myear 15, day 2289.1 m

Your town, then.

How sure is the flood?

The flood is a hypothesis backed by traces, not an observed fact. For it: the channel of about 390 km from the Gulf of Cádiz to the Algerian basin, which removed about 1,000 km³ of rock; the Noto Canyon and the chaotic deposit in the Ionian Sea; sand in ODP boreholes at the Miocene–Pliocene boundary, which a sediment-transport model finds consistent with a flow of about 100 million m³/s. And the sediments at the boundary record an exceptionally rapid return of a deep sea.

Against it: some researchers think the salt formed in a deep sea filled close to today's level (Lugli and colleagues, 2015), and put the erosion of the margins down to dense-water currents (Roveri and colleagues, 2014). The main objection to a kilometre-scale drop right before the Pliocene is the Lago-Mare deposits, with fresh- and brackish-water fauna, found near today's sea level. Others suggest the strait may never have closed completely, that the water came in earlier, 5.46 million years ago (Do Couto and colleagues, 2014), or that the crisis ended through a "pump" with no flood at all (Marzocchi and colleagues, 2016). Even García-Castellanos's own 2025 model gives the east a drop of under 1 km before the flood, which, the authors write, conflicts with the idea that the flood cut the Noto Canyon. García-Castellanos and colleagues wrote in 2020 that a large drawdown "is currently under strong debate and far from a consensus"; the outcome of three drilling proposals (DEMISE, DREAM, IMMAGE) will, they write, be crucial.

Switch the starting level on the map. With the 2009 lakes, the same physics gives 116 million m³/s and deepens the sill by 226 m; with lakes at −1,200/−1,800 m, 63 million m³/s and 193 m. Starting from −222 m, only 1.0 million m³/s, 157 years of flow and a cut of 86 m, less than half the one mapped under the strait. In this model, the deep channel needs a low sea.

Starting level:

The same physics, other assumptions, in the two-basin setup: the re-run here against the paper
AssumptionYears to the Sicily sill90% of the water (days)The paper, 2009
Reference (a = 1.5)14.146214.0 · 514
Another erosion law (a = 1)4.45044.2 · 510
Another erosion law (a = 3)3,0174603,072 · 790
First trickle 10 cm, not 1 m47.146247.1 · 514
Rock eroding like a river bed21812,275216 · 15,100
A 50 km ramp, not 100 km5.01545.3 · 168

These rows use today's basin split into west and east, as the paper does, so they differ from the run on the map (539 days for 90% of the water). The slow years come within 6% of the paper's; the peak and the length of the torrent differ more, by up to 43%, probably because the paper used a reconstruction of the Mediterranean floor that is not public.

Meanwhile, at the Iron Gates

The Mediterranean was not alone. To the north, the Paratethys, ancestor of the Black Sea, and the Pannonian lake on the Hungarian plain had levels of their own. In a 2025 model by Daniel García-Castellanos, lead author of the 2009 paper, with Dan Palcu of GeoEcoMar in Bucharest and other colleagues, the Pannonian lake spills into the Paratethys over the threshold near today's Iron Gates, the Danube gorge on the Romania–Serbia border. As the water deepens its outlets, at the Iron Gates and along the Bosphorus–Aegean route, the northern lakes shrink, and water no longer evaporated there reaches the Mediterranean.

In the last wet phase before the flood, this inflow lifts the Mediterranean to −222 m; then comes a final drop that, the authors suggest, helped breach Gibraltar. The third starting level on the map asks what would have happened had the flood started from there. In the same model, the Paratethys lake stays about 20 to 50 m below today's sea level. On the sea that covered Transylvania more than eight million years earlier: The sea that disappeared.

Under your beach

The salt is still there, beneath the Mediterranean floor: a layer up to 1.5 km thick in the west and 3.5 km in the east. The drilling that found it, in the 1970s, gave rise to the whole hypothesis. Today's strait, with its two currents, would be the flood's cut: by the 2009 paper's reading, the flood's own erosion pushed the sill 30–80 km west.

Next time you wade into the Mediterranean, look out to sea: tens to hundreds of kilometres offshore the floor was a white expanse of salt, and the water holding you up once came through Gibraltar, in the model, at up to 55 metres a second.

The Strait of Gibraltar: today's two currents · The sea that disappeared

How it was calculated

The model: equations 1–4 of the Methods of García-Castellanos and colleagues (Nature, 2009), with their parameters: erosion proportional to the water's drag to the power 1.5 (k = 1.3×10⁻⁴), Manning's formula with n = 0.05, channel width 1.2·Q^0.5, slope equal to the head divided by 100 km (head capped at 1,000 m), initial sill 1 m, time steps of 0.1 day, sea-water density 1,025 kg/m³ (the paper does not give it). Evaporation and rain are ignored, as in the paper.

The sea floor: ETOPO 2022 (NOAA), the 60″ grid averaged to 2′. Today's bathymetry stands in for the floor of the time: sediment laid down since the flood, the isostatic rise of the empty basin and crustal movement are ignored. Meijer and Krijgsman (2005) find that, for the whole basin, the area–depth curves of today and of the late Miocene are "largely the same". On the 60″ grid the Sicily sill lies at −345 m (the paper uses 430 m) and the Camarinal sill at −321 m (measured: 284 m). Today's areas: 0.84 million km² in the west, 1.67 million in the east.

The filling: a tree of the basin's pits (each lake fills to its sill, then spills into its lowest neighbour), with water entering at Gibraltar; the western lake's level sets the head in the equations. A coast's arrival day: the first day on which, within 20 km of today's shore, the floor is under water that is at least half Atlantic. Distances are straight lines on the sphere.

What cannot be checked: the model was tuned to cut the sill as deep as geophysics shows (240 m), so the channel is not an independent test. The length of the slow years rests on arbitrary assumptions. The days on the map are the model's, not a record.

Sources

  1. García-Castellanos, D., Estrada, F., Jiménez-Munt, I., Gorini, C., Fernàndez, M., Vergés, J. & De Vicente, R. (2009). Catastrophic flood of the Mediterranean after the Messinian salinity crisis. Nature 462: 778–781 (Methods and Supplementary Information). doi:10.1038/nature08555.
  2. García-Castellanos, D., Micallef, A., Estrada, F., Camerlenghi, A., Ercilla, G., Periáñez, R. & Abril, J. M. (2020). The Zanclean megaflood of the Mediterranean – Searching for independent evidence. Earth-Science Reviews 201: 103061. doi:10.1016/j.earscirev.2019.103061.
  3. García-Castellanos, D. et al. (2025). Science Advances 11: eads9752 (Paratethys and Pannonian spillways, the Mediterranean at −222 m before the Miocene–Pliocene boundary). doi:10.1126/sciadv.ads9752.
  4. Micallef, A., Camerlenghi, A., García-Castellanos, D. et al. (2018). Evidence of the Zanclean megaflood in the eastern Mediterranean Basin. Scientific Reports 8: 1078. doi:10.1038/s41598-018-19446-3.
  5. Meijer, P. Th. & Krijgsman, W. (2005). A quantitative analysis of the desiccation and re-filling of the Mediterranean during the Messinian Salinity Crisis. Earth and Planetary Science Letters 240: 510–520 (evaporite volume and thickness, hypsometry).
  6. Van Couvering, J. A., Castradori, D., Cita, M. B., Hilgen, F. J. & Rio, D. (2000). The base of the Zanclean Stage and of the Pliocene Series. Episodes 23(3): 179–187.
  7. Callède, J. et al. (2010). Les apports en eau de l'Amazone à l'Océan Atlantique. Revue des sciences de l'eau 23(3): 247–273. doi:10.7202/044688ar (206,000 m³/s at the mouth, 1972–2003).
  8. Sannino, G., Pratt, L. & Carillo, A. (2009). Hydraulic criticality of the exchange flow through the Strait of Gibraltar. Journal of Physical Oceanography 39: 2779–2799 (Camarinal sill, 284 m).
  9. Alternatives and objections, as summarised in García-Castellanos et al. (2020): Lugli, S. et al. (2015), deep-basin evaporites; Roveri, M. et al. (2014). Dense shelf water cascading and Messinian canyons. American Journal of Science 314: 751–784; Do Couto, D. et al. (2014), Lago-Mare near Málaga and an earlier flooding at 5.46 Ma; Marzocchi, A. et al. (2016). Mediterranean outflow pump: an alternative mechanism for the Lago-mare and the end of the Messinian Salinity Crisis. Geology 44: 523–526.
  10. NOAA National Centers for Environmental Information (2022). ETOPO 2022 15 Arc-Second Global Relief Model (60 arc-second bedrock grid, subset 10° W–42° E, 29–47.5° N). doi:10.25921/fd45-gt74.
  11. GeoNames (cities15000), CC BY 4.0, for the coordinates of the towns in the list.