5.96 million years ago (Late Messinian), when the connection between the Mediterranean Sea and the Atlantic Ocean was restricted, the entire Mediterranean became hypersaline, its marine fauna extinguished, and a kilometers-thick evaporite sequence was deposited on its floor1,2 within a short period of ~640 ky3,4. This Messinian Salinity Crisis (MSC) is an exceptional event in Earth’s history. The MSC sea-level fall is strongly debated for >50 years; with high-end estimates of >1500 m1,5,6,7,8, intermediate estimates of 800–1500 m29,10,11,12,13,14,15,or 630–900 m16, and even minor estimates of 200–500 m17,18. These estimates are ingredients of the applied depositional models: from a nearly desiccated Mediterranean, with salt deposited in residual shallow hypersaline lakes1,2,5, to a deep hypersaline-water body accumulating salt on its thousands-of-meters-deep floor17,18,19,20.

A leading evidence for a large MSC sea-level drop is the upstream canyon incision of rivers (e.g., the Nile and Rhone), buried under thick sequences of sediment2,6,21,22,23,24,25. Elevation differences along (~2 km) and across (~1.5 km) the buried Nile canyon are considered as a measure for base-level drop23 ignoring the possibility that at least a part of the canyon may have been subaqueous; moreover the thick Nile delta sediments also changed the vertical position of the buried canyon.

Here, we reconstruct the original elevation and shape of the Messinian Nile canyon, by correcting the buried Messinian erosion surface (MES6) for post-Messinian subsidence due to sedimentary loading (isostasy, lithosphere bending, and decompaction of underlying sediments). The results are the Messinian landscape immediately after its reflooding and before any Pliocene sedimentation (ignoring a few tens of m difference between the Early Pliocene and the present sea level26).

On the reconstructed landscape, we identify the geomorphologic base level of the MSC-Nile River, where the reconstructed river profile attains a subhorizontal slope. The paleo-shoreline should be along this relatively flat region, possibly where river profile steepens again (exact horizontal location within the subhorizontal region is of little relevance for restoration of vertical isostatic motions). With unloading the water above the base level, we restore the pre-flooding Messinian landscape, i.e, during maximal drawdown.

Finally, we show that our conclusion about limited drawdown is robust in spite of the uncertainties related to the restoration process, although we cannot exclude that the sea was temporarily at different levels during the drawdown.

Geological observations

Near the present-day coast, the Messinian Nile canyon is buried 3000 m under the flat delta plain (Fig. 1a). About 100 km to the south, Near the city of Cairo, the canyon base is ~1500 m below the surface and its shoulders nearly reach the surface23. Farther south, beyond the seismically-mapped plain, the canyon size and depth are uncertain. Approximately 1000 km south of Cairo, near Aswan, Early Pliocene brackish ostracods at the base of a 200 m deep canyon fill27,28 raised the hypothesis that an Early Pliocene marine invasion of the canyon had reached Aswan, portraying the boundaries of the earlier MSC incision ~1200 km away from the present day coast24,29. However, fauna reexamination30,31 showed a poor assemblage of rare and non-marine ostracods with a wide age range. This fauna could have been reworked from Cretaceous or Paleogene outcrops, as observed in many Pliocene sediments along the Nile Valley31; i.e., the near-Aswan canyon could have been subaerially excavated and buried before or after the MSC.

Fig. 1: The expected influence of the Nile delta load on the adjacent Messinian canyon.
figure 1

a Present-day topography and bathymetry. Red line outlines buried canyon walls23. Yellow lines are the 2000 and 3000 m50 thickness contour of the Pliocene-Quaternary section showing that the main sedimentary load is located north of the canyon. b Schematic illustration of the difference between 2D and 3D restoration of the original vertical position of the Messinian canyon. Upper panel shows the present day buried canyon under Pliocene-Quaternary rocks. Left panel illustrates that in 2D analysis, canyon remains fully below present day sea level. Right panel illustrates that 3D analysis, considering the off plane load, is expected to provide higher canyon shoulders. h - the height difference between canyon thalweg and shoulders. hr - the same after sediment unloading (hr < h). e - elevation of canyon shoulders asl. d - water depth in the submerged part of the canyon. The inferred sea level drop is Δsea level = hr − e − d.

Examining stratigraphic data along the Nile Valley may clarify the uncertain inland extent of the MSC Nile canyon. Near Helwan, ~15 km south of Cairo, a water well reached the bottom of the canyon at 550 m32,33 below sea level (mbsl) and a few km farther south another water well reached the bottom of the canyon at 200 mbsl29. This abrupt depth change from 1200 m to 200–550 along only 15 km, indicates a paleo-knickzone, separating a deep gorge from a much shallower upper valley. Moreover, this knickzone approximately coincides with an Early Pliocene sedimentologic transition in the canyon fill from deep-water claystone34 to shallow marine and tidal limestones that continue ~150 km farther south until Beni-Suef34,35,36,37,38. Farther inland, Pliocene exposures consist of fluvial-lacustrine sediments31,34,39,40. These sedimentological data indicate that the early Pliocene marine invasion filled a deep gorge up to the Cairo-Helwan region, continuing ~150 km southward as a narrow and shallow marine embayment. From there, non-marine lacustrine environments extended farther inland.

Flexure produced by the Nile delta isostatic load

Most MSC flexural studies14,41,42,43,44,45 are initiated with pre-MSC relief and forward model the vertical motions caused by deposition and erosion of pre-assumed masses of salt, water, and eroded material. Here we question the pre-assumed water mass and utilize a different approach; the present-day relief is the reference and we backward model (restore) the topography immediately after the MSC, by isostatically unloading sediments that accumulated since. Similar exercise15, conducted for the Po Plain and the north Adriatic Sea, concluded that an 800-900 m drawdown best explains buried erosional features; and the same approach applied for the Western Mediterranean yielded 1100–1500 m drawdown46. The advantage in backward modeling is the well-known current relief and the unloaded sediments mass from seismic and well data. The source of uncertainty in the tectonic correction, which could be large15 particularly in tectonically active areas. Sicily, the Apennines, Crete, and Cyprus have uplifted thousands of meters since the end of the Miocene; the Ionian, Herodotus, and northern Levant basins subsided hundreds or thousands of meters towards nearby subduction zones47,48,49. Fortunately, northern Egypt is far from any plate boundary with minimal tectonic activity. The continuous flow of River Nile along thousands of kilometers without significant post-Messinian incision, excludes the possibility of regional uplift. On the other hand, the lack of post-Messinian sedimentation outside of the Nile Valley indicates negligible regional subsidence. Moving offshore, the closest location of post-Messinian tectonic activity is observed 150–200 km north of the Messinian canyon along the Rosetta and Temsah Fault systems50. We estimate that this remote tectonic activity had minor influence on the restored canyon.

Taking advantage of the minor post-MSC tectonic activity in the Nile delta, we focus on the influence of the thick Nile delta load (2000 m and 3000 m contours, Fig. 1a) on the paleo-Nile canyon located > 100 km to the south. 2D-treatment of this problem44,45 provides uplifted canyon shoulders, but still below sea level (Fig. 1b). However, we hypothesize that a 3D-calculation considering a ~4 km-thick off-plane load should yields much higher canyon shoulders (e Fig. 1b). Our analysis indicates that the shoulders of the Messinian Nile canyon were significantly higher than the “normal” sea level before and after drawdown (e in Fig. 1b); due to the delta-load induced flexure, this positive topography was lowered after the Messinian. Without correcting for this pre-flexure topography, the magnitude of the sea-level drop is overestimated (Fig. 1b).

Geological markers for the fallen sea level

Another reason for overestimating the magnitude of the MSC drawdown is ignoring the possibility that the downstream portion of the canyon may have been, at least, partly submerged below the fallen sea level (d in Fig. 1b). The challenge is to recognize indicators of the fallen MSC sea-level, including 1) Changes in the slope of the river thalweg as indicators of land-sea transition45. 2) Sedimentary facies characterizing continental-to-marine or shelf-to-slope transitions15,51,52. 3) Buried scarps of shoreline terraces10,11,13. 4) Flat truncations indicating subaerial erosion11. 5) Geomorphic erosional surfaces/features distinguishing between subaerial and submarine channels16.

Here we combine two observations constraining the amplitude of sea-level drop: a) the knickzone height generated by the upstream-migrating Nile incision, which constrains the minimal (partial) drawdown, excluding the downstream part of the valley. b) The transition from a sub-horizontal to a steeper river profile marking the transition from fluvial to submarine flow. Such a transition is observed in rivers continuously extending on-to-offshore as submarine canyons; e.g., the Gauping River-Canyon53 (southwest Taiwan) and the LIobregat River-Foix Canyon54 (Spain), with nearly horizontal thalweg approaching the coastline or shelf edge and much steeper thalweg offshore. Over geological time scales, near their base level and mouth, river profiles flatten. Therefore, the near-horizontal river profile and its slope break with the submarine canyon indicate the presence of a regional base level.


Restoring the original topography

Restoring the original topography requires the Pliocene-Quaternary thickness (Fig. 2a), and involves calculating the subsidence due to compaction of pre-Pliocene sediments (Fig. 2b) and the flexural deflection due to the weight of the Pliocene-Quaternary sediment (Fig. 2c) and the evaporated water layer (Fig. 2d). Figure 2e, f present the observed and restored base Pliocene surface. Compaction maps calculated for various lithologies and deflection maps with varying elastic thickness and/or water layers are in Supplementary Fig. S1.

Fig. 2: Maps illustrating how the Messinian erosional surface (MES ~ base Pliocene) is restored to its original vertical position.
figure 2

a Thickness of rocks covering the base Pliocene surface. b Calculated subsidence due to compaction of pre-Pliocene sediments assuming shale lithology (for other lithologies see Supplementary Fig. S1). c Calculated flexural deflection due to unloading of post-Messinian rocks, using elastic thickness, Te = 30 km (for other Te values see Supplementary Fig. S2). d Calculated flexural deflection due to unloading of a 650-m-thick water layer (Te = 30 km, for other values of water thickness and Te see Supplementary Fig. S3). e Observed base Pliocene structural map50 (modified). f Restored base Pliocene surface after correcting for compaction (shale) and deflection (Te = 30 km) due to sediment unloading. No correction for water unloading has been applied here.

The restored landscape is a few hundred meters above the pre- and post-MSC sea level (Fig. 3b); its present depth below the flat delta plain expresses the post-Messinian subsidence due to flexural bending of the lithosphere and compaction of pre-Pliocene sediments (Fig. 2a, b). The canyon’s longitudinal profile (Fig. 4a) illustrates that this upward restoration reached ~750 m near Cairo and ~2000 m at the base of the paleo-continental slope. Confirming our hypotheses, the restoration quantifies the elevated canyon shoulders (e in Fig. 1b) and the original canyon depth (hr < h in Fig. 1b). For example, 60 km north of Cairo (Section 4, Fig. 4d), the restored shoulders are ~300 m above the pre- and post-MSC sea level, and the shoulder-to-thalweg relief drops from 1500 m in the buried surface to 1000 m in the restored landscape.

Fig. 3: 3D view of the Nile delta and the buried Messinian canyon illustrating landscape evolution being restored.
figure 3

a Base Pliocene structural map (= MES). b Restored topography for the earliest Pliocene immediately after sea level recovery and prior to any Pliocene deposition (shale parameters for decompaction and Te = 30 km for deflection, same as Fig. 2 f). The isobath 650 m represents the fallen MSC shoreline. c Present-day topography. Black lines mark locations of cross sections (Fig. 4). Part of the area annotated as “flat delta plain” in the present topography (c) was a few hundred m asl in the earliest Pliocene (b).

Fig. 4: Topography restoration illustrated in cross sections (location in Fig. 3c).
figure 4

a Along thalweg river profile illustrating the combined effect of sediment decompaction (shale parameters) and flexural deflection. Error bars represent the range of uncertainty in the decompaction (shale, sand, shaly-sand) and elastic thickness (Te = 20, 30, 40, and 50 km). b Interpretation of the restored river profile. The Cairo-Helwan knickzone is inferred as the southernmost location of the Messinian retrogressive incision, separating the MSC canyon gorge from the pre-MSC hanging valley. The transition from sub-horizontal to steeper gradient marks the MSC coast line 650 m below the present day sea level. c Restored profile across the subaerial segment of the canyon with incision down to 650 m below the pre- and post-MSC sea level. d Restored profile across the subaqueous part of the canyon.

The longitudinal river profile further illustrates that the height of the Cairo-Helwan knickzone is reduced from 650 m in the buried profile to 400 m in the restored profile (Fig. 4a). Downstream, the profile flattens to near horizontal with ± 100 m elevation differences for the 20 < Te < 50 km range of elastic thicknesses. This nearly horizontal segment extends ~40 km downstream to a slope break (knickpoint), where the concave profile abruptly steepens and becomes convex.

We interpret the Cairo-Helwan knickzone as the southernmost location of the Messinian retrogressive incision, separating the MSC canyon gorge from the pre-MSC upstream valley. This knickzone most probably was characterized by a series of high bedrock waterfalls55 held by hard Mesozoic carbonates23), accommodating most of the sea-level drop. From the base of this inferred knickzone, the river slope gradually decreased towards the inferred base level (Fig. 4a). The concave-convex knickpoint is interpreted as the coastline, where the subaerial river flow was transformed into a submarine turbiditic flow that formed a subaqueous canyon.

The depth of the inferred MSC coastline is ~650 mbsl for all curves calculated with elastic thickness 20 < Te < 50 km. Differences of 100–200 m between Te = 20 and Te = 50 km are observed at the far ends of the profile (Figs. 4a and 5a). This restoration is calculated for a filled Mediterranean, a short moment after it returned to “normal” (pre- and post-MSC) sea level (the absence of large glaciations at the time allows to assume Zanclean sea level close to present26). To restore the topography immediately before re-flooding, we unloaded a 650 m-thick water layer, generating an isostatic rebound that brings the MSC coastline to 600 mbsl (Fig. 4b) and, thus, conclude that a 600 m drawdown best explains the observations. Considering that sea level of a closed basins commonly fluctuates, we cannot rule out short episodes of lower or higher sea level. However, the observed sub-horizontal segment of the river profile indicates relative stability of the base level around 600 mbls for a duration that is long enough to produce such a profile.

Fig. 5: Uncertainties in river profile restoration.
figure 5

a Calculated subsidence during burial of pre-Pliocene sediments using compaction parameters of shale, sand, and shaly-sand. Calculated deflection due to unloading of post-Messinian sediments using elastic thickness Te = 20, 30, 40, and 50 km. b Restoration based on the assumption that pre-Pliocene sediments are mostly shales (same as Fig. 4a). c Restoration based on the assumption that pre-Pliocene sediments are mostly shaly sand. The difference between b and c is 100 m in the sea level restoration (650 m versus 750 m, respectively).

Uncertainties related to decompaction and deflection

The unknown lithology and spatiotemporal distribution of the sediments buried under the Pliocene-Quaternary section, produces an uncertainty related to the parameters used for the decompaction correction. To illustrate the range of this uncertainty, we used parameters of three lithologies: sand, shale, and shaley sand (Fig. S1). Considering that at least since the Late Eocene, huge amounts of clastics started arriving to the Egyptian continental margin, which accumulated mainly shales29,56,57,58,59, we corrected for decompaction using shale parameters and reached the conclusion of 600 m drawdown (Fig. 5b). In addition, we demonstrate that using shaly sand lithology for the pre-Pliocene sediments, the amplitude of the reconstructed drawdown may increase to 750 m (Fig. 5c).

Another source of uncertainty is related to the elastic thickness used for the deflection correction. Previous studies dealing with lithospheric strength in the studied region argued that rigidity increases along with decreasing crustal thickness60. Te = 45 km was used for the Levant margin61, Te = 25 km for the Arabian plate61, Te = 30 km for the Levant Basin, assuming that this is approximately the depth of the 3500 isotherm62, and Te of 10–35 km for the Levant continental margin63.

In light of this uncertainty, here we calculated the deflection four times using Te = 20, 30, 40, and 50 km. We show that this range significantly ( ±400 m) influences the results in the central basin (Fig. S2), but its influence on the restored coastline is smaller than 100 m (Fig. 5b, c).

Unloading of a 650 m thick water layer causes an isostatic rebound of 210 m in the deep basin and zero inland (Fig. S3, Fig. 4b). To further show this effect we recalculate water unloading for a 500 and 750 m thick water layer, for Te = 30 amd 40 km (Fig. S3). The effect of these variations on the restored river profile are negligible (Fig. S3).


The restored topography, during the time immediately after re-flooding and just before initial Pliocene deposition, indicates the existence of a deep-water embayment extending to the Cairo-Helwan knickzone, and a 200 m deep water body continuing farther inland (Figs. 2e and 3b). We argue that the deep embayment was excavated by Messinian incision whereas the upper valley south of the Cairo-Helwan knickzone, may have been shaped by pre-Messinian ancestors of River Nile and unaffected by the Messinian incision. This conclusion is consistent with the sedimentological data presented above indicating a transition from deep water deposits north of Cairo to shallow water and lacustrine and fluvial deposits in the Beni-Suef region and farther south along the Nile Valley; it disagrees with suggestions that the Messinian incision and subsequent marine invasion had reached Aswan.

Another indication for limited retrograde incision up to Cairo (~100 km) and not further upstream to Aswan (~1200 km) is the absence of a thick Messinian fan delta, which is expected in a case of enormous erosion along 1200 km. The Late Messinian Qawasim and Abu-Madi Formations, exceed thickness of 1000 m only in a limited (100 × 25 km) region at the mouth of the canyon23,29,64,65,66. Outside this region these formations rapidly thinn to a few 10 s or 100 s of m (<300 m) resembling distal deep water deposits29,67,68. Such a canyon fill with a relatively small lowsrtand fan is better reconciled with limited incision of the Messinian Nile canyon up to Cairo.

Messinian deposits of the Abu Madi and Qawasim Fm. overlying the MES offshore Egypt23 and the MSC Stage 3 deposits overlying the salt in the Levant Basin (100–200 m thick7,69), indicate that the major drawdown occurred during the second stage of the MSC alongside massive salt deposition in the deep basin. However, since the thickness of the post salt Messinian deposits (MSC stage 3) are negligible in relations to the Pliocene-Quaternary section, here we assume that the top salt and the base Pliocene surfaces approximately coincide. Thus, the restored landscape at the end of Stage 2 (5.55 Ma17) and at the end of Stage 3 (5.33 Ma17) are practically the same.

The 200 m depth of the pre-Messinian upper valley south of Helwan is difficult to explain, considering that Oligocene-Miocene eustatic drawdowns only reached a few tens of meters26. To reconcile this discrepancy geodynamically, ~150 m regional subsidence should be invoked, as proposed from modeling of mantle-flow-sourced topography70; such post-Messinian tectonic correction implies reduction of the amplitude of drawdown from 600 m to 450 m. On the other hand, the presence of Pliocene marine rocks in the upper Nile Valley, 50 m above sea level34, indicates that northern Egypt has not been subsiding significantly, at least since the Pliocene and, therefore, 450 m for drawdown, seems unrealistic to our opinion.

Another possibility is that the canyon depth near Helwan is not 550 mbsl32,33, as we used for modeling, but only 200 mbsl as observed in a water well, located a few km to the south29. Applying 200 mbsl in the model, the restored upper valley rises to ~100 mbsl; considering post-Messinian 50 m of tectonic subsidence, the base of the restored Miocene valley rises to 50 mbsl, which is within the range of eustatic falls. With such a tectonic correction (50 m), the inferred MSC drawdown is reduced from 600 to 550 m. It should be mentioned, however, that such details are within the uncertainties of the technique (e.g., sediment thickness, compaction, and elastic thickness) as illustrated in the Supplementary Figs. S1S3.

Within the range of 550–750 m drawdown, we prefer the value of 600 m, which fits the assumptions of shale dominated lithology for pre-Pliocene lithology and tectonic stability of the Egyptian continental margin. ~600 m drawdown is 2–4 times smaller than previous estimations deduced from the Nile canyon23,44,45. We claim that previous studies underestimated the post-Messinian subsidence by ignoring the compaction of the pre-Pliocene sediments and not considering the 3D-effect of flexure. For instance, the formerly45 estimated post-Messinian subsidence of 750–1000 m at the base of the continental slope, becomes ~2000 m when including decompaction and flexural isostasy. We further propose that base level drop and incision depth were overestimated by ignoring the uplifted canyon shoulders (e) and its subaqueous downstream portion (d) (Fig. 1b). For instance, Barber23 measured a > 2000 m relief along the buried gorge and associated it with base-level drop. Here, we show that after isostatic unloading, the river profile reveals a flat region consistent with a steady lowered base level during the MSC at about −650 m.

The difference between our result of a 600 m drawdown and the 800–900 m drawdown inferred from the Po Plain and north Adriatic Sea15 (same method) is within the range of uncertainty of the method. The discerepancy may therefore be related to 1) limitations of the technique (uncertainties are shown in Fig. 5 and Supplementary Figs. S1S3); 2) The possibility of disconnected Adriatic Sea from the Eastern Mediterranean, consequently having different base levels15, or 3) Involvement of difficult-to-correct tectonic motions in the Po Plain, located between the active Apennines and Alps. We emphasize again that, in contrast, the tectonically-stable Egyptian continental margin provide ideal conditions for restoration.

Finally, we cannot dismiss the possibility that geomorphological markers identified at various depths below the Mediterranean seabed7,8,11, represent different and possibly shorter-lived stages of the MSC. Our analysis is based on the incision along the Messinian Nile canyon, the largest erosional feature formed during the MSC and hence we interpret it as representative for the average sea level during most of the duration of the drawdown stage.


For unloading the post-Messinian (Pliocene-Quaternary) sediments from the base Pliocene surface, we first expanded the recently compiled base Pliocene map50 a few hundred kilometers south of Cairo. In practice this requires to reconstruct the depth of the buried canyon under the present-day Nile Valley. Outside the valley the thickness of post Messinian rocks is negligible. Based on the data presented above, we considered a linear gradient of 1/30 (1000 m over a distance of 30 km) between Cairo and Helwan and a much smaller gradient of 1/3333 (300 m per 1000 km) farther south. In this map the bottom of the canyon is 1200 m near Cairo, 550 m in Helwan32,33, and 400 m at the southern boundary of the map (Fig. 2a).

For decompaction of underlying sediment, we used a regional sediment thickness map71, which is >10 km deep in most of the study area. The decompaction results are insensitive to the accuracy of this map, because rocks deeper than ~5 km retain nearly negligible porosity15,72. The uncertainty related to a range of decompaction parameters (shale, sand, and shaly-sand) are presented in Fig. 5.

For flexural backstripping we apply an elastic thin-plate approach with a pseudo-3D (planform) flexural procedure15,46,73. For density of the post-Messinian sediments we use 2000 kg/m3, consistent with seismic and well log data71,72. For elastic thickness we use 20, 30, 40, and 50 km and discuss the associated uncertainty in the Supplementary Fig. S2. For decompaction of underlying sediment, we used a regional sediment thickness map74, which is >10 km deep in most of the study area. The decompaction results are insensitive to the accuracy of this map, because rocks deeper than ~5 km retain nearly negligible porosity15,72. The uncertainty related to a range of decompaction parameters (shale, sand, and shaly-sand) are shown in the Supplementary Fig. S1.