The formation of the Isthmus of Panama can be tracked through the signatures that it left on the surrounding biotic and abiotic environment and, more generally, on the biosphere. Current beliefs about the age of the final formation of the isthmus are mainly based on interpretation of geological evidence, which is not sufficient to reconstruct the palaeogeographical history of the area, given the difficulty of making direct geological observations and the fact that much of the geological record is eroded as a result of the complex geological history of the area61. Although biogeographical information can be useful in palaeogeography and has often been invoked in the past (e.g. ref.62), previous efforts to determine the palaeogeography of the Isthmus of Panama (e.g. ref.1) did not manage to separate a clear and detailed biogeographical signal. Here, I use all of the available phylogeographical information to infer a specific palaeogeographical scenario suggested by the data. I then consider the congruence between this scenario and the available geotectonic, biostratigraphic, oceanographic, and paleoclimatic data. Regarding the latter, I investigate whether the formation of the Isthmus of Panama is related to the occurrence of the great paleoclimatic change known as the MPT.
2.1. Phylogeographic data
Molecular divergence dates of transisthmian geminate species and populations have been widely used to infer conclusions regarding the formation of the Isthmus of Panama (e.g. ref.1). Such divergence dates can be considered reliable only if they are calibrated on independent calibration points rather than on a benchmark date for the formation of the isthmus, because the a priori assumption of a benchmark date for the formation of the isthmus leads to circularity if the goal is to examine the timing of dispersal or vicariance across the isthmus1. On the other hand, the characterisation of species or populations as “geminate” is itself subject to limitations and uncertainties, which can influence the phylogeographic conclusions37. A proper selection of geminate species should make clear how the invoked speciation is related to the formation of the isthmus, because different forms of life have different capabilities to overpass specific biogeographical barriers, so their biogeographical significance in paleogeographic reconstructions varies accordingly.
In the present work, I categorise the phylogeographical data by qualitatively evaluating its importance and clarity regarding the general concept of a recent isthmus formation as well as the exposure of specific transisthmian seaways. The formation of the Isthmus of Panama is assumed to have predominantly caused vicariant speciation of marine biota and terrestrial biota that dwell exclusively in coastal areas and are unable to cross inland features such as hills, mountains, and forests. Geminate coastal species of this category (Category A) are currently distributed on either side (Atlantic and Pacific) of the isthmus, because the inland of the isthmus developed a physical barrier to interbreeding between Atlantic and Pacific populations of the common ancestor and thus led to vicariant speciation (see the graphical example in Supplementary Fig. 1).
Squirrel monkeys (genus Saimiri) currently inhabit each of the areas of vertebrate endemism in the Amazon (i.e. the Atlantic side of South America), but one species, Saimiri oerstedii, exists only in a small coastal area on the Pacific side of Central America42. The restricted distribution of S. oerstedii is accredited to specialisation to a lowland coastal niche, which is thought to have been a characteristic of the last common ancestor of S. oerstedii and the other Saimiri species42. The divergence between S. oerstedii and Saimiri sciureus, which is found in the northern Amazon, is estimated to have occurred ~ 0.95 Mya42. Sea birds are another example of coastal organisms for which inland areas present a barrier63. The most recent common ancestor of regional populations of masked boobies (Sula dactylatra) on either side of the Isthmus of Panama is estimated to have lived ~ 0.64 Mya43.
Mangrove coastal vegetation is often considered to harbour the signatures of historical vicariance events64. It has been argued that because mangrove or coastal habitats were the last to disappear during the final closure of the CAS, the species that inhabited those environments were probably the last to be separated by the closure event65 (Supplementary Fig. 1). Therefore, the divergence times of mangrove species should correspond accurately to the final completion of the isthmus65. The Pacific and Atlantic populations of the black mangrove (Avicennia germinans) are estimated to have evolved independently since ~ 0.84 Mya44. Equivalent divergence dates have been reported for the mangrove-inhabiting fishes of the transisthmian genera Dormitator33, Anisotremus1, and Mulloidichthys1 (Table 1).
With regards to marine biota vicariance, there are several shallow-water fishes, decapoda, arthropods, and bivalves dwelling on either side of the isthmus whose divergence times have been dated close to, or since, 1 Mya using independently calibrated molecular clocks (Table 1). Most of these species are inhabitants of mangrove environments, suggesting a shallow-water property in the last transisthmian seaways. However, the existence of some deep-water species in Category A suggest that at least one of the seaways maintained a deep sill.
In addition to the examples in Category A, there are many other potential geminate species with molecular divergence dates that are comparable with a < 1 Mya age of the formation of the Isthmus of Panama, including plants, insects, reptiles, and birds (e.g. Supplementary Table 8 in ref.1). Invocation of these species as direct evidence of the age of the formation of the isthmus would require composite dispersal scenarios with a substantial degree of uncertainty, however, so they are counted here as indirect evidence. The most significant of them have been included in a separate Category (Category B) that includes species with distinct (Category B1; Supplementary Table 1) or overlapping (Category B2; Supplementary Table 2) distributions on both sides of the proposed former transisthmian seaways.
Detailed palaeogeographical information about the exposures of the transisthmian seaways in specific locations can be extracted from the phylogeography of terrestrial species with limited dispersal, that is, species (mainly endemic) that inhabit the same area for a long time. Accordingly, one can suppose that genetically distinct populations that today inhabit regions on both sides of former seaways will have molecular divergence dates that correspond to the last exposure of the seaways. Species in this category include freshwater dwellers such as fishes and frogs (Supplementary Table 3 and Supplementary Appendix). For example, genetic variants of the freshwater catfish Pimelodella chagresi with an estimated divergence time ~ 0.9 Mya66 are currently distributed on either side of the ancient locations of the Barú, Canal, and Atrato Seaways (see Supplementary Appendix). In addition, genetic variants of the mosquito A. albimanus67 can be used to help reconstruct the positions of the Early Pleistocene seaways just before the final formation of the Isthmus of Panama. The rationale for that approach is that as the relative sea level fell during the formation of the isthmus, saltwater swamps with mangrove vegetation would have developed in the emergent lowlands that were previously occupied by seaways, creating favourable conditions for the saline-tolerant A. albimanus to disperse across the isthmus. A. albimanus is one of the approximately 5% of mosquito species that live in either brackish or saline water, having adapted a highly developed system to regulate salinity68. The continuing sea level fall finally dried the saltwater swamps and separated the ancestral population of A. albimanus into isolated regions.
2.2. Geotectonic data
The hypothesis that the final formation of the Isthmus of Panama occurred ~ 3 Mya is summarized in the latest Neogene-Quaternary palaeogeographical reconstructions of Coates and Obando8 (Fig. 1B & 1C). According to that scenario8, interoceanic marine communication through the isthmus was maintained until the Pliocene via three persistent seaways (Fig. 1C): the Nicaraguan Seaway (through the Nicaragua Depression), the Canal Seaway (through the area of the current Panama Canal), and the Atrato Seaway (through the Atrato Basin of Colombia) (see also ref.69). Another seaway, here called the Barú Seaway, although shown in Coates and Obando’s Late Miocene (~ 6 Mya) palaeogeographical reconstruction (Fig. 1B), was considered to be subaerially exposed in their 3 Mya reconstruction (Fig. 1C). In the following sections, I review the available geotectonic data in order to determine whether recent developments in geoscience support this palaeogeographical scenario.
2.2.1. The Nicaraguan Seaway
The Nicaraguan Depression is a prominent, 40–70 km wide tectonic graben extending ~ 1000 km from the Caribbean side of Costa Rica in the southeast to the northern Gulf of Fonseca in El Salvador (Figs. 2 & 3). The depression has been interpreted as having probably developed since the Pliocene70,71. The most recent structural studies of the El Salvador Fault Zone show that the graben structures are the result of a two-phase evolution starting with an initial extensional phase that occurred between 7.2–6.1 Mya (latest Miocene) and 1.9–0.8 Mya (Early Pleistocene)72. At the northwest side of the Nicaraguan Seaway (sensu Coates and Obando8), the shallow-water El Salto Formation unconformably overlies older sequences (the Rivas, Brito, Masachapa, and El Fraile Formations) of various ages from Cretaceous to Pliocene71. While the fossil sites in Nicaragua have so far produced only land mammals of Late Pleistocene age, a baleen whale was found in the Mine K-11 locality within the El Salto Formation73. Pyroclastic-alluvial deposits of the Las Sierras Group succeed the marine deposits of the El Salto Formation, so the age of the El Salto deposition corresponds to the minimum age of the Nicaraguan Seaway at its northwest end. Based on molluscan fossil findings, an Early Pliocene age has long been assigned to the El Salto Formation; however, that age is considered to be poorly constrained (ref.73 and references therein). Correlation of offshore commercial wells in the Sandino basin and onshore stratigraphic studies showed74 that although the El Salto Formation includes the complete Pliocene, its top (sequence 12, SB12) reaches up to the early-to-mid Pleistocene Transition ~ 0.9 Mya. This age is congruent with the end of the second, transtensional phase of the formation of the Nicaraguan Depression (1.9–0.8 Ma; Early Pleistocene), implying that the associated extensional structures are contemporaneous with, and might have led to, the deposition of the El Salto Formation, which has a thickness of 100 m onshore and up to 1,000 m offshore.
The two large freshwater lakes, Lake Nicaragua and Lake Managua, contained in the Nicaraguan Depression are also considered to have been formatted since the Early Pleistocene75. Investigations of whether and when the lakes were last connected to the oceans did not reach a clear conclusion76. Nevertheless, the occurrence of marine-like nematodes such as Theristus setosus (Btitschli) Filipjev (an inhabitant of marine and brackish environments77), Polygastrophora octobulba (an inhabitant of marine and freshwater environments78), and the endemic Viscosia nicaraguensis (considered to originate from the marine Viscosia papillate79) in the lakes does not exclude the possibility that the lakes were once connected directly to the sea (ref.76 and references therein).
In any case, Coates and Obando8 independently showed that the Pacific seashore reached at least the middle of the Nicaraguan Seaway (Venado Formation, San Carlos Basin) during the Early Pleistocene (see also ref.80) and maybe had a southward connection via the Tempisque Basin as well (see map in ref.14,81). They further suggested that thereafter, during the Pliocene, the Nicaraguan Seaway was connected (although somehow restricted by the Sarapiquí Arch) to the Caribbean Sea through the Northern Limón Basin.
The Northern Limón Basin is an undeformed, normal faulted back arc basin that is separated from the Southern Limón Basin by the Moín High, and from the San Carlos Basin to the west by the Sarapiquí Arch80. Miocene regional uplifting of the inner arc formed a series of intermountain basins on the western, inland portion and shallow marine conditions on the eastern, Caribbean side. Subsequently, prograded fan deltaic, continental fan deposits, and patch reef development produced bay conditions corresponding to the Plio-Pleistocene Suretka Formation80. Such a transition is displayed in the classic stratigraphic scheme of the Limón Basin,82 where the Middle Miocene–Early Pliocene, deltaic shallow marine Río Banano Formation is unconformably succeeded by the shallow marine and continental rocks of the Plio-Pleistocene Suretka Formation (e.g. ref.83 and references therein). According to this very generalised stratigraphical interpretation, the conglomerates of the Suretka Formation demonstrate the southwards exposure limit of the Nicaraguan Seaway toward the Atlantic. This view is reproduced in the palaeogeographical reconstruction of Coates and Obando,8 which is based on a 2.5 Mya estimation for the age of the upper boundary of the Río Banano Formation84. That age cannot be assumed for the younger marine depositions in the Northern Limón Basin, however, because it is based entirely on cross-section data from the Southern Limón Basin,84 which experienced a different geodynamic history and deformation than the Northern Limón Basin and is much more poorly studied80,83. For example, although the Northern Limón and San Carlos Basins cover a 12.000 km2 onshore area, only three deep wells have been drilled on land,80 and outcrops are extremely rare85. On the other hand, the age of the Suretka Formation is unapproved, as no index fossil that dates a geological age was ever found86. Therefore, the stratigraphical data from the Southern Limón Basin are not representative of the palaeogeography of the Northern Limón Basin.
Recent geological research based on well and seismographic analyses showed that continuous marine Pleistocene deposits can be laterally traced for several tens of kilometres on the northwest-to-southeast coastline northwards of the Moín High83,87 and across the whole coast of the Northern Limón Basin (see ref. 80: Fig. 10). Therefore, given the potential for extended Pleistocene deposits in the Northern Limón Basin, the most recent age corresponding to the complete transoceanic exposure of the Nicaraguan Seaway is not defined by the conglomerates of the Suretka Formation, but instead by the Early Pleistocene deposits of the Venado Formation in the San Carlos Basin. Today, only 34 m of relief separates Caribbean and Pacific waters in the area of the San Carlos basin88, and it is likely that the area has been slightly uplifted under the tectonic effect of the Cocos Ridge subduction beneath Costa Rica during the Early Pleistocene.
2.2.2. The Barú Seaway
In the Late Miocene (~ 6 Mya) palaeogeographical reconstruction of Coates and Obando8 (Fig. 1B), the Atlantic Bocas del Toro Basin and the Pacific Burica Basin are connected by a marine trough, which I refer to as the Barú Seaway because of its location where the Barú volcano (3,474m) is today (Figs. 2 & 3). Also in the vicinity of the Barú Seaway is the Panama Fracture Zone, an active right lateral-moving transform fault that forms part of the tectonic boundary between the Cocos and Nazca Plates and the larger southeast-moving triple junction between the Cocos, Nazca, and Caribbean Plates89. As the oceanic crust of the Cocos Plate moved north eastward, it was partially subducted beneath the volcanic arc system of Costa Rica and western Panama, causing rapid elevation of the Central American Isthmus from the Arenal volcano in Costa Rica to El Valle in Panama. The result of this orogenic procedure is the mountain range of Cordillera de Talamanca (eastern Costa Rica) and Cordillera Central (western Panama)89,90, which cuts off the Barú Seaway and other interoceanic basins across the isthmus in the Late Pliocene (~ 3 Mya) reconstruction of Coates and Obando8, supporting the view of a complete formation of the Isthmus of Panama around the same time as the rise of the mountain range. Therefore, the exact age of the orogenic uplift due to the Cocos Plate collision is fundamental to scenarios of the final formation of the Isthmus of Panama.
The timing of Cocos Ridge subduction is currently one of the most widely discussed debates in Central American tectonics,91,92 with estimates ranging from as old as 8 Ma (e.g. ref.93) to as young as 0.5 Ma (e.g. ref.94). The youngest estimates for the Cocos Ridge arrival (0.5–3.5 Ma) at the Middle America Trench are derived from onshore rock uplift and subsidence patterns as well as plate reconstructions, whereas the oldest estimates (8–5 Ma) are largely based on the cessation of “normal” (non-adakitic) calc-alkaline volcanism within the Cordillera de Talamanca (ref.91 and references therein). Recent studies have concluded that the cessation of arc volcanism and the onset of Cocos Ridge collision (< 3 Ma) are separate events, however, reflecting recent changes in the configuration of the plate boundary system91,92. Hence, scenarios for a young age of Cocos Ridge subduction are increasingly favoured today.
Deposition facies of the Burica Peninsula on the Pacific side of the isthmus have been studied as a proxy for the tectonic movements of the Cocos Plate, because the peninsula is located at the eastern edge and front of the plate. The facies nomenclature used here is based on ref.95. The most comprehensive study of the area29 concluded that the marine faunal evidence contained in the Pliocene to Early Pleistocene (3.5–1.5 Mya) Burica Member of the Charco Azul Formation suggests strong subsidence and a bathyal environment (~ 2,000 m) in the area during the Early Pliocene, followed by gradual shoaling from 2,000 m to 1,400 m during the Late Pliocene through Early Pleistocene. Overlain strata assigned to the lowermost Armuelles Formation (late Early Pleistocene) were found to contain faunas indicative of water depths between 1,200 m and 1,300 m. By contrast, shallow-water molluscs in stratigraphically higher exposures of the Armuelles Formation are indicative of shelf deposits. Therefore, within the deposition interval of the Armuelles Formation, rapid shoaling took place at a time corresponding to the “Early Pleistocene–Late Pleistocene boundary”. These findings led Corrigan et al.29 to conclude that rapid uplift due to Cocos Plate subduction took place after the Early Pleistocene. In agreement with that view, Gardner et al.,94 using a radiometrically calibrated geodynamical model, concluded that the collision took place 0.5 Mya.
In contrast to these views, Collins et al.90 considered the whole Armuelles Formation to include very shallow-water deposits (< 10 m) and proposed that the rapid 2,000 m uplift in the Burica Peninsula falls within the interval of deposition of the underlain Burica Member of the Charco Azul Formation from the Late Pliocene. Furthermore, Collins et al.90 suggested that the subduction of the buoyant Cocos Ridge and the orogeny of the Cordillera de Talamanca began about 3.6 Ma, or 2 to 3 million years earlier than the date proposed by Corrigan et al.29 and Gardner et al.94. In particular, based on a ~ 1.6 Ma age for the top of the Moín Formation (i.e. the age of the most recent marine deposits in the Southern Limón Basin at the Atlantic coast to the north), Collins et al.90 concluded that the subaerial exposure of the Southern Limón Basin was caused by uplift from the delivery of the subducted Cocos Plate. Therefore, they placed the whole orogenesis within the interval 3.6–1.6 Mya. However, subsequent studies of the Southern Limón Basin revised the age of the basin uplift to < 1 Ma96, and more recent studies concluded that the whole Colón carbonate platform in the Bocas del Toro region of Panama (south of the Southern Limón Basin) was finally subaerially exposed during a Middle Pleistocene regional uplift97. Geological research in the Bocas del Toro region of Panama found that although neritic deposits persisted in the Northern region (Colón platform) until the end of the Early Pleistocene, there was a hiatus in deposition over a distance of ~ 80 km between Isla Colón and the Escudo de Veraguas Island (yellow line in Fig. 3) during the last 3.5 Myr and over an even greater distance within the last 1.8 Mya98. Therefore, the available data are not sufficient to refute the existence of a seaway with bathyal property during the hiatus interval.
In the middle of the proposed Barú Seaway, where the Barú volcano (3,474 m) stands today, there is independent evidence that low-elevation relief (< 500 m) may have persisted until as recently as the mid-Pleistocene, whereas rapid uplift took place thereafter21. Moreover, the age of the Barú volcano formation is considered to be ~ 0.5 My. It is therefore likely that the uplift that closed the Barú Seaway was a recent event that took place around the Early to Middle Pleistocene boundary (1–0.5 Mya). Indeed, most recent geotectonic analyses are congruent with the palaeobythometric interpretation of Corrigan et al.29 rather than the interpretation of Collins et al.90 (e.g. ref.92,95), suggesting that neither the Osa Peninsula nor the Burica Peninsula was emergent until at least 1 Ma, when the axis of the Cocos Ridge reached the Middle American Trench92.
2.2.3. The Canal Seaway
The Canal Basin (in a broad sense) is an elongate, northwest-southeast–trending transisthmian sedimentary trough controlled by a fault system known as the Canal Discontinuity,99 which extends across the isthmus within an 80-km wide zone (Figs. 2 & 3). Today, it results in a major topographic discontinuity between high volcanic mountain ranges to the east and west and lowlands punctuated by small (generally < 300 m high) topographic features of unclear tectonic and/or volcanic origins100. It remains uncertain whether the topographic highs in the Canal Basin are erosional remnants of ancient volcanic landforms that could have obstructed an interoceanic strait or instead of islands within a more recent interoceanic seaway101.
In the simplest terms, the geology of the Canal Basin consists of a pre-middle Eocene volcanic basement overlain by late Eocene to Late Miocene marine deposits interbedded with volcanic and volcaniclastic rocks102. Younger deposits are generally poorly exposed. The relatively recent scenario for the formation of the Isthmus of Panama depicted in the Pliocene palaeogeographical reconstruction of Coates and Obando8 is based on the assumption of a Pliocene age for the most recent marine deposits in the Canal Basin (i.e. those at the top of the Chagres Sandstone; e.g. ref. 61,103). In fact, the Chagres Sandstone, which conformably overlays the Late Miocene Gatún Formation, has been dated to be at least as old as the Late Miocene69 (see also ref.104). On the other hand, the only extended marine deposits that provide evidence of an interoceanic communication via the Canal Basin is the Early Miocene ‘‘La Boca Formation’’ (more recently re-interpreted as being the lower part of the Culebra Formation104). There is, however, certain other geological evidence suggesting that the Canal Basin might have experienced marine transgression episodes much more recently.
The Canal Basin is broken at the point where the isthmus reaches its lowest topographic elevation, originally ~ 84 m above sea level88, making it the best location for the construction of the interoceanic Panama Canal and also a possible location for the past exposure of a natural transoceanic seaway. The question is when the current topographical highs in the Canal Basin that currently separate the two oceans emerged. It is known that the Panama Arc began rising around 6 Ma and has continued to rise until the present day. On the other hand, the area maintained a low topography, which was certainly influenced by glacio-eustasy during the Quaternary (e.g. ref.7). The detailed chronological sequence of events is not well known, however, because of the complex geological history of the area and the difficulty of making direct geological observations61.
Within the Canal Zone, the most pronounced expression of low topography in the Canal watershed area is the once-extensive swamp within the broad valley of the Chagres river, which is now covered by the Panama Canal’s Gatún lake105,106. The valley penetrated to within 25 km of the Pacific Ocean in central Panama, where the Pacific–Caribbean drainage divide descends to one of its lowest elevations in Central America (< 200 m) in the low saddle of the Culebra Cut (also known as the Gaillard Cut) along the Panama Canal100. The saddle of the Culebra Cut marks the southern boundary of the Río Chagres basin, the geomorphology of which is the result of four continent-shaping movements that followed the cessation of intense volcanic activity in the Early Miocene and resulted in the erosive and depositional intervals that created the present day land mass105,107. During the first movement, the central portion of the isthmus was elevated above the coastal lines, developing the present morphology of the Central and Pacific portions of the isthmus. The second movement elevated the terrain to more than 90 m in the Atlantic area of the isthmus. Slow settling of the land surface characterised the third movement, during which the lower parts of the isthmus were overtaken by the sea, as evidenced by the layers of marine deposits with strictly fluvial beds in the “Atlantic mud” 107. The Atlantic mud and “Pacific mud” (also known also as sludges or more generally as Quaternary deposits) are informally known deposits with similar physical properties and appearance that unconformably overlay the Miocene marine formations of the Canal Basin61. The age of the muds is considered to be Holocene to Late Pleistocene61,103, which is supported by radiochronological estimations108. The muds are contained in swamp and stream deposits extending as far inland as Gamboa on the Caribbean side of the Canal Zone and as far as the Miraflores Locks on the Pacific side61, separated by the saddle of the Culebra Cut. The muds were deposited upon a stream-eroded topography of considerable relief105, so that any previous marine deposits in the area were eroded and lost. From this view, the uplift of the area around the saddle of the Culebra Cut might correspond to a more recent age in which there was a marine connection between the Atlantic and Pacific Oceans through the Canal Basin.
Seismic imaging along the Caribbean coast indicates that faults beneath the Limón Bay may be part of a more extensive set of predominantly north and northeast-trending faults, which are also exposed in the Culebra Cut between Gatún Lake and the Pacific coast106. Detailed geologic surveys conducted during the Culebra Cut widening project (1959–1969) revealed over 100 normal, reverse, and strike-slip faults along a single 1.8 km section of the cut. The geologic impression given by the surveys is that the area has been subjected to immense stresses and thoroughly shattered99. The faulting in the Canal Zone post-dates Late Miocene strata, and, although the minimum age of displacement is unconstrained, the available data imply a Pliocene or Quaternary age106. Indeed, after the cessation of folding in eastern Panama in Plio-Quaternary times81, the modern tectonics of the Canal Zone are dominated by strike-slip faults, such as the Río Gatún and Pedro Miguel faults (the latter is near and on the right of the Culebra Cut area), which were modelled by uplift and the emergence of central Panama within the last 3 Mya109. Geomorphic analysis suggests that the Pedro Miguel fault continues southward offshore into the Pacific Ocean, where Taboga Island may indicate an uplift at a left step on the fault110. In particular, on the basis of the slip-rate (4–7 mm/year) and total displacement (~ 5–10 km) along the Pedro Miguel fault, Farris et al.111 concluded that the fault must be younger than 1–3 Mya.
In conclusion, prominent geomorphic lineaments, topographic breaks, and bends in river courses are all consistent with a young, fault-controlled landscape100. Therefore, the available data cannot rule out an age < 1 Mya for the uplift of the topographical high separating the low topographies north and south of the Culebra Cut. Such a high could have constituted a terrestrial barrier to a recent interoceanic marine communication through the low topography of the Panama Canal area.
2.2.4. The Atrato Seaway
The Late Pliocene (~ 3 Mya) reconstruction of Coates and Obando8 displays another transoceanic seaway through the Atrato and Chucunaque Basins (Fig. 1C), the Atrato Seaway, which was originally proposed by Woodring112. Recent stratigraphical research showed, however, that the seaway passing through the Chucunaque Basin was subaerially exposed by 5.6 Mya81 (see also ref.113) and therefore no longer in existence during the Late Pliocene (Figs. 2 & 3). Based on stratigraphic data from a single well in the Atrato Basin (Opogado-1), Coates et al.81 postulated that the Atrato trough was also subaerially exposed by 4.8 Mya. This age corresponds to the minimum age of the marine deposits at the top of the Munguido Formation in the Opogado-1 well (for detailed biostratigraphy see ref.114,115). Coates and Obando8 and Coates et al.81 considered the Mongoido Formation to be the last marine formation in the Atrato Basin, but that assessment now appears to be incorrect. The Mongoido Formation is overlain by the Quibdó Formation in wells both north and south of the Opogado-1 well (see ref.116: Fig. 21), indicating that the stratigraphic scheme of the Opogado-1 well is not representative of the wider Atrato Basin area and therefore cannot be used to infer general palaeogeographical conclusions about the area.
The Quibdó Formation consists of sandstones and occasional conglomerates including marine faunas, and it is overlain by gravels, sands, and sandstones corresponding to alluvial deposits of the Atrato River117. On the basis of a Late Pliocene age for the Quibdó Formation, O’Dea et al.7 argued that the Atrato Seaway persisted until 3.1 Mya and was overlain thereafter by terrestrial sediments. The minimum age of the Quibdó Formation is therefore critical for bracketing the age of the exposure of the Atrato Seaway. Although the age of the Quibdó Formation is uncertain, it has been estimated to be Late Pliocene, but also Pliocene to “Quaternary (?)”, with a question mark indicating the possibility that the upper boundary falls within the Quaternary (e.g. ref.117,118). Other recent stratigraphical schemes display the minimum age of the Quibdó Formation to be ~ 1.8 Mya, albeit also accompanied by a question mark indicating that it might be even younger (e.g. ref.116: Fig. 20; ref.119: Fig. 17). Stratigraphical schemes in some technical reports show the Quibdó Formation to be Late Pliocene to Middle Pleistocene in age (e.g. ref.120,121). The confusion over the age of the Quibdó Formation stems from the purely marine biostratigraphic markers that have been recovered from it so far (see ref.119: Fig. 18). These are the benthic foraminifera Cassidulinella pliocenica and Ammonia cf. beccarii and the planktonic foraminifera Orbulina cf. universa (ref.119: Fig. 17). While the former is congruent with an age as young as the Late Pliocene122, the latter two are indicators of an age spanning from the Miocene to recent123,124. In the absence of well biostratigraphic markers, geotectonic data provide additional clues regarding the age of the Quibdó Formation.
The importance of the Quibdó Formation for geologists is not mainly stratigraphical but tectonical, as it is considered to describe the emergence of the Serranía de Baudó area, known as the Baudó Event116,119. The age (usually referred to as “8–4 Ma ?”) and the mechanisms responsible for the Baudó Event remain uncertain, because the available data are insufficient to fully explain the kinematics of the Baudó Range emplacement. In any case, the Baudó Event is believed to have led to the formation of the western margin and closure of the Atrato Basin125 (Fig. 2). Thus, while the eastern margin of the Atrato Basin records the shallowing of the basin, large deformations toward the western margin in the Baudó Range can be seen as a result of orogenic activity. A noteworthy example of such a dynamic procedure is printed in the Río Murrí section, where the Eocene Salaquí Formation outcrops in fault contact with the Quibdó Formation116,119. Late Miocene to Pliocene orogeny in the Northern Andes was likely triggered by the onset of a flat-slab subduction of the Nazca plate underneath the northernmost Andes of Colombia126. Thus, a Pliocene age of the Quibdó Formation matches well with the latest Neogene tectonic activity in the Northern Andes. It is also evident, however, that orogeny and uplift in the area occurred since the Early Pleistocene (1.8 Mya) until recently127. Therefore, the final configuration of the Serranía de Baudó area and subaerial conditions in the Atrato Seaway could be the results of a more recent procedure that occurred in two steps. In this context, the presence of Quibdó Formation deposits in the Salaquí river area may suggest a past westward marine connection of the Atrato Seaway to the Pacific via the northern Baudó mountain range, which was severed during the Early Pleistocene orogeny.
Northwards, a likely connection between the Atrato Seaway and the Atlantic Ocean passes through the Urabá Basin, which is separated from the Atrato Basin by the Mandé magmatic arc128. Detailed surveys of the region showed that the sediments in the Atrato Basin extend into the Urabá Basin7,128. In the Urabá Basin, based on the stratigraphic records of the onshore Apartadó-1 and Chigorodó-1 wells, as well as the interpretation of seismic lines, four seismostratigraphic sequences can be defined, ranging from the Lower Miocene to the Pliocene (?) (ref.119: Fig. 22; the question mark indicates that the data are purely biostratigraphic). Even the lithostratigraphic relations among the drilled units that outcrop on the west and east flanks of the basin are uncertain because of a lack of detailed stratigraphic studies in these areas119. Nevertheless, the stratigraphic record of the Necoclí-1 well on the right bank of the Urabá Gulf along with the related seismic line transect (see ref.129: Fig. 9 & ref.119: Fig. 27) reveal the geometry of the Urabá Basin and allow a precise biochonostratigraphic estimation for the facies of the whole foredeep basin. According to this idealised stratigraphic scheme129, the oldest recorded sedimentary rocks in the Urabá Basin are Lower Miocene deep-water facies, which grade to Pliocene–Early Pleistocene shallow-water siliciclastic deposits and are overlain by more recent alluvial sediments (ref.129: Fig. 5c). Therefore, an Early Pleistocene exposure of the Atrato Seaway from the Urabá Basin to the Pacific via a marine trough in the northern Baudó mountain range is congruent with the available data. Τhis seaway is not related to the findings of ref.,6 because it was located northwards, not southwards, of the Mandé Batholith (see ref.6: Fig. 3).
A likely second connection of the Atrato Seaway southwards to the Pacific passes from the adjacent San Juan Basin. There, the Mayorquín Formation can be correlated with the Quibdó Formation on the basis of similarity in age (e.g. ref.119). However, the Atrato River Valley to the north and the Pacific Plain to the south were separated by the uplift of the “San Juan Paleohigh,” resulting in the southernmost extension of the “Istmina Deformation Zone.” This uplift was the product of dextral tectonics from the collision of the Panama arch with the Colombia Pacific during the Late Neogene, but the precise age of the event is not well constrained130.
2.4. Palaeoceanographic and Stratigraphic Data
Oceanographic models support the hypothesis that the open CAS permitted relatively fresh and cool Pacific water to flow into the North Atlantic, affecting buoyancy by adding freshwater into the Caribbean and weakening the AMOC (e.g. ref.18). For example, a narrow (~ 100 km) but deep (~ 2,000 m) transisthmian seaway could exert a profound effect on the global oceanic circulation pattern18. The open/close modes of the transisthmian seaway(s) can be traced through changes in ocean salinity and the circulation patterns of ocean surface currents, as well as the circulation of bottom currents and their erosional effect on the ocean floor. Specifically, during periods when the transisthmian seaways were open, low-salinity Pacific waters could penetrate into the high-salinity Caribbean waters, reducing the intensity of the Gulf Stream and the North Atlantic Current on the surface of the Atlantic (Fig. 4) and the NADW on the bottom of the Atlantic.
The dominance of the planktic foraminifera Globorotalia truncatulinoides along with enriched δ13CNd isotope values in ODP Hole 994C, Blake Ridge, NW Atlantic, have been interpreted to indicate the presence of high-salinity, less-productive surface water in the northern Atlantic derived from an intensified Gulf Stream flow due to closure of the CAS136. Such changes have been recorded since 0.9 Mya and thereafter, as well as during earlier intervals136. Enhanced Gulf Stream flow is accompanied by increased North Atlantic Deep Water formation (as part of the AMOC). Periods of increased AMOC strength have been linked to increased salinity and warm water transport from the Mediterranean Outflow Water current (MOW)47. The increased AMOC strength is marked by depositional hiatuses on the route of the MOW, indicating erosion by bottom currents due to increased volumes circulating into the North Atlantic. Such hiatuses occurred during three periods, the last of which was from 0.9 Mya to 0.7 Mya47 (Fig.4D). Estimation of the variation in Northern Component Water (NCW) overflow, the ancient counterpart of the NADW, also highlights a culmination at 0.9 Mya46 (Fig.4C). Furthermore, the increasing domination of the Caribbean planktic foraminiferal assemblages by the salinity-tolerant species Globigerinoides ruber is considered to reflect increased Atlantic surface-water salinity due to cessation of sustained flow between the Pacific and the Caribbean9. The relative abundance of G. ruber culminated from 0.9 Mya to 0.6 Mya3,9 (Fig.4B).
Figure 4. Paleobiogeographic and oceanographic data supporting a major step in the restriction of the transisthmian seaways of the Isthmus of Panama, 0.89 Mya and ~ 0.7 Mya. A) Southern Caribbean Neogene biogeographical units15. B) Relative abundance of the high-salinity-tolerant planktic foraminiferal species Globigerinoides ruber, as a proxy of surface-water salinity9. C) Estimation of the variation in Northern Component Water (NCW) overflow46. D) Drill core data of IODP Expedition 339 sites, showing major hiatuses caused by increased saline and warm water transport from the Mediterranean Outflow Water current (MOW)47.
2.5. Paleoclimatic data
Climate models developed since the early 1990s have suggested that closure of the CAS would strengthen the production of warm, high-salinity Gulf Stream water, enhancing heat transport to the North Atlantic and the upper NADW formation in the Labrador and Nordic Seas (ref.12,137 and references therein). Accordingly, if transisthmian seaways existed in the early Quaternary as remnants of the CAS, their exposure intervals might have been printed in the Quaternary record of changes in Gulf Stream intensity. Because the transisthmian seaways were generally shallow before their final subaerial exposure, Quaternary glacioeustatic sea level changes should have influenced their open/close modes and, thus, the strengths of the Gulf Stream and the North Atlantic Current (Fig. 5). By this way, during glacial stages, sea level falls should have strengthened the Gulf Stream and the North Atlantic Current, increasing ocean surface temperature and transferring heat northwards, contra to the low temperatures that generally prevail during glacial stages. Thus, increased temperatures in North Atlantic surface waters during glacial stages might correspond to intervals of temporary closure and/or essential restriction of the transisthmian seaways. In Supplementary Fig. 2, proxies of surface temperatures of the Gulf Stream (IODP Site U1313) and the North Atlantic Current (ODP Sites 980 & 982) are correlated with the mean global temperature proxy and show that, indeed, such intervals existed between ~ 1 Mya and 0.6 Mya. Moreover, as the heat transport through the North Atlantic Current is accompanied by humidity transport, it has long been argued138,139, and was recently shown by coupled ocean–atmosphere general circulation models140,141,142,143, that closure of the CAS would result in enhanced precipitation over Greenland. Stratigraphical evidence from ODP Site 646 on the continental rise off southern Greenland (58°12.56N, 48°22.15 W)19 shows that such an event took place in the eye of the glacial maximum of Marine Isotope Stage 22 (MIS22; Supplementary Fig. 3). This event is evidenced by increases in pteridophyte and spermatophyte spore and pollen records (Supplementary Fig. 3E), respectively, suggesting that south (and probably east) Greenland experienced an unprecedented local climatic optimum characterised by increased vegetation growth. The event lasted ~ 20,000 years, from 0.89 Mya to 0.87 Mya, and was associated with an episodic increase of ~ 4.5°C in SSTs across the track of the North Atlantic Current (Supplementary Fig. 3). Two positive SST culminations in the record of ODP Site 98053 are particularly well matched with two contemporaneous positive peaks in the spermatophyte pollen record of MIS22 from South Greenland (Fig. 6 and Supplementary Fig. 3H), suggesting that the intensification of the North Atlantic Current was among the reasons for the unprecedented climatic optimum in South Greenland.