Pangea Map: Reconstruct 330 Million Years of Continental Drift

How to use the tools on this page

Start with the slider. Drag it left and the continents move, not along a path somebody drew but by rotating each block about a published Euler pole, the same arithmetic the research software uses. At 0 you get the world you know. Somewhere around 200 million years ago the Atlantic closes up. By 250 you are looking at Pangea.

Two things are worth doing before anything else. Turn on ice and coal, then set the slider to 300 Ma. The blue circles are Late Palaeozoic glacial deposits and the dark squares are Carboniferous coal basins. Today they are scattered across five continents and make no sense together. At 300 Ma the ice gathers into the far south and the coal lines up along the equator. That is the whole argument for continental drift in one picture, and it was not put there by hand. It falls out of the rotations.

Then switch to the globe and drag it. Pangea was a hemisphere of land facing a hemisphere of water, and a flat map hides that badly. On the globe you can see that the great ocean, Panthalassa, was bigger than the Pacific is now, and that almost none of its floor survives.

The second tab answers a question most people have as soon as they see the map. Pick your city, or type any coordinates, and the tool traces that exact point back through 330 million years and plots its palaeolatitude. Chicago sat within a couple of degrees of the equator at 300 Ma, which is why Illinois has coal. London was in the tropics. Cape Town was inside the polar circle when the Karoo ice sheet was on top of it.

The third tab is the evidence, twelve lines of it, with the objection each one faced at the time and what modern work has since changed. It is not a list of things that turned out to be right. Two of the twelve have real caveats attached today, and the page says so.

The last tab is the one most pages leave out. It sets out how much of this reconstruction is measured and how much is assumption, including a direct comparison of two published reference frames that put the same African point 13.4 degrees apart at 200 Ma. That is about 1,500 kilometres of disagreement between two respectable models, and it is the honest size of the uncertainty.

What Pangea actually was

Pangea was a supercontinent that existed from roughly 299 million years ago to roughly 200 million years ago, in which almost every continental block on Earth was joined into a single landmass surrounded by a single ocean.

The name comes from Alfred Wegener, who assembled it from Greek roots meaning all earth. He was not the first person to notice that the Atlantic looks like a tear. He was the first to build a scientific case that it was one.

Three facts are worth fixing before anything else. Pangea was not the only supercontinent, it was simply the most recent. Pangea did not contain quite all the land, and the exceptions matter. And the shape of Pangea that appears in textbooks is a hypothesis with a serious rival, not a measurement.

Almost all the land, but not all of it

North China and South China were separate blocks throughout, sitting out in the Palaeo Tethys ocean. So were the Cimmerian terranes, a long ribbon of crust that became Iran, Afghanistan, the Karakoram, Qiangtang, Lhasa and Sibumasu.

Cimmeria is the interesting case, because it rifted away from Gondwana at about 295 to 290 million years ago, which is right around the time Pangea is usually said to have finished assembling. The northern edge of the supercontinent was already shedding pieces before the southern half had finished welding on. Cimmeria then drifted north for roughly 150 million years before colliding with Eurasia, closing the Palaeo Tethys in front of it and opening the Neo Tethys behind.

So the picture of one clean landmass forming, sitting still, and then splitting is wrong in both directions. Assembly and dispersal overlapped. The careful phrasing is that Pangea contained almost all of Earth's landmasses, and the word almost is doing real work.

One ocean, and why that matters

Panthalassa surrounded the whole thing. It was larger than the modern Pacific, and it is the reason deep time reconstruction is so much harder than it looks.

Ocean floor is created at ridges and destroyed at trenches, and nothing survives the round trip. The oldest ocean floor anywhere on Earth today is about 180 million years old. Every square metre of Panthalassa has been subducted. That has a direct consequence for this page: the magnetic stripe record, which is the single best constraint we have on plate motion, simply does not exist for the age of Pangea. Everything you see at 250 Ma is built from palaeomagnetism, from geology matched across ancient sutures, and from assumptions.

Where your own town was

The second tab does something the abstract map cannot. It takes one point, yours if you want, and follows it.

Pick Chicago. At 300 million years ago it sat at 2.3 degrees south, essentially on the equator, and at 250 Ma it was at 12.3 degrees north. Illinois has coal because Illinois was an equatorial swamp, and the tool gets that without being told.

Pick London and you get much the same story, a little further east in the same tropical belt. Pick Cape Town and the story inverts: at 300 Ma it was deep inside the southern polar circle, under the ice sheet that left the Dwyka tillite. Pick Delhi and you watch India sit in the far south with the rest of Gondwana, then break away and run north across the Tethys at nearly 20 centimetres a year, the fastest sustained plate motion in the geological record, before slamming into Asia and raising the Himalaya.

Two cities in the list will not work, and the failure is the interesting part.

Reykjavik returns nothing, because Iceland is about 16 million years old and sits on ocean floor that had not been made yet. Manila is similar: it sits on an oceanic arc rather than on continental crust. A point in an ocean basin usually did not exist at all during Pangea, which is a harder idea than it first sounds. Continents persist for billions of years. Ocean floor does not.

Read the latitude, be careful with the longitude

The north to south number is the one to trust. Palaeomagnetism measures the angle a compass needle would have dipped at, and that angle depends only on latitude, so a rock carries its own latitude of formation with it.

The east to west number is different in kind. The geomagnetic field is symmetric about the spin axis, which means you can rotate a continent around that axis by any amount at all and its magnetic signature will not change by a hair. Palaeomagnetism therefore cannot constrain longitude. Ever. It is not a matter of better instruments.

Longitude in any reconstruction comes from somewhere else: from seafloor magnetic anomalies where they survive, which is only back to about 180 Ma, from hotspot reference frames, from the positions of subducted slabs imaged in the mantle, or from an optimisation that minimises implausible plate speeds. Those are reasonable methods. They are not measurements of the same quality as latitude, and the tool says so every time it shows you a longitude.

What the coloured blocks actually are

The shapes on the map are not coastlines, and that is deliberate.

They are continental blocks taken from the GPlates static polygon set, filtered to the polygons that represent continental crust rather than ocean floor. That includes the submerged continental shelf, which is why Europe looks fatter than the atlas and why there is a broad apron around Argentina and northern Australia.

Coastlines are the wrong line for this job. Sea level has moved by more than two hundred metres over the Phanerozoic, and a coastline is simply where that surface happens to intersect the land today. Fit two coastlines together and you are fitting an accident. Fit two continental margins and you are fitting the actual edges of the crustal blocks that rifted apart.

Edward Bullard and his colleagues settled this in 1965. They tested four different depth contours as candidate fitting lines and found the 500 fathom contour, about 914 metres, worked best, because it usually lies on the steepest part of the continental slope and therefore marks the structural edge rather than an erosional one.

The residual misfit at that contour is worth knowing. South America against Africa comes out at 0.93 degrees, about 88 kilometres. The whole Atlantic assembly comes out at 2.15 degrees, about 130 kilometres. There is a genuine 230 kilometre gap between Newfoundland and the English Channel that only partly closes when Iberia is rotated separately.

Those numbers are the answer to the oldest objection to drift, which was that any two wiggly lines can be made to fit. They can, roughly. They cannot be made to fit to within 88 kilometres over 6,000 kilometres of margin unless they were once joined.

How this reconstruction is computed

It is worth knowing what is under the slider, because it is simpler than most people expect and it explains where the uncertainty lives.

A plate's position at a past time is described by a finite rotation: a pole somewhere on the sphere, given as a latitude and longitude, plus an angle to turn through about that pole. That is all. Two numbers and an angle move an entire continent.

To move a point, you apply the Rodrigues rotation formula. Take the unit vector to the point, the unit vector to the pole, and the angle. The rotated vector is the point times the cosine of the angle, plus the cross product of pole and point times the sine of the angle, plus the pole times the dot product of pole and point times one minus the cosine. Convert back to latitude and longitude and you are done.

Published models supply these rotations at specific ages, usually the ages of magnetic anomalies rather than round numbers. Between those ages you have to interpolate, and how you interpolate matters. Blending the pole position and the angle separately gives the wrong answer, because a rotation is not the sum of its parts in that sense. The correct approach, and the one the research software uses, is to convert both rotations to quaternions, interpolate along the shortest arc between them, and convert back. This page does the same.

Rotations also compose. A plate is usually specified relative to a parent plate, which is specified relative to its parent, and so on up a hierarchy to whatever is held fixed. To get South America relative to Africa you multiply along that chain. To get an absolute position, the kind that gives you a real palaeolatitude, you multiply the whole chain by the absolute rotation of the anchor plate.

That last step is where the trouble is. Relative rotations between neighbouring plates are tightly constrained by the ocean floor between them, and different published models agree on them closely. Absolute rotations are a different problem: they require knowing where the whole plate system sat with respect to the Earth's spin axis, which needs palaeomagnetism, or with respect to the deep mantle, which needs an assumption about hotspots. Different models make different choices and their answers separate with age.

This page uses relative rotations from one published global model with Africa held fixed, and an absolute frame from a published apparent polar wander path for Africa with a true polar wander correction applied. Continental outlines are static polygons filtered to continental crust and simplified to a fifth of a degree, which cuts fifteen thousand vertices down to about two and a half thousand without changing any shape you can see at this scale.

Reading a palaeogeographic map

What the coloured blocks show is continental crust. What they do not show is where the sea was, and the difference is large.

Continental crust is not the same thing as dry land. Large parts of it flood when sea level rises, producing shallow epicontinental seas that can cover most of a continent. The Western Interior Seaway split North America in two during the Cretaceous. The Zechstein sea flooded northwest Europe in the Permian. On this map those areas are drawn as continent, because they are continental crust, which is the right thing to draw for a reconstruction but the wrong thing to read as coastline.

The map also does not show topography. The Central Pangean Mountains ran through the middle of the supercontinent at Alpine scale, and they governed where rain fell and where sediment went, but a rotation model has nothing to say about elevation.

And it does not show the ocean floor, because there is none left to show. The blank blue on a Pangea map is not empty. It is Panthalassa, floored by plates that were being created and destroyed the whole time, carrying island arcs and seamounts and microcontinents, every trace of which has been subducted. The blankness is a statement about what has been lost, not about what was there.

One more thing worth watching for on any reconstruction. Rigid block models produce small overlaps and gaps near rifted margins as continents are brought back together. Those are not errors in the fit. Margins genuinely stretch and thin during rifting, sometimes by tens of percent, so the present outline of a margin is wider than the original edge that separated. Modern deforming plate models handle this explicitly. Rigid ones, including this one, do not.

The evidence, one line at a time

No single observation proved Pangea. What did the work was a stack of independent lines that all pointed the same way, each of which had a plausible alternative explanation on its own and none of which had a plausible alternative explanation together.

The fossils that could not swim

Mesosaurus is a small aquatic reptile, about a metre long, from rocks of roughly 282 to 275 million years old. It occurs in the Irati Formation of the Parana Basin in Brazil, in the Whitehill Formation of the Karoo in South Africa and Namibia, and in the Mangrullo Formation in Uruguay. It occurs nowhere else on Earth.

The counterargument was that it swam. The reply was that it was a freshwater or at best brackish animal with limb proportions unsuited to open ocean, and that a 6,000 kilometre crossing of the Panthalassic ocean is not a small ask for something the size of a cat. A 2018 study argued it was semi aquatic rather than fully aquatic, which if anything strengthens the case. Its exact salinity tolerance remains unsettled.

Glossopteris is the better argument, because it is not one species but a whole flora. The same assemblage of seed plants turns up in South America, Africa, India, Australia and Antarctica, and in each place it is the dominant vegetation. Scott's polar party collected the Antarctic specimens in 1912 and the fossils were found with the bodies.

Lystrosaurus is the most persuasive single datum, and it arrived late. Peter Barrett found the first Antarctic land vertebrate, a fragment of jaw, in December 1967. The 1969 to 1970 field season at Coalsack Bluff in the Transantarctic Mountains, at 84 degrees south, produced abundant Lystrosaurus. The same animal is common in the Karoo of South Africa, in the Panchet Formation of India, in China and in Russia.

By 1970 there was essentially nobody left to argue with. Seafloor spreading had already been accepted. The Antarctic dicynodonts confirmed the answer rather than deciding it.

One caution the textbooks skip. Cynognathus is usually drawn sweeping right across southern Pangea from Argentina to South Africa. The African and east African record is solid, including well established occurrences in Tanzania and Zambia published in 2017. Some of the South American material is fragmentary and the referral is debated. The map overstates what the bones support.

One mountain range on four continents

Walk the Appalachians north into Newfoundland, cross the Atlantic, and pick the same belt up in Ireland, Scotland, Scandinavia, Svalbard and east Greenland. That is the Caledonian orogen, built between about 490 and 390 million years ago when Laurentia, Baltica and Avalonia collided and closed the Iapetus ocean. The suture runs, more or less, along the border between England and Scotland.

A second and younger belt runs through the southern Appalachians, the Ouachitas, Morocco and on into Iberia, France, Germany and Bohemia. That is the Variscan and Alleghanian orogen, built between roughly 380 and 260 million years ago as Gondwana closed on Laurussia. End to end the combined chain exceeded 10,000 kilometres, the largest collisional mountain belt of the Palaeozoic. Shortening in the Appalachians of North Carolina and Tennessee reached about 300 kilometres.

Two corrections to the popular version are worth making. These are two different orogens of different ages and they are routinely conflated. And the High Atlas you can see today is a Cenozoic range built on Variscan basement, so the line that the Atlas mountains are the Appalachians is true of the basement and not of the topography.

Ice in the wrong places

The strongest single line is glacial. Scattered across South America, Africa, India, Australia and Antarctica are Late Palaeozoic tillites, dropstones and striated pavements, all recording ice sheets. The Dwyka Group in the Karoo, the Itarare Group in the Parana Basin, the Talchir Formation in India, the Tarija beds in Bolivia.

Several of these are now in the tropics. That alone can be explained away by proposing a much colder Earth. What cannot be explained away is the direction the ice was flowing. The striations and pavements record ice moving outward from centres, and at Nooitgedacht near Kimberley the ice was flowing southward, which is to say off the modern coast and out to sea. On a reassembled Gondwana that direction makes sense. On the modern map it requires an ice sheet in the middle of the South Atlantic.

Modern work has revised the picture in a way worth knowing. It was not one continuous Gondwanan ice sheet. Current syntheses describe several ice centres of different sizes, from full ice sheets down to mountain glaciers, waxing and waning out of phase in episodes lasting one to eight million years across an interval from about 362 to 256 million years ago. The argument survives that revision intact, because the ice flow directions are what matter.

Coal on the equator, salt in the subtropics

At the same time as the southern ice, a belt of coal swamps ran across the tropical part of Pangea. The Appalachian, Illinois, Ruhr, South Wales and Donets basins are all the same phenomenon: equatorial rainforest, buried.

Run them back to 300 million years ago in the tool and they land at 8.0 south, 4.7 south, 5.7 north, 2.8 north and 12.1 north. Every one within 13 degrees of the equator. There is a second, quite different coal province in the far south, in the Karoo, in India, in Australia and in Antarctica, formed in cold, high latitude peatlands rather than tropical swamp.

A little further from the equator sit the evaporites and the reefs: the Zechstein basin of northwest Europe, the Castile Formation of the Delaware Basin in west Texas, the Capitan reef complex of the Guadalupe Mountains. Those form in the arid subtropical belt, at roughly 15 to 35 degrees, and that is exactly where the reconstruction puts them, worked out independently from magnetism.

The two pole paths that could not both be right

The measurement that changed professional minds was not a fossil. It was magnetism.

Rocks lock in the direction of the magnetic field as they cool or as sediment settles. Measure enough of them, of enough different ages, on one continent, and you can plot where the magnetic pole appears to have been through time. Ken Creer, Ted Irving and Keith Runcorn did this at Cambridge from 1954.

The result was that Europe gave one path and North America gave another. At 200 million years ago the pole calculated from North American rocks plots in China, while the pole from European rocks plots in the Pacific. There is only one magnetic pole. Two paths are impossible unless the continents have moved relative to each other.

Runcorn had started out sceptical of drift. He changed his mind in 1956 and became one of its proponents. That is a reasonable marker for the moment the argument stopped being about intuition and started being about numbers.

The tape recorder on the seafloor

The last piece was the mechanism, and it came from the ocean rather than the land.

Vine and Matthews published in September 1963, proposing that the symmetrical magnetic stripes surveyed either side of mid ocean ridges were a record of geomagnetic reversals frozen into crust as it spread outward. Lawrence Morley reached the same conclusion independently and had his paper rejected. Within three years, symmetric profiles such as the one collected by the research vessel Eltanin over the Pacific Antarctic Ridge had made it undeniable.

That settled continental drift by making it a consequence of something else. It also set the limit that shapes this entire page. Because no ocean floor older than about 180 million years survives, magnetic anomalies say nothing at all about the configuration of Pangea. The best tool we have runs out exactly where the interesting question starts.

A worked example you can repeat

Take the Illinois Basin coal, at roughly 39 north and 88 west today, and put it back to 300 million years ago.

The block it sits on carries a finite rotation for that age: a pole somewhere on the sphere and an angle to turn through. Rotating a point about that pole uses the Rodrigues formula, which takes the point vector, the pole vector and the angle, and returns the rotated vector. Written out it is the point times the cosine of the angle, plus the cross product of pole and point times the sine, plus the pole times the dot product of pole and point times one minus the cosine.

Do that and Illinois lands at 4.7 degrees south. Do the same for the Appalachian field and it lands at 8.0 south. For the Ruhr, 5.7 north. For South Wales, 2.8 north. For the Donets Basin, 12.1 north.

Now do the Karoo. It lands at 70.3 degrees south. The Transantarctic Mountains land at 78.8 south.

Nothing in that calculation knows what coal is or what a tillite is. The rotations were derived from magnetic and geological data with no reference to where the coal or the ice happened to be. The two datasets are independent, and they agree.

That is the test. If Pangea were an artistic reconstruction, there would be no reason for the coal to line up on the equator when you ran it through somebody else's rotation poles.

Doing it the other way

You can also run the test backwards, and it is worth doing because it catches errors.

Take a palaeolatitude you believe on independent grounds. The Dwyka tillite requires a polar or near polar setting. Ask the model where the Karoo was at 300 Ma and it says 70 degrees south. Agreement.

Ask it where the Karoo was at 200 Ma, by which time the Karoo Basin was accumulating the Elliot and Clarens formations, which are desert sandstones with aeolian dune bedding. The model says 37.8 degrees south, in the arid subtropical belt where deserts belong. Agreement again, and from a completely different kind of rock.

That progression, from ice at 300 Ma to desert at 200 Ma, is recorded in the Karoo section itself as a vertical succession you can walk through. The tool reproduces it as a consequence of Gondwana drifting north. Neither was tuned to match the other.

How Pangea came together

Supercontinents do not appear. They are assembled, by closing oceans, and the closing takes hundreds of millions of years.

The story that ends in Pangea starts with the breakup of the previous supercontinent. Rodinia assembled between roughly 1,300 and 900 million years ago and came apart in pulses at about 825, 780 and 750 million years ago, driven by episodes of plume magmatism that show up on several continents at once.

Out of that dispersal came the Iapetus ocean, opening at about 600 to 570 million years ago as southeastern Laurentia rifted away from Amazonia and Baltica. Then, at about 490 to 480 million years ago, a second ocean opened behind the first: the Rheic, formed when the terranes of Avalonia and Carolina rifted off Gondwana. There is a good argument that the Rheic opened because the Iapetus was already closing to the north, and the slab pulling down there pulled Avalonia away. The Rheic eventually exceeded 4,000 kilometres in width.

Closing those two oceans is what built Pangea.

The Iapetus went first. Laurentia, Baltica and Avalonia converged through the Caledonian orogeny, from about 490 to 390 million years ago with the main collision between 420 and 405. The result was Laurussia, sometimes called the Old Red Sandstone continent after the distinctive red terrestrial sediments shed off the new mountains.

The Rheic went second and took longer. The Variscan orogeny welded northwest Africa to southern Europe from about 380 to 300 million years ago. The Alleghanian and Ouachita orogenies welded Africa and South America to North America from about 325 to 260. That last collision is the final stitch. When it finished, Pangea existed.

The date usually given for full assembly is somewhere between 299 and 273 million years ago, in the Early Permian. That is a range rather than a date because the closure was not simultaneous along its length, and because there is a live argument about the arrangement even after assembly, which the last section of this page deals with.

Pangea's world

A single landmass spanning pole to pole, with one ocean around it, produces a planet that behaves differently from ours.

The monsoon with nothing to stop it

The most important consequence was climatic. Continents heat and cool faster than oceans, so a large landmass straddling the equator sets up a seasonal pressure contrast that drives monsoonal circulation. Pangea was the largest landmass the Earth has had since Rodinia, and it was arranged roughly symmetrically about the equator, particularly in the Triassic.

The result is generally called the megamonsoon, and it was stronger than any monsoon operating today. The evidence for it is not modelling alone. Aeolian sandstones on the Colorado Plateau preserve the direction ancient dunes were migrating, and those palaeowind directions rotate from north northeast in the Permian to northwest in the Triassic, which is what a strengthening monsoonal circulation predicts. Evaporites and red beds turn up in equatorial positions where a simple zonal climate model says they should not. Peat, which needs reliable water, migrated from low latitudes to high ones across the Carboniferous and Permian boundary.

Modelled seasonality in southern Pangea reaches a cold month mean of minus 30 degrees and a warm month mean of plus 25. That is a continental interior with a 55 degree annual swing, which is more extreme than anywhere on Earth today outside Siberia.

Deserts, and how far from the sea you could get

Parts of Pangea lay more than a thousand kilometres from any coastline. Moisture simply did not reach them.

Red beds spread through the Permian and Triassic, and they are worth reading carefully. Red beds require alternating wet and dry conditions, not permanent aridity, because you need water to move iron and oxygen to fix it. Their spread records strong rainfall seasonality rather than uniform desert. Triassic evaporites accumulated faster and across a greater area than in any other interval in Earth history.

The mountains down the middle

The Variscan and Alleghanian collision left a mountain chain running through the centre of the supercontinent, the Central Pangean Mountains, and it is routinely described as Himalayan.

The published measurements do not support that. Stable isotope palaeoaltimetry, which reads elevation from the hydrogen isotope composition of minerals that grew during deformation, gives 2,500 metres plus or minus 900 for the Armorican Massif and 3,300 plus or minus 1,000 for the French Massif Central in the Late Carboniferous. Those are Alpine numbers, not Himalayan ones, at least for the segment that has been measured. Anyone quoting 8,000 metres is quoting an assumption.

Sea level and the air

Sea level during the Late Palaeozoic Ice Age swung with the ice, in cycles ranging from a few tens of metres up to about 120 metres. The Palaeozoic maximum, roughly 225 metres above present, was much earlier, in the Late Ordovician. The Palaeozoic minimum fell in the early Late Permian, on a fully assembled Pangea, which is what you would expect: a single continent has less flooded shelf than several scattered ones.

Atmospheric oxygen was probably higher than today, and here the honest answer is a range. The consensus curve puts the Permo Carboniferous maximum above 25 percent, charcoal based estimates suggest around 30 percent in the Early Permian, and one widely quoted model gives 30 to 35 percent. Different methods disagree by five to ten percentage points, so a single number should be treated with suspicion.

Carbon dioxide went the other way. It was close to pre industrial values through much of the Carboniferous, which is part of why there was an ice age, then climbed steeply to around 2,000 parts per million in the Triassic with an individual maximum near 3,700 at 215 million years ago.

The worst day in the history of animals

The end Permian extinction happened on a fully assembled Pangea, about 251.9 million years ago, or about 251.4 depending on which dating study you follow.

It was fast. The extinction interval is constrained to roughly 60,000 years, with an uncertainty of 48,000 either way, and the carbon isotope excursion that marks its onset took less than 10,000 years. Estimates of loss range from about 81 to 95 percent of marine species depending on method and taxonomic rank.

Pangea's role in it is more modest than popular accounts suggest, and this is worth being straight about. The proximate cause was the Siberian Traps: an enormous volcanic province releasing carbon dioxide, driving about 8 degrees of ocean warming and widespread anoxia. Pangea contributed context. A single landmass means less continental shelf, and a single ocean basin circulates more sluggishly, which makes anoxia easier. The older argument that reduced shelf area caused the extinction through a species area effect is no longer treated as a primary mechanism.

How Pangea came apart

Breakup began where the continent was weakest, along the Variscan and Alleghanian suture that had welded it together in the first place. New rifts tend to follow old scars.

The event that opened it was volcanic. At 201.6 million years ago the Central Atlantic Magmatic Province erupted, covering something like 11 million square kilometres with 2 to 6 million cubic kilometres of basalt in four pulses spread over roughly 600,000 years. Individual eruptions may have lasted no more than a century each. It is the largest continental flood basalt province known, and its remnants are now scattered across four continents, which is itself a drift argument.

The Triassic and Jurassic boundary, and the mass extinction that goes with it, sits inside that interval. The continental extinction is dated to 201.564 million years ago, plus or minus 0.015. That is a level of precision that would have astonished Wegener.

Seafloor spreading in the Central Atlantic followed, somewhere between 200 and 170 million years ago. The exact onset is genuinely controversial and you should treat any single number for it with suspicion.

Gondwana went next. The Karoo and Ferrar flood basalts erupted from about 200 to 170 million years ago, and East Gondwana began separating from Africa. The oldest magnetic anomalies in the Weddell Sea date from 180 to 160 million years ago.

The South Atlantic opened last, and it opened like a zip, from south to north. The southern segment started at about 134 million years ago, marked by magnetic anomaly M9 off Cape Town, and the Parana and Etendeka flood basalts erupted at 134 to 129. The central segment did not open until about 112, held up for somewhere between 5 and 20 million years by the Florianopolis fracture zone, which acted as a barrier to the propagating rift.

India then did something remarkable. Having separated from Madagascar at about 70 million years ago, it ran north across the Tethys, reaching a peak convergence rate of 18 to 19.5 centimetres a year at around 57.5 million years ago. That is the fastest sustained plate motion in the geological record, more than twice the rate of most plates today. It decelerated from about 18 to about 5 centimetres a year between 50 and 35 million years ago, which is the signature of a collision.

The collision itself is dated, on the best palaeomagnetic evidence, to about 50 million years ago at roughly 24 degrees north. That number is contested. Published proposals span from about 59 to about 34 million years ago, and there are two stage models in which a soft collision long preceded a hard one. Anyone who gives you a single confident date for the start of the Himalaya is compressing a live argument.

The man who was right for the wrong reason

Alfred Wegener presented continental drift for the first time on 6 January 1912, in a lecture to the Geological Association at the Senckenberg Museum in Frankfurt. He published the full argument as a book in 1915, and revised it in 1920, 1922 and 1929.

He was not first to see the fit. Abraham Ortelius wrote in 1596 that anyone considering the facing coasts of Europe, Africa and America carefully would see the vestiges of a rupture, a passage that was forgotten for nearly four centuries until a classicist rediscovered it in 1994. Antonio Snider Pellegrini drew the first before and after maps in 1858, using matching coal plant fossils in North America and Europe as his evidence, and buried the observation inside a catastrophist religious framework that discredited it.

Wegener's contribution was to assemble the geological, palaeontological, palaeoclimatic and geodetic evidence into one argument and press it hard. The response was hostile, particularly in the English speaking world.

The set piece attack was the 1926 symposium of the American Association of Petroleum Geologists in New York. Rollin Chamberlin of Chicago described the hypothesis as of the footloose type, taking considerable liberty with the globe and being less tied down by awkward facts than its rivals. He is also credited with saying that believing it would require forgetting everything learned in the previous seventy years and starting again.

What he actually got wrong

The evidence was not the problem. The mechanism was.

Wegener proposed that continents were driven by Polflucht, the equatorward component of the Earth's rotational force, and by tidal forces from the sun and moon. He described continents ploughing through oceanic crust the way an icebreaker ploughs through ice.

Harold Jeffreys, the most formidable of his critics, showed that both forces are orders of magnitude too weak for that and that the crust is far too strong to be ploughed through. Jeffreys was right. Wegener knew it. In the 1929 edition he wrote that the Newton of drift theory had not yet appeared.

He also made a tactical error that cost him dearly. He cited astronomical longitude measurements as evidence that Greenland was moving away from Europe at metres per year. The published figures vary between accounts, from about 11 metres a year up to 32 or 36, and one account gives 250 centimetres a year for North America against Europe. The real Atlantic opening rate is about 2.0 to 2.5 centimetres a year, so his figures were somewhere between a hundred and a thousand times too fast.

The measurements were simply not accurate enough to detect what he was looking for, and he flagged the uncertainty himself. It did not help. Handing critics a demonstrably wrong number is a good way to lose an argument you are otherwise winning.

The one who supplied the mechanism, and was ignored

Arthur Holmes gave the answer in a lecture in January 1928 and in papers published between 1928 and 1931. Radioactive decay heats the mantle. The mantle convects. Convection currents drag continents apart and create new ocean floor in the gap between them.

That is seafloor spreading, thirty years early, and it is essentially correct.

It was ridiculed. A contemporary recalled students likening his rising convection currents to the froth in a glass of beer. Holmes himself was candid about the weakness, writing that the treatment had been almost entirely qualitative and stood in need of quantitative revision.

So the common claim that Wegener had no mechanism deserves nuance in two directions. Wegener had a mechanism and it was wrong. Holmes had a mechanism and it was right, and the profession rejected it anyway.

Du Toit, who had the better case

Alexander Du Toit, working in South Africa, made the geological argument far more rigorously than Wegener ever did. His 1927 comparison of South America with South Africa and his 1937 book Our Wandering Continents replaced Wegener's single Pangea with two landmasses, Laurasia to the north and Gondwana to the south, separated by the Tethys. That two part structure is closer to what the models show today than Wegener's version was.

Du Toit had walked the Karoo and the Parana. He knew the glacial pavements and the Glossopteris beds firsthand. His case was persuasive in the southern hemisphere, where geologists could see the rocks, and largely ignored in the north, where they could not.

How the argument was actually won

Continental drift was not accepted because someone finally found a convincing fossil. It was accepted because a different question got answered.

Between 1960 and 1968, four things happened. Harry Hess proposed seafloor spreading in 1960 and published in 1962, calling it geopoetry because he knew the evidence was thin. Vine and Matthews explained the magnetic stripes in 1963. Tuzo Wilson described transform faults in 1965, supplying the third kind of plate boundary and making a self consistent rigid plate kinematics possible. McKenzie and Parker in 1967 and Morgan in 1968 wrote that kinematics down formally.

Bullard's computer fit, published in 1965, arrived in the middle of that and mattered because it was quantitative in a way no previous fit had been.

The shift took roughly three years, which is one of the fastest reversals in the history of the physical sciences. Wegener had been dead for thirty five years, having died on the Greenland ice cap in 1930.

The lesson is not that the sceptics were fools. Jeffreys was a superb geophysicist and his objection to Wegener's mechanism was correct. The theory as Wegener stated it, with continents ploughing through the ocean floor, deserved to be rejected. What got accepted in the 1960s was a different theory that happened to produce the same map.

What one landmass did to life

Joining every continent together does something specific to biology: it removes barriers, and barriers are what generate diversity.

Through the Carboniferous, when the continents were still converging, land faunas across the world were noticeably provincial. Different regions had different assemblages because they had been isolated long enough to diverge. By the Late Permian and Early Triassic, on a fully assembled Pangea, that provincialism had collapsed. Lystrosaurus is the emblem of it. The same genus turns up in South Africa, India, China, Russia and Antarctica, which is a distribution no land vertebrate could achieve on the modern map.

Homogenisation of that kind is usually bad for total diversity even though it looks like success for individual species. When previously separated faunas meet, some lineages spread everywhere and many others are outcompeted, and the global species count falls even as individual ranges grow. The pattern is visible in the marine record too, where a single continuous shelf around one landmass supports fewer distinct provinces than several scattered continents do.

There is a second effect that gets less attention. A supercontinent reduces the total length of coastline and the area of shallow shelf, and shallow shelf is where most marine life lives. Less shelf means less habitat. That was once offered as a primary explanation for the end Permian extinction through a species area relationship, and it is no longer treated that way, but as background it is real. Pangea was a low diversity world before anything went wrong with it.

Then the breakup ran the process in reverse. As the Atlantic opened and Gondwana fragmented, populations that had been continuous were cut into pieces, and those pieces diverged. Much of the biogeography that seems obvious today is a direct consequence: marsupials in Australia and South America but not Africa, lemurs on Madagascar and nowhere else, ratite birds scattered across the southern continents in a pattern that traces the order of Gondwanan breakup rather than any flight path.

The Madagascar case is the crispest. It separated from Africa early, at around 160 million years ago, and from India at about 88 million years ago, and then sat alone. Almost everything that lives there evolved in isolation for longer than the Atlantic has existed.

The exception that proves the rule

Not every disjunct distribution is a drift story, and this is where the reasoning goes wrong most often.

Plants and small animals do cross oceans. Rafting on floating vegetation is well documented, and some plant groups with distributions that look Gondwanan turn out, when dated with molecular clocks, to have dispersed long after the continents separated. The southern beech genus Nothofagus, once the textbook example of a Gondwanan relict, has a more complicated history than the simple vicariance story allows.

That is why the fossil evidence for Pangea rested on animals that plainly could not cross, on whole floras rather than single species, and on assemblages rather than individual occurrences. A single shared species is an anecdote. An entire flora shared across five continents, in rocks of the same age, with the same associated fauna, is not.

What we do not know about Pangea

Every Pangea map you have seen, including the one on this page, is drawn with more confidence than the evidence supports. Here is where the joints are.

Longitude is not measured

This is the big one and it is structural rather than technical.

A magnetised rock records the inclination of the ancient field, which gives latitude, and the declination, which gives the orientation the continent was facing. It records nothing about longitude, because the field is symmetric about the spin axis. Rotate Africa 40 degrees eastward about the Earth's axis and every palaeomagnetic measurement ever made on it stays exactly the same.

Where does longitude come from then? From seafloor magnetic anomalies, which run out at about 180 million years. From hotspot reference frames, which assume plumes are fixed, an assumption known to be imperfect. From matching geology across sutures, which is a strong constraint locally and a weak one globally. And from optimisations that pick the longitude minimising implausible plate speeds.

The practical consequence is measurable. Comparing two published reconstructions in an Africa fixed frame, the position of North America relative to Africa agrees to within a couple of degrees at 200 and 250 million years ago, but by 400 million years ago the two models differ by 65 degrees. Eurasia differs by 75. That is not a small correction. It is a completely different world, and it happens because before Pangea assembled at around 320 million years ago there is no preserved seafloor to tie Laurussia to Gondwana at all.

That is why the tool on this page stops at 330 Ma rather than running back to 500 or 600 the way many animations do. Past assembly, the models are not refining a picture. They are disagreeing about which picture.

Two published frames, 1,500 kilometres apart

To put a number on the softer uncertainty inside the range we do cover, we ran the same point in the Congo through two published absolute reference frames: the moving hotspot frame used in one widely cited model, and a palaeomagnetic frame with a true polar wander correction from another.

Out to 100 million years ago they agree to within a degree. At 140 Ma the gap is 4.9 degrees. At 180 it is 7.3. At 200 it is 13.4 degrees, which is roughly 1,500 kilometres of latitude on the same continent at the same moment, from two models both of which are perfectly respectable.

Neither is wrong. The difference comes from whether and how true polar wander is corrected for, and that is an open methodological question. Almost nobody shows this, which is why a Pangea map tends to look more settled than it is.

Pangea A or Pangea B

There is a live disagreement about the shape of the supercontinent itself.

The familiar arrangement is Pangea A. An alternative, Pangea B, places Gondwana roughly 3,000 kilometres further east relative to Laurasia in the Early Permian. If Pangea B is right, then getting to Pangea A by the Triassic requires an enormous strike slip shear between the northern and southern halves, of about 3,000 kilometres, completed in perhaps 25 million years, which implies relative plate speeds of 10 to 15 centimetres a year along a continental transform.

Supporters point out that Pangea A does not fit the Permian palaeomagnetic poles well and Pangea B does. Critics reply that the misfit is an artefact of inclination shallowing, a known bias in sedimentary rocks where compaction flattens the recorded magnetic direction and makes the rock appear to have formed closer to the equator than it did.

This is unresolved. The map on this page uses a Pangea A style model, because that is the mainstream choice, not because the question is settled.

One silent failure worth naming

Building this page turned up a trap that anyone doing their own reconstruction should know about.

Plate rotation models cover a stated time range. Ask them for a rotation outside that range and they do not always tell you. Asked for Greenland at 250 million years ago, the underlying rotation file used here quietly returns an identity rotation, which means no rotation at all, which means Greenland is drawn at its present day position on a Pangea map. It looks like an answer. It is the absence of one.

The same file returns identity for India, Australia and Antarctica past 370 million years ago. The tool detects these cases and, for Greenland, holds it welded to North America instead, which is where it belongs until the Labrador Sea opens in the Cretaceous.

If you build a reconstruction from published rotation files, check every plate for silent identity rotations at every age before you trust a single frame. It is the sort of error that produces a picture that looks perfectly plausible and is completely wrong.

The supercontinent cycle

Pangea was not a one off. Continents appear to gather and disperse repeatedly, on a period of very roughly 400 to 600 million years.

Rodinia is the best established predecessor, assembled between 1,300 and 900 million years ago and dispersing from about 825. Before that, Nuna, also called Columbia, reached final assembly at about 1,820 million years ago. Before that, possibly, Kenorland at about 2,720.

Confidence falls off fast with age, and it falls off differently for existence and for shape. Rodinia's existence is not seriously doubted. Its configuration is: the standard synthesis states outright that there is no consensus on how many cratons took part, how they were arranged, or the chronology of assembly and breakup. Kenorland's existence is itself disputed, with a serious competing model treating the Slave and Superior cratons as two separate supercratons rather than parts of one landmass.

Pannotia is the sharpest argument. Proposed at about 633 to 573 million years ago, it is defended by researchers who argue the supporting evidence is stronger than for any supercontinent except Pangea, and rejected by others who argue the palaeomagnetic data and precise dating do not support it, and that accepting it would make the supercontinent cycle implausibly fast. That is a genuine live dispute and this page does not take a side.

The next one

Four scenarios are seriously argued for the supercontinent after this one, and they disagree about which ocean closes.

Pangaea Proxima, proposed by Christopher Scotese in 2003, has the Atlantic reverse and close within about 250 million years. That is called introversion: an interior ocean shuts again. He renamed it from Pangaea Ultima to avoid implying it would be the last.

Novopangaea, proposed by Ted Nield in 2007, has the opposite happen. The Atlantic keeps opening and the Pacific keeps closing, so the Americas eventually collide with Antarctica and Asia. That is extroversion, and it is often argued to be the most probable simply because it requires nothing to change from what is happening now.

Aurica, from Joao Duarte and colleagues in 2018, has both oceans close and a new Pan Asian ocean open across Siberia, Mongolia and India.

Amasia, from Mitchell and colleagues in 2012 using a name coined by Paul Hoffman in 1992, has the continents gather around the North Pole, at ninety degrees to Pangea's centroid, following the ring of subduction zones inherited from Pangea's breakup. That is orthoversion. The predicted timing has been given as about 200 million years, though later work by the same group suggests substantially longer.

None of these is a prediction in the sense a weather forecast is. They are extrapolations of present kinematics under different assumptions about what happens when an ocean gets old.

What is taught wrong about Pangea

That it was the only supercontinent

It was the most recent one, and the only one whose breakup left ocean floor we can still read. Rodinia, Nuna and arguably Kenorland came before it. Pangea is famous because the evidence survives, not because it was unique.

That it contained all of Earth's land

North China, South China and the Cimmerian terranes were separate blocks. Cimmeria had already rifted away from Gondwana by 295 to 290 million years ago. There were island arcs and microcontinents scattered through Palaeo Tethys. Almost all is the correct phrase.

That continents float on liquid magma

The asthenosphere is solid rock. It flows over geological time, at roughly the rate fingernails grow, but a handful of it would be rock, not liquid. Magma exists in comparatively small volumes near subduction zones and spreading centres. The only fully liquid layer in the Earth is the outer core, and it is 2,900 kilometres down.

That Wegener had no mechanism

He had two, Polflucht and tidal forces, and both were far too weak. He conceded the point himself. Meanwhile Arthur Holmes had published a mechanism that works, mantle convection driven by radioactive heat, in 1928, and was ridiculed for it. The accurate summary is that a correct mechanism existed by 1930 and the profession did not want it.

That palaeomagnetism gives you longitude

It gives latitude and orientation, and it is mathematically incapable of giving longitude. This is the single most consequential thing people get wrong about deep time reconstructions, because it means the east to west positions in any Pangea map are a modelling choice rather than a measurement.

That the shape of Pangea is known precisely

The Pangea A against Pangea B question is unresolved and the difference between them is 3,000 kilometres. The familiar textbook map is one hypothesis rendered without error bars.

That Africa and South America fit at their coastlines

They fit at the continental slope, at about 914 metres depth, and even there the misfit is 88 kilometres. Coastlines are sea level accidents. Rifted margins are also stretched and thinned during breakup, so their present outlines are not the shapes that originally separated.

That continental drift and plate tectonics are the same theory

They are different mechanical claims and one of them is false. Wegener had continents ploughing through ocean floor. Plate tectonics has rigid lithosphere, ocean and continent together, moving as units over a ductile asthenosphere, created at ridges and destroyed at trenches. Continents are passengers, not bulldozers. The modern theory vindicated Wegener's conclusion and discarded his mechanism.

Mistakes that catch people out

Reading a palaeolongitude as if it were a measurement. It is the softest number on any reconstruction and it gets softer with age.

Comparing two reconstructions without checking they use the same reference frame. Africa fixed and absolute frames give different answers, and two absolute frames give different answers from each other. Most published disagreements about deep time positions turn out to be frame disagreements.

Assuming a rotation model covers the age you asked for. It often does not, and it often does not say so.

Treating the tropics and polar circles as fixed lines. The tropics are set by the Earth's axial tilt, which varies, and the climate belts they imply have never been perfectly zonal, least of all on Pangea where the monsoon overrode them.

Expecting a point in an ocean basin to have a history. It usually does not. Ocean floor older than about 180 million years does not exist anywhere.

Assuming the continents were rigid. They were not, especially at rifted margins, which stretch and thin by tens of percent. Rigid block reconstructions produce small overlaps and gaps near margins that are real deformation, not error.

Common questions about Pangea

What was Pangea?

Pangea was a supercontinent in which almost every continental block on Earth was joined into a single landmass, surrounded by a single ocean called Panthalassa. It existed from roughly 299 million years ago to roughly 200 million years ago. The name was given by Alfred Wegener and comes from Greek roots meaning all earth.

When did Pangea exist?

Full assembly is usually placed between 299 and 273 million years ago, in the Early Permian. Breakup began in the Late Triassic, with the Central Atlantic Magmatic Province erupting at 201.6 million years ago and seafloor spreading starting somewhere between 200 and 170 million years ago. So Pangea existed as a single landmass for roughly a hundred million years.

Was Pangea really all the land on Earth?

Almost, but not quite. North China, South China and the Cimmerian terranes, which became Iran, Afghanistan, the Karakoram, Tibet and parts of southeast Asia, were separate. Cimmeria had already rifted away from Gondwana by 295 to 290 million years ago, before the southern half of Pangea had finished assembling. There were also island arcs and microcontinents scattered through the Palaeo Tethys ocean.

What is the evidence for Pangea?

Several independent lines. The continental margins of the Atlantic fit together to within 88 kilometres. Identical fossils, including Mesosaurus, Glossopteris, Cynognathus and Lystrosaurus, occur on continents now separated by oceans. One mountain belt runs continuously across four continents when they are reassembled. Late Palaeozoic glacial striations record ice flowing in directions that only make sense on a joined Gondwana. Carboniferous coal basins line up along a single equator. And two continents give two different apparent polar wander paths for the same time, which is impossible unless they moved relative to each other.

Who discovered Pangea?

Alfred Wegener made the first scientific case, presented in January 1912 and published as a book in 1915. Others had noticed the Atlantic fit earlier, including Abraham Ortelius in 1596 and Antonio Snider Pellegrini in 1858. Alexander Du Toit built a far more rigorous geological case in 1927 and 1937, and it was Du Toit who split Pangea into Laurasia and Gondwana.

Why was continental drift rejected for so long?

Because of the mechanism, not the evidence. Wegener proposed that continents were pushed by the Earth's rotation and by tidal forces, ploughing through the ocean floor. Harold Jeffreys showed those forces are orders of magnitude too weak and the crust far too strong. That objection was correct, and Wegener conceded it in 1929. Drift was accepted only after seafloor spreading supplied a mechanism that worked.

What broke Pangea apart?

Rifting along the old Variscan and Alleghanian suture, the same line that had welded the supercontinent together, accompanied by the eruption of the Central Atlantic Magmatic Province at 201.6 million years ago. That province covered about 11 million square kilometres. The underlying driver is mantle convection, with slab pull from subducting plates as the dominant force once spreading is underway.

What was the climate of Pangea like?

Extreme and strongly seasonal. A single landmass straddling the equator drove a monsoonal circulation stronger than any today, usually called the megamonsoon, which peaked in the Triassic. Continental interiors more than a thousand kilometres from the sea were arid. Modelled seasonality in southern Pangea reaches a cold month mean of minus 30 degrees and a warm month mean of plus 25. There was also a large ice sheet system in the far south for much of the Carboniferous and Permian.

Where was my city during Pangea?

The tool on this page answers that for 56 cities or for any coordinates you type. Chicago was within a couple of degrees of the equator at 300 million years ago. London was in the tropics. Cape Town was inside the southern polar circle under an ice sheet. Two cities in the list return nothing, Reykjavik and Manila, because they sit on crust that did not exist yet.

Why do the continents fit but not perfectly?

Because coastlines are the wrong line to fit. Sea level moves and coastlines follow it. The real edges are the continental margins, at roughly 914 metres depth, and there the fit is good to 88 kilometres for South America against Africa. Rifted margins are also stretched and thinned during breakup, so the shapes you see today are not exactly the shapes that separated.

What was Panthalassa?

The single world ocean that surrounded Pangea. It was larger than the modern Pacific. Almost none of its floor survives, because ocean crust is continually destroyed at subduction zones, which is why reconstructions of the Pacific basin at this age are much weaker than the continental ones.

What was the largest supercontinent?

Pangea is the largest for which the evidence is solid, but the honest answer is that comparing sizes across supercontinents is difficult because the configurations of earlier ones are poorly known. Rodinia may have been comparable. Statements ranking supercontinents by area should be treated with caution.

Will Pangea form again?

Something like it probably will, in roughly 200 to 250 million years, but which arrangement is disputed. Four scenarios are seriously argued. Pangaea Proxima closes the Atlantic. Novopangaea closes the Pacific instead. Aurica closes both and opens a new ocean across Asia. Amasia gathers everything around the North Pole. These are extrapolations of present plate motion, not predictions in the everyday sense.

How fast did the continents move?

Typically a few centimetres a year, comparable to fingernail growth, which is also the rate the modern Atlantic opens at, about 2.0 to 2.5 centimetres a year. The record holder is India, which reached 18 to 19.5 centimetres a year at about 57.5 million years ago on its run north to collide with Asia. Wegener's published rates were between a hundred and a thousand times too fast, which badly damaged his case.

How accurate are Pangea maps?

Less accurate than they look. Latitude is well constrained by palaeomagnetism. Longitude is not constrained by it at all, because the geomagnetic field is symmetric about the spin axis, so east to west positions come from other assumptions. Two published reference frames put the same African point 13.4 degrees apart at 200 million years ago. And the overall shape is disputed between two configurations, Pangea A and Pangea B, that differ by about 3,000 kilometres.

What came before Pangea?

Rodinia, assembled between 1,300 and 900 million years ago and dispersing from about 825. Before that Nuna, also called Columbia, with final assembly at about 1,820 million years ago. Before that possibly Kenorland at about 2,720 million years ago, though whether Kenorland was a single supercontinent is disputed. Pannotia, proposed at 633 to 573 million years ago, is the subject of an active argument about whether it existed at all.

Related Calculators and Reference

• Seafloor Spreading Calculator and Magnetic Stripe Explorer works the mechanism that took Pangea apart, and the magnetic stripe record that dates every step of it back to about 180 million years.
• Subduction Zone Explorer and Megathrust Calculator covers the other half of the system, where Panthalassa went and why none of it survives to constrain this map.
• Interactive Geological Time Scale puts every date on this page, from Rodinia to the Himalaya, on one dated axis.
• Rock Cycle Diagram and Identification Key covers the tillites, coals and red beds that this reconstruction is tested against.
• Mohs Hardness Scale and Mineral Identification covers the minerals whose magnetic remanence carries the palaeolatitude signal used here.
• Rock Density Calculator turns a mineral assemblage into a bulk density, which is what decides whether a block of crust rides or sinks.
• GPlates is the research software whose rotation conventions and static polygon geometry this page follows.