The relief is the real sea floor (ETOPO 2022, 1.85 km per cell). Today's sea floor is not the old land surface: sediment has buried valleys. The turquoise rivers are computed, not mapped.Turquoise rivers are computed; today's sea floor is not the old land.
A whole continent, now sea floor
At the peak of the last ice age, part of the oceans' water is locked in ice sheets, and the sea around Southeast Asia falls by 114 to 121 metres, depending on the record: the one made on the Sunda Shelf itself gives −116 m, the global coral curve gives −121 m. The sea floor between the Malay Peninsula, Sumatra, Java and Borneo is very flat and shallow, so almost all of it comes out of the water.
The land on the map is computed from the measured relief of the sea floor. Inside the dotted frame, about 2.18 million square kilometres of today's sea floor come out, as much as Greenland. The number depends on where the frame is drawn: Voris (2000) counted 1.53 million square kilometres of new, contiguous sea floor south and east of the Isthmus of Kra. This frame is drawn and counted differently, so the two numbers cannot be compared directly.
Gold is land that lies under water today; green is today's land; the pale line is today's coast, kept as a reference. The tide staff and the “Play the flood” button move the water across the whole map.
about 2.18 million square kilometres of new land
How low the sea stood: two records, one test
The global sea-level curve comes from corals at Barbados and Tahiti. The Sunda Shelf has a record of its own: Hanebuth, Stattegger and Grootes (2000) radiocarbon-dated samples from its sediment and calibrated the ages to calendar years. The chart above sets them side by side. Tap it and the sea on the map follows.
The two curves do not agree. At 19,000 years the Sunda record stands 5 m higher than the coral curve (−114 m against −119 m). At 14,600 years they differ by only 1.8 m. At 14,300 years the gap is 15.5 m: here the sea passes −80 m 14,300 years ago, while in the coral record it does so only 13,800 years ago, about 500 years later. At 13,100 years the gap is 5.9 m. It may come from the dating methods (uranium-thorium on corals, calibrated radiocarbon on the shelf) and from local crustal movement.
The meltwater pulse of about 14,600–14,300 years ago raised the sea by 16 m in 300 years, or 5.3 m a century, according to Hanebuth et al. The Sunda record stops at 13,100 years and the page continues from there on the coral curve. That is why every strait shallower than 60 metres closes at the same dates in both versions; the versions differ only for what happened earlier.
Tap the chart: the sea on the map follows it.
| Years ago | Sunda | Corals | Gap |
|---|---|---|---|
| 19,000 | −114 m | −119 m | +5 m |
| 14,600 | −96 m | −97.8 m | +1.8 m |
| 14,300 | −80 m | −95.5 m | +15.5 m |
| 13,100 | −64 m | −69.9 m | +5.9 m |
up to 15.5 m between two curves, at 14,300 years
Rivers that now run across the sea floor
On a plain this flat, water gathers into long rivers. Voris (2000) distinguishes four great river systems: North Sunda, East Sunda, Siam and the Strait of Malacca. The geologist Molengraaff mapped some of the North Sunda valleys as early as 1921.
On the map the turquoise lines are computed: for the land at −120 m, water flows downhill over a relief smoothed over about 15 km and gathers where the largest basins collect. The computed paths carry the water toward the edges of the shelf, at several points.
They are traces of the great flow, not mapped channels. The sea floor in the model is interpolated between sparse soundings and covered by 20,000 years of sediment, and the straight lines that cut across the shading in some areas look like survey ship tracks, not valleys. That is why the map shows only the large drainage routes.
four great river systems in the literature, flow traces on the map
When each road closed
How far must the sea fall for Sumatra, Java and Borneo to join? The map answers for each pair of places: it finds the highest sea level at which a land path exists between them, then turns that level into a date with both records and a ±5 m allowance for the sill.
The result can be checked against the literature. Voris (2000) writes that at −40 m the Peninsula, Sumatra and Java stayed joined, and that at −20 m the large islands were probably separate; Natuna stayed attached at −75 m and was isolated by −50 m. The map puts Natuna at −69 m. For Sumatra–Java the published sill is 20–40 m (Setiawan et al., 2015), shallower than the model's −45 m; for Java–Bali the 1.85 km grid does not resolve the strait, and the published sill is about 60–70 m.
Press a connection in the table: the sea drops below the sill and the road lights up. Then raise the sea with the staff and watch it break.
| Connection | Closes at | Date (years ago) | Published sill |
|---|---|---|---|
| Malay Peninsula – Sumatra | −30 m | 8,700–9,900 | ≈ −30 m (Shintani et al., 2011) |
| Singapore – Borneo | −35 m | 9,300–10,400 | ≈ −30 m / ≈ −36 m (Bird et al., 2006; Weiss et al., 2022) |
| Sumatra – Java | −45 m | 10,400–11,200 | −20 … −40 m (Setiawan et al., 2015) |
| Java – Borneo | −45 m | 10,400–11,200 | — |
| Sumatra – Bangka | −14 m | 7,200–8,000 | — |
| Borneo – Natuna | −69 m | 12,600–13,800 | −50 … −75 m (Voris, 2000) |
| Java – Bali | −19 m (grid) · ≈ −60 … −70 m (published) | 11,300–13,900 | ≈ −60 … −70 m |
The sea has dropped below the bridge. Raise it with the staff and watch the road break.
Sumatra separates from Java between 10,400 and 11,200 years ago
Savanna or forest? The evidence disagrees
Did the large animals of the new land have a wall of equatorial forest to cross? Bird, Taylor and Hunt (2005) propose an open savanna corridor, in their minimal scenario 50–150 km wide, running north-west to south-east between the wet forests of Borneo and the Malay Peninsula. Isotopes in bat guano support a savanna that spread at least 400 km on the Malay Peninsula (Wurster et al., 2010).
A vegetation model (Cannon, Morley and Bush, 2009) shows that at the ice-age peak there was more rainforest in total than today, because the shelf was dry land. Slik et al. (2011) identify the shelf's sandy soils as a barrier to forest dispersal, finding no evidence for a continuous savanna corridor, but without saying it did not exist.
The corridor is missing from the map: its extent has not been securely established. Forest and savanna can coexist: more forest overall does not rule out an open strip somewhere. The corridor's continuity, width and age remain in dispute.
one hypothesis on which four studies disagree
People on the lost land
At Sangiran on Java, Homo erectus most likely appears about 1.3 million years ago, and less than 1.5 million (Matsu’ura et al., 2020). The last known, from the Solo River, lived between 117,000 and 108,000 years ago (Rizal et al., 2020).
In 2025 the first hominin fossils from submerged Sundaland were described: two skull fragments from the Madura Strait, close to the Homo erectus of Java. The sediment was dated to 162,000 ± 31,000 and 119,000 ± 27,000 years; the bones were not dated directly. The place is a shallow sea floor, dry in periods when the sea fell.
Modern humans are on Sumatra, at Lida Ajer, between 73,000 and 63,000 years ago, according to the authors the earliest in a rainforest; on Borneo, at Niah, about 40,000 years ago; on Palawan, at Tabon, a tibia dated to about 47,000 years. Toward Sahul, Australia and New Guinea, there was no land bridge: several sea crossings were needed (Bird et al., 2018, 2019), and the first occupation of the Australian site Madjedbebe is dated to about 65,000 years ago, a date that is disputed.
The places on the map
73,000–63,000 years: the teeth of Lida Ajer
The slow flood: under a metre a century, with one leap
The sea did not come back at once. The shelf is so flat that after the sea had risen about 56 of its 116 metres, about 73% of the new land was still above water. Half is lost only around −46 m, about 11,100 years ago; the average rate over the whole climb was under a metre a century.
Between −96 m and −80 m, in the meltwater pulse of 14,600–14,300 years ago, the map loses about 130,000 square kilometres of land in three centuries, as much as Greece.
Scroll or press “Play the flood”: time runs from 21,000 years ago to today, and the map computes the shore for every level.
about 130,000 km² in one three-century pulse
The sea that never dried
Pull the sea all the way down and look at the eastern edge. Between Bali and Lombok the water stays. The sill of the Lombok Strait, about 300 metres, stayed sea through the depths of the glaciation: the sea had fallen by less than half its depth.
This line of deep water carries the name of Alfred Russel Wallace. West of it lies Sunda, with Asia's large mammals; east of it, Wallacea, then the world of Australia, with marsupials. Land animals had no bridge to cross on.
The table shows which large animals live in the wild on the three islands today. Of the three islands, the tiger survives only on Sumatra: the Java and Bali populations are gone, and none lives on Borneo now. Today's differences come from recent extinctions as well as from the sea.
| Sumatra | Borneo | Java | |
|---|---|---|---|
| Tiger | ● | — | — |
| Asian elephant | ● | ● | — |
| Orangutan | ● | ● | — |
| Sun bear | ● | ● | — |
| Sunda clouded leopard | ● | ● | — |
| Malayan tapir | ● | — | — |
300 m: the Lombok Strait sill, more than two and a half times the 116 m fall of the sea
Your city, on the drowned continent
Pick two cities. The map looks for the shortest road over the land that existed at the sea level on the staff and draws it. If the sea is too high, it tells you at what level the road would open and roughly when it closed.
Lengths are distances on a broken line across cells of about 11 km, without slopes, rivers or swamps. They give the order of magnitude, not a route to walk.
You can also tap the cities on the map.
22 cities, two maps of the same place