Imagine standing at the edge of what used to be forest. The trees are gone. In their place, a cold, dry plain stretches to the horizon, the soil pale and cracked, the air carrying nothing but dust and silence. This is what most of the world looked like 20,000 years ago, at the height of the Last ice age called Glacial Maximum, when the Earth’s climate locked into one of its deepest freezes.
Global temperatures had dropped by around 5 to 6 degrees Celsius. That might not sound like much until you consider what it did to the atmosphere’s capacity to hold moisture. Rainfall across the tropics collapsed. The great rainforests of Africa, South Asia, and much of South America shrivelled. What had been dense, layered canopy became fragmented scrubland or open savanna. The diversity of life that those forests carried, millions of species built up across tens of millions of years was compressed, isolated, and in many places simply lost.
But not everywhere.
Three regions held on. Three patches of forest, spread across different continents and shaped by completely different geographies, managed to stay warm enough and wet enough to keep going. They didn’t just survive passively, the way seeds survive in frozen soil. They remained functional, living, generating ecosystems. And when the ice finally retreated and the world began to warm again, they became the source from which life rebuilt itself across entire continents.
Those three forests are the Tropical Andes-Amazon complex in South America, the Sundaland forests of Borneo and Sumatra in Southeast Asia, and the Daintree Wet Tropics of northeastern Australia. Scientists call places like these glacial refugia. Understanding how each one survived , and what happened after changes how you think about forests, about biodiversity, and about what we stand to lose right now.
How the Tropical Andes Held On
The Andes mountain range runs like a spine down the western edge of South America, and during the last ice age, it turned out to be one of the most important pieces of geography on the planet.
While temperatures dropped and rainfall faltered across much of the Amazon basin, the Andes did something unusual: they kept their moisture. Research combining paleoclimate models with records preserved in cave mineral deposit shows that moisture from Atlantic trade winds continued flowing inland and rising up the mountain slopes throughout the glacial period . The mountains intercepted that moisture and held it. The forests at mid-elevation stayed wet when forests everywhere else were drying out.
But there was something else happening too. The steep elevation gradient of the Andes created what scientists now describe as a vertical escape route. When temperatures dropped, species didn’t have to migrate thousands of kilometres to find a viable climate. They moved up the slope. A few hundred metres of altitude shift is enough to find a meaningfully different temperature zone in mountain terrain. Birds, insects, plants, frogs; they tracked their preferred conditions vertically rather than horizontally, and the mountains gave them enough variation to work with.
The result was extraordinary. Rather than simply sheltering a reduced population of the species that had been there before, the Andes became an engine of new life. Each isolated valley, each ridge separated from the next by a slightly different altitude and microclimate, became its own small evolutionary chamber. Species that got separated on different slopes over thousands of years began to diverge. New varieties emerged. The area that acted as the ice age’s best shelter became, over the same period, the most species-rich place on Earth.
Today the Tropical Andes hosts over 30,000 species of vascular plants and nearly 2,000 bird species — roughly a quarter of all bird species on the planet (Conservation International). The survivorship of the Last Glacial Maximum didn’t just preserve what was there. It multiplied it.
Read more: Most Biodiverse Place on Earth: Why the Tropical Andes Leads the World
How Borneo and Sumatra Defied Expectations
The story of Sundaland which is the ancient landmass that encompasses Borneo, Sumatra, and the surrounding shallow seas of Southeast Asia begins with a counterintuitive discovery.
When scientists set out to model what happened to these forests during the Last Glacial Maximum, the expectation was fragmentation. Ice ages, in the standard model, cause rainforests to shrink and break apart into isolated pockets. That is what happened across most of the tropical world. In Sundaland, the models said otherwise.
A study tracking the distribution of 317 species of dipterocarp trees, the towering dominant family of Southeast Asian rainforests, across glacial and interglacial periods found that the forests of central Sundaland didn’t fragment during the Last Glacial Maximum. They expanded. The reason was that same 120-metre sea level drop that exposed continental land bridges elsewhere in the world. In Southeast Asia, it exposed vast areas of shallow continental shelf, creating new land that the forests marched across. Rather than retreating, they grew.
Northern Borneo was the heartland of this survival. While surrounding regions dried and contracted, the uplands of Borneo maintained dense forest cover. The equatorial position of these islands meant they were never exposed to the temperature extremes that hit forests at higher latitudes. Combined with the orographic rainfall generated by upland terrain, this created conditions stable enough to carry 130 million years of continuous forest growth through even the harshest glacial periods.
When sea levels rose again as the climate warmed, the newly populated shelf flooded. But the populations that had expanded across it didn’t simply vanish. They got isolated in the areas that remained above water, seeding new pockets of biodiversity across the archipelago. Scientists describe this process as a species pump: a cycle of expansion, isolation, and diversification that, repeated over millions of years, distributed Southeast Asian biodiversity across an enormous geographic range
Read more: Borneo Deforestation: What’s Being Destroyed and Why It Can’t Be Replaced
How the Daintree Waited Out the Cold
Australia’s Wet Tropics present a different kind of survival story, quieter, more contained, and in some ways more remarkable for it. The Daintree rainforest in northeastern Queensland is the oldest surviving tropical lowland forest on Earth, with lineages tracing back approximately 180 million years. That continuity means it was ancient long before the last ice age arrived. What makes the ice age chapter of its story unusual is how it survived given what it was working with.
Unlike the Andes, the Daintree had no elevation gradient extensive enough to serve as a migration corridor. Unlike Sundaland, it had no exposed continental shelf to expand across. Australia is an isolated continent, and the Wet Tropics sat at its edge with nowhere to go. Its species couldn’t retreat to somewhere warmer or wetter when conditions deteriorated. They had to hold on where they were, or not at all. climate modelling identified three specific pockets within the Wet Tropics bioregion where forest conditions remained viable, while the surrounding landscape shifted to savanna. The jagged peaks of the Great Dividing Range trapped the retreating rain clouds at precisely the right points. Three zones stayed wet enough. Everything else didn’t.
What happened inside those zones over tens of thousands of years was intense. Isolated populations with nowhere to move and no genetic exchange with other regions had to adapt or disappear. The species that survived became extraordinarily specialised. The Daintree today has the highest concentration of primitive flowering plant families found anywhere on Earth, representatives of lineages that diverged from the rest of the world’s flora during the age of the dinosaurs and have been evolving in isolation ever since. Its density of endemic vertebrates, species that exist nowhere else, directly corresponds to those three ancient refugia zones.
Read more: Daintree Rainforest: The World’s Oldest Tropical Rainforest Explained
What Happened After the Ice Retreated
When the Last Glacial Maximum ended roughly 12,000 years ago and global temperatures began to climb, something extraordinary happened across every continent where these refugia existed. Life came back. Not spontaneously, and not quickly by human standards, but with a directionality and source that scientists can now trace through genetic analysis of modern species.
From the Andes, the genetic signatures in cloud forest plants show them to be the source populations for post-ice age recolonisation of the broader Amazon basin. Bird species that today range across lowland South America carry genetic lineages pointing back to Andean populations. The mountains that sheltered them during the freeze became the staging ground for a repopulation that unfolded over thousands of years, species by species, valley by valley, until the Amazon recovered something close to its pre-glacial richness.
From Borneo and Sumatra, populations that had expanded across the continental shelf during the glacial period and been isolated by rising seas spread outward again as climates stabilised. Molecular studies of orangutans, Asian plant families, and tropical invertebrates consistently point to Bornean and Sumatran source populations as the origin of the biodiversity found across the broader Indo-Pacific region today.
From the Daintree, species expanded southward and outward in waves, repopulating the broader Australian rainforest zones that had contracted during the cold. You can still read the evidence of this in the pattern of Australian biodiversity today: species diversity in rainforest fauna declines as you move south, with the gradient pointing directly back to the Wet Tropics as the source.
Three forests. Three continents. Three completely different survival strategies. And then the same result: the rebuilding of entire ecosystems from the populations that made it through.
The Same Forests, A Different Threat
Here is where the story becomes urgent rather than simply remarkable.
The forests that survived the last ice age are being cleared right now at rates that would have been unimaginable to the climatic forces that shaped them. Borneo has lost an estimated 92% of its original primary forest cover, with Indonesia clearing around 264,000 hectares in 2024 alone, driven largely by palm oil and pulpwood operations. The Tropical Andes has lost roughly 75% of its original extent, with deforestation accelerating in Bolivia and Peru in recent years. The Daintree, while protected at its core by World Heritage listing, faces mounting pressure from climate shifts that are altering the rainfall patterns its refugia properties depend on.
The difference between an ice age and industrial deforestation isn’t just speed, though the speed is striking. An ice age changes conditions gradually enough that species can shift, adapt, find new pockets of stability. Clearance removes the habitat entirely. When a patch of primary forest in the Ucayali is cleared for agriculture, the species it contained don’t move somewhere else. Most of them disappear. The genetic lineages that survived 20,000 years of glacial cold get ended in an afternoon.
There is also the question of fragmentation. A refugium doesn’t function as a single point on a map. It functions as a connected system of habitats that allows species to move in response to changing conditions, exactly as they moved up and down Andean slopes during the last ice age. Fragment that system enough and the refugia properties disappear even if some trees remain. The forest is still there on a satellite image, but its capacity to shelter and regenerate life is gone.
The Forests That Rebuilt the World
What the science of glacial refugia reveals, when you follow it through, is something that reframes the entire conservation conversation. These aren’t just forests we’d prefer not to lose. They’re the specific places from which the world has repeatedly rebuilt itself after catastrophe. They are, in the most literal biological sense, the source code for the ecosystems that cover vast stretches of the planet today.
The Amazon you picture when you think of tropical rainforest grew back from the Andes. The orangutan populations of Sumatra and the archipelago beyond trace their ancestry to Bornean upland refugia. Australia’s rainforest fauna radiates outward from three small pockets in the Wet Tropics that happened to sit under the right mountains to stay wet when everything else dried out.
Knowing this changes what it means to protect them. Every hectare of primary forest preserved in the Tropical Andes, or in the uplands of Borneo, or in the Daintree, is a hectare of that regenerative capacity kept intact. The forests that outlasted the last ice age have earned a different kind of consideration — not just as habitat, not just as carbon storage, but as the planet’s proven mechanism for recovery.
They did it before. Given the chance, they can do it again.
Read more: Rainforest Refugia: What They Are and Why They Kept Life Alive
Sources
- Baker, P.A., Fritz, S.C., Battisti, D.S., et al. (2020). Beyond Refugia: New Insights on Quaternary Climate Variation and the Evolution of Biotic Diversity in Tropical South America. Neotropical Diversification: Patterns and Processes, Springer. https://doi.org/10.1007/978-3-030-31167-4_3
- Raes, N., Cannon, C.H., Hijmans, R.J., et al. (2014). Historical distribution of Sundaland’s Dipterocarp rainforests at Quaternary glacial maxima. PNAS, 111(47): 16790-16795. https://doi.org/10.1073/pnas.1403053111
- Wurster, C.M., Bird, M.I., Bull, I.D., et al. (2010). Forest contraction in north equatorial Southeast Asia during the Last Glacial Period. PNAS, 107(35): 15508-15511. https://doi.org/10.1073/pnas.1005507107
- Salles, T., et al. (2021). Quaternary landscape dynamics boosted species dispersal across Southeast Asia. Communications Earth and Environment, 2: 240. https://doi.org/10.1038/s43247-021-00311-7
- Hilbert, D.W., Graham, A., and Hopkins, M.S. (2007). Glacial and interglacial refugia within a long-term rainforest refugium: The Wet Tropics Bioregion of NE Queensland, Australia. Palaeogeography, Palaeoclimatology, Palaeoecology, 251(1): 104-118. https://doi.org/10.1016/j.palaeo.2007.02.020
- Graham, C.H., Moritz, C., and Williams, S.E. (2006). Habitat history improves prediction of biodiversity in rainforest fauna. PNAS, 103(3): 632-636. https://doi.org/10.1073/pnas.0505754103
- Rangel, T.F., Edwards, N.R., Holden, P.B., et al. (2018). Modeling the ecology and evolution of biodiversity: Biogeographical cradles, museums, and graves. Science, 361(6399): eaar5452. https://doi.org/10.1126/science.aar5452
- Arora, N., van Noordwijk, M.A., Ackermann, C., et al. (2010). Effects of Pleistocene glaciations and rivers on the population structure and genetic diversity of Bornean orangutans. PNAS, 107(50): 21376-21381. https://doi.org/10.1073/pnas.1010169107
- Williams, S.E. and Pearson, R.G. (1997). Historical rainforest contractions, localized extinctions and patterns of vertebrate endemism in the rainforests of Australia’s Wet Tropics. Proceedings of the Royal Society B, 264(1382): 709-716. https://doi.org/10.1098/rspb.1997.0101
- Conservation International CEPF. Tropical Andes Biodiversity Hotspot. https://www.cepf.net/our-work/biodiversity-hotspots/tropical-andes
