Climate science

The resilience of Amazonian forests

Isotope evidence suggests that, during dry periods associated with the most recent ice age, the Amazonian forest survived in a region that is sensitive to rainfall changes — settling a debate about Amazonian aridity. See Letter p.204

A long-standing controversy exists over whether the most recent ice age was dry enough to cause savannahs to replace forests across much of Amazonia1,2. We know that Amazonia cooled by as much as 5 °C during ice ages3, but, crucially, we do not know by how much precipitation changed at the peak of the most recent ice age — the period known as the Last Glacial Maximum, about 24,000 to 18,000 years ago. On page 204, Wang et al.4 provide empirical data for the amount of precipitation over the past 45,000 years in a region of Amazonia that is especially sensitive to conversion to savannah. They also report evidence indicating that the forest in this region did not become savannah, even during the driest period of the most recent ice age.

Ice-age aridity in Amazonia was proposed to account for the origins of Amazonian biodiversity1. It was suggested that Amazonian rainforests were replaced by savannah during ice ages, and that forest fragments (refugia) remained only in the wettest locations. In these refugia, genetically isolated populations of plants and animals could have speciated. When wetter conditions returned, the range of Amazonian forests and of the newly formed species would have expanded to exclude savannah, thus forming the modern pattern of enriched biodiversity. Wang et al. set out to provide a climatic history testing the hypothesis of ice-age savannah expansion.

The authors' work focuses on samples collected from Paraíso Cave, which lies in one of the driest regions of Amazonia: the dry corridor (Fig. 1). They recovered seven stalagmites from the cave that, taken together, grew for a period spanning the past 45,000 years. Apart from a few brief gaps, the stalagmites provide a detailed history of conditions during the most recent ice age and the Last Glacial Maximum, and of the relative warmth of the Holocene epoch (which began some 11,700 years ago).

Figure 1: The dry corridor.

Mark B. Bush

The coloured region of the map shows the current average annual precipitation in Greater Amazonia (shown in millimetres per year; data from An area that usually receives less than 2,000 mm yr−1 is known as the dry corridor, and its ecology is especially sensitive to changes in precipitation. Wang et al.4 report isotope data from stalagmites collected in Paraíso Cave. The data provide empirical evidence for the amount of precipitation that fell in the dry corridor during the past 45,000 years, and indicate that the forest in this region did not become savannah during prolonged dry periods.

The relative abundance of oxygen isotopes (quantified by the parameter δ18O) in the stalagmites depends on δ18O in the drip water from which the stalagmites were produced. The δ18O of the drip water is, in turn, influenced by evaporation and condensation within the clouds and rain that formed the drip water. Higher δ18O values in the carbonate of a stalagmite can thus be interpreted to correlate with reduced precipitation. Similarly, δ13C values (which quantify the abundances of carbon isotopes) in the stalagmites provide a record of the dominance of forest versus grassland in the region, because they reflect contributions from the different carbon-isotope signatures of trees and grasses.

The composite stalagmite record shows a strong concordance between Amazonian rainfall and sea surface temperatures in the Atlantic Ocean — findings that are in line with geological records taken from caves at the foot of the Andes5. Most interestingly, Wang et al. find that, during the Last Glacial Maximum, the Amazonian landscape remained as forest even when precipitation above the Paraíso Cave fell to a level about 42% lower than modern rainfall at the site. The Amazonian forests were saved by cooling: the lower temperatures reduced evaporation rates from the forests and offset the loss of rainfall, so that the net effect was an approximately 20% decline in moisture availability.

Wang and colleagues' empirical data are consistent with evidence from Amazonian lake sediments, which indicate that the driest time in the most recent ice age was between 26,000 and 15,000 years ago6. However, lake records often contain a gap in sedimentation during this period, so knowledge about the climate conditions at that time is lacking. Wang et al. fill that gap with empirical data and highlight that the dry corridor — the landscape most vulnerable to change7 — remained forested during the period. Their data end the long debate over Amazon refugia.

Precipitation and temperature can interact to enhance or lower the probability of savannah expansion, but other factors, such as forest fires, can also have important roles. Amazonian forests are not fire-tolerant, because even a single fire causes tremendous tree mortality. This mortality generates dead wood, which increases the potential fuel load for subsequent fires and reduces shade, quickly causing the forest floor to dry and thus raising the probability of subsequent fires8. Multiple fires can rapidly transform a rainforest system into a savannah-like state.

Fire can transform a relatively wet forest into savannah, but in the absence of fire a lot of drying is needed for such a conversion to occur. Wang and colleagues' findings imply that fires must have been extremely rare for forest to have survived the net 20% drying that occurred over the course of the most recent ice age. This implication is supported by an analysis of sediments recovered from the Amazon delta9. The discovery that fire rarity contributed greatly to the maintenance of forest at that time has profound implications for future conservation of the region.

The Paraíso data also show that, during the most recent ice age, Amazonian plant and animal assemblages experienced a period of flux consisting of wet and dry cycles that each lasted up to a few thousand years. These cycles may have had particularly major consequences in the dry corridor that bisects Amazonia. This corridor was an important route that allowed species living in dry forest regions north and south of the Amazon basin to encounter each other. Phylogenetic studies of Amazonian plant and animal species indicate that there must have been connectivity across the Amazon basin within the past few million years, but the nature of that connection is debated10. Wang and colleagues' findings indicate that the millennial-scale wet–dry climatic oscillations would have caused the connection across this corridor to vary through time, and that the dry corridor could have supported dry-forest habitats, but probably not savannah.

A major distinction between ice-age and future conditions is that modern climate change takes place against a backdrop of elevated, rather than depressed, temperatures. Wang et al. suggest that future Amazonian climates responding to warmer sea surface temperatures will be wetter than modern climates. However, there is currently no accord regarding future Amazonian precipitation patterns in climate models. What is evident is that destabilized climates are increasing the frequency and severity of extreme drought and flood events. These extremes, rather than mean climate, may decide the future of the region.

Even so, Wang and colleagues' data offer hope that Amazonian forests will be resilient in a fire-free world, even if communities of rainforest species are not. Fire in modern Amazonia is almost exclusively a product of human activity, and shows marked increases in occurrence in dry years11. Our best path for buying time in the face of climate change is to promote land use that excludes fire — otherwise, destabilized climates will lead rapidly to habitat degradation.Footnote 1


  1. 1.

    See all news & views


  1. 1

    Haffer, J. Science 165, 131–137 (1969).

    ADS  CAS  Article  Google Scholar 

  2. 2

    Colinvaux, P. A., Irion, G., Räsänen, M. E., Bush, M. B. & Nuñes de Mello, J. A. S. Amazoniana 16, 609–646 (2001).

    Google Scholar 

  3. 3

    Colinvaux, P. A., De Oliveira, P. E., Moreno, J. E., Miller, M. C. & Bush, M. B. Science 274, 85–88 (1996).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Wang, X. et al. Nature 541, 204–207 (2017).

    ADS  CAS  Article  Google Scholar 

  5. 5

    Mosblech, N. A. S. et al. Nature Geosci. 5, 817–820 (2012).

    ADS  CAS  Article  Google Scholar 

  6. 6

    Bush, M. B. & De Oliveira, P. E. Biota Neotropica (2006).

  7. 7

    Bush, M. B. J. Biogeogr. 21, 5–17 (1994).

    Article  Google Scholar 

  8. 8

    Barlow, J. & Peres, C. A. Phil. Trans. R. Soc. B 363, 1787–1794 (2008).

    Article  Google Scholar 

  9. 9

    Maslin, M. A., Ettwein, V. J., Boot, C. S., Bendle, J. & Pancost, R. D. J. Quat. Sci. 27, 451–460 (2012).

    Article  Google Scholar 

  10. 10

    Honorio Coronado, E. N. et al. Divers. Distrib. 21, 1295–1307 (2015).

    Article  Google Scholar 

  11. 11

    Aragão, L. E. O. C. et al. Phil. Trans. R. Soc. B 363, 1779–1785 (2008).

    Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Mark B. Bush.

Related links

Related links

Related links in Nature Research

Hydrology: When wells run dry

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Bush, M. The resilience of Amazonian forests. Nature 541, 167–168 (2017).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing