Amazonian and neotropical plant communities on glacial time-scales: The failure of the aridity and refuge hypotheses

Plants respond to Pleistocene climatic change as species, not as associations or biomes. This has been demonstrated unequivocally by paleobotanical data for temperate latitudes. In the far richer vegetations of the tropics species populations also fluctuated independently in response to climatic forcing, from their longlasting glacial states to the patterns of brief interglacials like the present and back again. We use pollen data to reconstruct the vegetation of the Amazon basin in oxygen isotope stages 3 and 2 of the last glaciation in order to measure how the plant populations of the Amazon responded to the global warming at the onset of the Holocene. We find that plant communities of the neotropics vent copious pollen to lake sediments and that this pollen yields powerful signals for community composition. Three continuous sedimentary records reaching through oxygen isotope stage 2 are available from the Amazon lowlands, those from Carajas, Lake Pata and marine deposits off the mouth of the Amazon River. All three records yield pollen histories of remarkable constancy and stability. By comparing them with deposits of equal antiquity from the cerrado (savanna) of central Brazil, we show that most of the Amazon lowlands remained under forest throughout a glacial cycle. This forest was never fragmented by open vegetation as postulated by the refugia hypothesis. Instead the intact forest of glacial times included significant populations of plants that are now montane, suggesting that the global warming of the early Holocene resulted in the expulsion of heat intolerant plants from the lowland forest. Pollen data from the Amazonian flank of the Andes and from Pacific Panama provide evidence that populations of these heat intolerant plants survive the heat of interglacials in part by maintaining large populations at cooler montane altitudes. Our conclusion that the Amazon lowlands were forested in glacial times specifically refutes the hypothesis of Amazonian glacial aridity. Accordingly we examine the geomorphological evidence for glacial aridity and find it wanting. Of the three paleodune systems reported for tropical South America, that of NE Brazil was active in the Holocene as well as the Pleistocene. Parts of NE Brazil were actually moister than now in late-glacial times. Paleodunes in the Pantanal have never been seen on the ground, and those in the Orinoco Llanos are undated and may bie of any age since the Tertiary. Arkosic sands in the Amazon fan deposits came from the Andean foothills or from down cutting by rivers and cannot be evidence of a former arid land surface. White sands of Amazonia formed as podzols, not by aeolian activity. Such Amazonian stone lines as have received critical scrutiny are concretionary pisolites in stratigraphic formations that are more than ten million years old. Although the Amazon was never arid, modeling cooler glacial tropics gives plausibility to a somewhat drier Amazon in glacial times, a concept given substance by pollen data for the movement of ecotones in Rondonia, by stream histories in the Bolivian Andes, and by evidence for lowered lake levels at Carajas and Lake Pata. But this reduced precipitation was never enough to fragment the forest in the Amazon lowlands themselves. Pleistocene mammals of the Napo river valley in Ecuador were able to live along the river system in a forested landscape. Our data suggest that the Amazon forests have been stable since the start of the Pleistocene, a fact that has contributed to the storage of vast diversity. The coming anthropogenic global warming and CO2 enrichment will add to the global warming already endured by Amazon biota in the Holocene. We think it possible that the expulsion from the lowland forests of heat intolerant species is already complete and that the forest property of maintaining its own microhabitat will allow the high species richness to survive more global warming, provided large enough tracts of forest are preserved. (C) 1999 Elsevier Science Ltd. All rights reserved.