Vegetation cover of Brazil in the last 21 ka: New insights into the Amazonian refugia and Pleistocenic arc hypotheses

Abstract

Aim: The two main hypotheses about the Neotropical palaeovegetation, namely that of Amazonian refugia by Haffer and of the Pleistocene arc by Prado and Gibbs, are still constantly debated. We offer new insights on this debate using ecological niche modelling with combined climate-soil predictors to test both hypotheses, reconstruct the palaeovegetation of the Last Glacial Maximum (LGM; 21 ka) and Mid-Holocene (Mid-H; 6 ka) and indicate the configuration of refugia areas. Location: Brazil.Time period: Last 21 ka.Major taxa studied: Biomes. Methods:We modelled the environmental space of the 10 most representative biomes with the RandomForest classifier, using climate predictors from three atmospheric general circulation models (CCSM4, MPI-ESM-P and MIROC-ESM) and soil predictors, the same for the different situations. Based on the consensus among the models, we reconstructed the palaeovegetation cover for LGM and Mid-H and used fossil pollen sites to validate the reconstructions in a direct comparison.Results: The climate in the past was cooler and wetter throughout most of the territory. The Amazon basin region was the most affected by climate change in the last 21 ka, with equatorial rain forest retracting to refugia areas, while the tropical rain forest (with climatic preferences similar to the Atlantic forest) expanded in the basin. In southern Brazil, the mixed forest (Araucaria forest) shifted to lower latitudes, while the grasslands expanded. In most biomes, the greatest changes occurred in the ecotonal zones, supported by pollen fossils.Main conclusions: With regard to Haffer's hypothesis, the forests of the Amazonian lowlands retreated to refugia areas, while the colder and wetter climate of the basin created a favourable niche for another type of forest, instead of savanna. The advance of dry vegetation was restricted to ecotonal conditions, preventing the formation of a continuous Pleistocene arc, predicted by

From pampa to puna: Biogeography and diversification of a group of Neotropical obligate grassland birds (Anthus:Motacillidae)

“…There is no increase in the number of lowland lineages during the late Pleistocene, and the divergence between northern and southern populations of A. lutescens precedes the LGM, contradicting the refugium theory. The older origins of this split agrees with older splits found in many forest‐based taxa (Patel et al., ; Ribas, Aleixo, Nogueira, Miyaki, & Cracraft, ) suggesting one of the many alternative theories explaining Amazonian speciation (Arruda et al., ; Burkart, ; Bush & Oliveira, ; Colinvaux, ; Patel et al., ). Only A. lutescens, and to a limited extent A. nattereri, occur exclusively in the tropical lowlands, which may be caused by ecological or physiological limits to the expansion of Anthus into tropical lowland grasslands.…”

Similar but different: Revealing the relative roles of species‐traits versus biome properties structuring genetic variation in South American marsh rats

“…The most dramatic effect is across Amazonia, where the more northern population of H. sciureus tend to be genetically quite similar despite fairly large distances separating the sampled individuals (Figure ). This geographic area that individuals map based on their genetic makeup is consistent with an estimated area of habitat stability based on ENMs for the past and present (Figure ), which suggests a potential role of climatic shifts, and coincidentally is estimated to be less stable relative to the southern sampled populations (Arruda, Schaefer, Fonseca, Solar, & Fernandes‐Filho, ; Bush et al., ; Mayle, Burbridge, & Killeen, ; Urrego, Silman, & Bush, ). In contrast, in southern regions of the distribution of H. sciureus , there is a strong correspondence between the genetic and geographic position of individuals (Figure ), and their general genetic distinctiveness relative to the more northern populations suggests a history of relative regional isolation within the biome (e.g.…”

Impacts of landscape composition, marginality of distribution, soil fertility and climatic stability on the patterns of woody plant endemism in the Cerrado

“…In addition to the geomorphological processes during the Late Miocene, the Pleistocene climate is proposed as a key driver of taxa diversification in the Cerrado (Arruda, Schaefer, Fonseca, Solar, & Fernandes‐Filho, ; Bueno et al, ; Costa et al, ; Santos, Nogueira, Giugliano, & Colli, ; Werneck, Nogueira, Colli, Sites, & Costa, ). In general, climatically stable regions are considered to be good predictors of diversity at both interspecific (Jansson, ; Werneck, ) and intraspecific levels (Buzatti, Lemos‐Filho, Bueno, & Lovato, ; Carnaval & Moritz, ; Carnaval, Hickerson, Haddad, Rodrigues, & Moritz, ; Collevatti et al, ; Correa Ribeiro, Lemos‐Filho, de Oliveira Buzatti, Lovato, & Heuertz, ; Lima, Telles, Chaves, Lima‐Ribeiro, & Collevatti, ).…”

“…Large‐scale hypothesis testing for the identification of refugia has been conducted for several regions around the world, with important biogeographical outcomes (Svenning, Normand, & Kageyama, ; Thuiller, ). Mostly, single‐species patterns or simulated dataset models have been applied in past distributional reconstructions for the Cerrado (Arruda et al, ; Collevatti, Castro, Souza Lima, & Campos Telles, ; Lima, Telles, Chaves, Lima‐Ribeiro, & Collevatti, ; Werneck et al, ), whereas multiple species approaches are less common (Bonaccorso, Koch, & Peterson, ; Bueno et al, ; de Siqueira & Peterson, ). These studies addressed future climatic changes based on a conservation biology approach (de Siqueira & Peterson, ), simulated the history of Amazonian forest fragmentation and the potential occupation of open areas by Cerrado species (Bonaccorso et al, ) or reconstructed the distribution of woody savanna vegetation under past scenarios (Bueno et al, ).…”

“…Large-scale hypothesis testing for the identification of refugia has been conducted for several regions around the world, with important biogeographical outcomes (Svenning, Normand, & Kageyama, 2008;Thuiller, 2004). Mostly, single-species patterns or simulated dataset models have been applied in past distributional reconstructions for the Cerrado (Arruda et al, 2018;Collevatti, Castro, Souza Lima, & Campos Telles, 2012;Lima, Telles, Chaves, Lima-Ribeiro, & Collevatti, 2017;Werneck et al, 2012), whereas multiple species approaches are less common (Bonaccorso, Koch, & Peterson, 2006;Bueno et al, 2017;de Siqueira & Peterson, 2003).…”