Drawbacks to palaeodistribution modelling: the case of South American seasonally dry forests
Species distribution modelling (SDM) has increasingly been used to predict palaeodistributions at regional and global scales in order to understand the response of vegetation to climate change and to estimate palaeodistributions for the testing of biogeographical hypotheses. However, there are many sources of uncertainty in SDM that may restrict the ability of models to hindcast palaeo‐distributions and provide a basis for hypothesis testing in molecular phylogenetics and phylogeographical studies.
Seasonally dry forests (SDFs) in South America.
We addressed the problem of using palaeodistribution modelling for SDFs based on the projection of their current distribution into past environments (21 ka) using 11 methods for SDMs and five coupled atmosphere–ocean global circulation models (AOGCMs) for 16 species.
We observed considerable uncertainty in the hindcasts, with the most important effects for AOGCM (median 12.2%), species (median 15.6%) and their interaction (median 13.6%). The effects of AOGCMs were stronger in the Amazon region, whereas the species effect occurred primarily in the dry areas of central Brazil. The log‐linear model detected significant effects of the three sources of variation and their interaction on the classification of each map in supporting alternative hypotheses. An expansion scenario combining the Pleistocene arc and Amazonian expansion, and Pennington's Amazonian expansion alone, were the most frequently supported palaeodistribution scenarios.
As a basis for evaluating a given hypothesis, hindcast distributions must be used in direct association with other evidence, such as molecular variation and the fossil record. We propose an alternative framework concerning hypothesis testing that couples SDM and phylogeographical work, in which palaeoclimatic distributions and other sources of information, such as the pollen fossil record and coalescence modelling, must be weighted equally.
No Supplementary Data
No Article Media
Document Type: Research Article
Publication date: 01 February 2013