Aim 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 palaeodistributions and provide a basis for hypothesis testing in molecular phylogenetics and phylogeographical studies.Location Seasonally dry forests (SDFs) in South America. MethodsWe 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. ResultsWe 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.Main conclusions 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.

We investigated here the demographical history of Tabebuia impetiginosa (Bignoniaceae) to understand the dynamics of the disjunct geographical distribution of South American seasonally dry forests (SDFs), based on coupling an ensemble approach encompassing hindcasting species distribution modelling and statistical phylogeographical analysis. We sampled 17 populations (280 individuals) in central Brazil and analysed the polymorphisms at chloroplast (trnS-trnG, psbA-trnH, and ycf6-trnC intergenic spacers) and nuclear (ITS nrDNA) genomes. Phylogenetic analyses based on median-joining network showed no haplotype sharing among population but strong evidence of incomplete lineage sorting. Coalescent analyses showed historical constant populations size, negligible gene flow among populations, and an ancient time to most recent common ancestor dated from ~4.7 ± 1.1 Myr BP. Most divergences dated from the Lower Pleistocene, and no signal of important population size reduction was found in coalescent tree and tests of demographical expansion. Demographical scenarios were built based on past geographical range dynamic models, using two a priori biogeographical hypotheses ('Pleistocene Arc' and 'Amazonian SDF expansion') and on two additional hypotheses suggested by the palaeodistribution modelling built with several algorithms for distribution modelling and palaeoclimatic data. The simulation of these demographical scenarios showed that the pattern of diversity found so far for T. impetiginosa is in consonance with a palaeodistribution expansion during the last glacial maximum (LGM, 21 kyr BP), strongly suggesting that the current disjunct distribution of T. impetiginosa in SDFs may represent a climatic relict of a once more wide distribution.

Abstract.-Studies in biogeography and macroecology have been increasing massively since climate and biodiversity databases became easily accessible. Climate simulations for past, present, and future have enabled macroecologists and biogeographers to combine data on species' occurrences with detailed information on climatic conditions through time to predict biological responses across large spatial and temporal scales. Here we present and describe ecoClimate, a free and open data repository developed to serve useful climate data to macroecologists and biogeographers. ecoClimate arose from the need for climate layers with which to build ecological niche models and test macroecological and biogeographic hypotheses in the past, present, and future. ecoClimate offers a suite of processed, multi-temporal climate data sets from the most recent multi-model ensembles developed by the Coupled Modeling Intercomparison Projects (CMIP5) and Paleoclimate Modeling Intercomparison Projects (PMIP3) across past, present, and future time frames, at global extents and 0.5° spatial resolution, in convenient formats for analysis and manipulation. A priority of ecoClimate is consistency across these diverse data, but retaining information on uncertainties among model predictions. The ecoClimate research group intends to maintain the web repository updated continuously as new model outputs become available, as well as software that makes our workflows broadly accessible.