New records and modelling the impacts of climate change on the black-tailed marmosets

permanent license for the collection of zoological material (Number: 6825–1) from the Chico Mendes Institute for Biodiversity Conservation—ICMBio, on behalf of the Museu Paraense Emilio Goeldi, through the Biodiversity Authorization and Information System—SISBIO. Thus, for the execution of this research and the eventual disclosure of data, we are supported ethically and legally. As specimens studied were road-kill victims, ethics committee permission to sacrifice the animals was not required since, according to Brazilian legislation, this kind of collection does not require other legal documents, just the possession of a generalized license for the collection of zoological material. Ecological Niche Modeling (ENM) Environmental variables. The literature recommends a cautious approach to choosing the size of the area to be analysed, as this can greatly affect the performance of the models [29, 30]. Size of chosen area must also be commensurate with study aims: accordingly, our models were generated for the entire neotropical region [31], since we aimed to identify all new suit- able sites with for the target species, so using a more circumscribed area could limit the identi- fication of such areas. To produce the ENM, we obtained 19 bioclimatic variables from the WolrdClim 2.0 [32] database. We also used ten edaphic and topographic variables, that is, six variables with soil physical properties (depth to bedrock, bulk density, clay, coarse, sand, and silt) from SoilGrids database (https://soilgrids.org/) and three topographic variable: elevation, obtained from the

EarthEnv database (http://www.earthenv.org/); and aspect and slope, obtained using the ter- rain function of the raster package from R [33, 34] (see S1 Table in S1 File). The variables were clipped for the neotropical region using a shapefile mask (available at http://ecoregions2017. appspot.com/), and then were resampled at a 5-minute arc-sine resolution (~ 10k). To test for collinearity between the selected variables, we performed two Principal Compo- nent Analysis (PCA). PCA 1 was generated for the bioclimatic variables, from which we selected the first six axes as model predictors, since these explained ~ 95% of data set variation over the neotropical region. We performed a second PCA (PCA 2) for the edaphic and topo- graphic variables, selecting the first two axes, which represented 66% of the variation within the same region. We attributed less weight to the variables in this second PCA because biocli- matic variables form the central focus of the current study. We generated projections for the bioclimatic variables, to 2070, under two different green- house gas emission scenarios generated by the Intergovernmental Panel on Climate Change (IPCC): a more optimistic scenario (SSP2-4.5), and the more pessimistic (SSP5-8.5). We chose 2070 as the period in which to understand the effects of climate change on the distribution of this species to provide an extensive, but still tangible, period of time, and to reduce the effects of uncertainties arising from climate projections beyond 2090 [4].

These simulated variables were produced based on the projections of the Coupled Model Intercomparison Project Phase 6 (CMIP6) [33]. The CMIP6 provides different scenarios based on the socio-economic pathway chosen (SSP) [34]. We did not run a new PCA on the variables of the future, as this would cause a loss of correspondence between the PCA axes produced for the current and the future variables. Using the PCA Projection function of the R package ENMGadgets [5], we produced a PCA using current variables, and designed the linear coeffi- cients of the current PCA using the future variables. This controlled for the percentage of explanatory power of each variable for each PCA axis in both the present and the future with- out losing correspondence between the PCAs [35]. PLOS ONE Modeling of Mico nigriceps PLOS ONE | https://doi.org/10.1371/journal.pone.0256270 September 7, 2021 5 / 14