Biogeographic Distribution Patterns of South American Amphibians: A Regionalization Based on Cluster Analysis

Abstract

We devise a regionalization system based on the geographical distribution of 2,265 amphibian species in South America. We used range maps of amphibians to obtain a presence/absence data grid resolved to 9310 km 2 . Biogeographical regions were generated by submitting the dataset to k-means clustering combined with v-fold cross-validation. The boundaries of the three clusters generated by multiple runs of the analysis are congruent with broadly defined biome structure in South America: 1) the Andes, the Atacama desert, Patagonia, and subtropics including grassland in southern Brazil, Argentina and Uruguay; 2) Amazon forest; 3) Atlantic forest and the Cerrado-Caatinga-Chaco complex. A few runs further distinguished Atlantic forest and Caatinga biome from other drier, more open biomes. The variable most strongly associated with the distribution of clusters was species richness, but climate also had moderately strong explanatory power. The regionalization scheme based on clustering is less finely resolved than previous schemes generated by expert opinion and rates of endemism but provides a general overview of the biogeographic signal contained in the current distribution patterns of amphibian species.

Expected impacts of climate change threaten the anuran diversity in the Brazilian hotspots

“…Because different anuran species pools have been identified between and within the analyzed hotspots (e.g., Valdujo, Carnaval, & Graham, ; Vasconcelos, Rodríguez, & Hawkins, ; Vasconcelos et al., ), we expected that the main component structuring the patterns of beta diversity of both hotspots, irrespective of the climate scenarios, would be accounted for the species replacement along their geographic extents. These expectations were confirmed, but we also found different and interesting results for the climate change impacts on the anuran beta diversity in the AF and CER: (a) the anuran composition along the CER extension tends to be more homogeneous under the climate change scenarios, as indicated by the lowest beta diversity dissimilarity values across time in this hotspot, with an associated decrease in the importance of the turnover and nestedness values; (b) although the dissimilarity values of beta diversity in the AF are similar across time, which means that the degree of changes in the anuran composition along the AF is expected to be the same among the different time slices, the nestedness component has increased values in the climate change predictions while the turnover component decreases.…”

“…Because different anuran species pools have been identified between and within the analyzed hotspots (e.g., Valdujo, Carnaval, & Graham, 2013;Vasconcelos, Rodríguez, & Hawkins, 2011;, we expected that the main component structuring the patterns of beta diversity of both hotspots, AF anurans that resulted in species losses per cell (e.g., Lemes et al, 2014;Loyola et al, 2014;present study), which in turn have led more grid cells being subsets of other richer ones. Spatially speaking, there is no such a deep change in the distribution of beta diversity in the AF across time, so the use of beta diversity distribution patterns for conservation purposes in the AF (e.g., Legendre, 2014;Socolar et al, 2016) should not result in different strategies between current and future climate scenarios.…”

Biogeographic Distribution Patterns and Their Correlates in the Diverse Frog Fauna of the Atlantic Forest Hotspot

“…The frog fauna from the inland semideciduous forest is made up of a mix of typical Cerrado and AF species, most of which are widely-ranged species [7], [41]. Hence, the most common feature shared by the SEMID species is the fact that they are both less diverse [11] and more widely distributed compared to species from other clusters (see also Figure S2 and Figure S3).…”

Introducing the biogeographic species pool

“…Evaluation methods, such as the v folds technique, are used to explore a range of potential cluster sizes to fi nd the optimum number (Rueda et al 2010, Vasconcelos et al 2011. An evaluation of the performance of three diff erent modularity indices in detecting biogeographic regions was performed recently (Th é bault 2013); however, further work is needed do not require an a priori decision regarding the number of clusters.…”

“…Non-hierarchical clustering approaches require an a priori choice of cluster number. Evaluation methods, such as the v folds technique, are used to explore a range of potential cluster sizes to fi nd the optimum number (Rueda et al 2010, Vasconcelos et al 2011. Such techniques are also appropriate for hierarchical methods.…”