Figure legends
Fig. 1: Analytical scheme used to test whether evolution is faster at ecotones, which involved 1) calculating tip-based ancestral trait state and its change over time and 2) spatializing changes from the ancestral trait state using assemblage-level metrics (aTR, aST, aLT), and 3) propagating uncertainty across the previous steps (gray arrow in the background). To calculate tip-based metrics at the species level, we mapped and estimated ancestral states using stochastic mapping of discrete traits via Bayesian inference, which allows calculating the time at which a trait changed along phylogeny nodes. The tip trait state is taken into account when calculating TR (as seen for Sp. 1). Note that transitions not fixed at the nodes are not considered when calculating TR (e.g., the brief transitions between n1 to n2 from plant→ insect to insect→plant), although such brief transitions do reduce ST and LT. Also note that ST is the maximum time length between two nodes, and LT is the sum of branch lengths with reconstructed traits equal to the tip trait. Values of tip-based metrics are equal for sister species (Sp. 6 and 5, Sp. 4 and 3) because trait change occurred exactly in the same nodes.
Fig. 2: Density plots of the intercept (expected mean) of assemblage transition rates aTR, and regression coefficient (deviation from the mean) of the most important variables. In each plot, the intercept is represented by the gray line and the regression coefficient is represented by the black line. Estimates were extracted from Linear Mixed Models that consider ecoregion-scale variables as fixed effects, ecoregion ID as random effect, and exponential correlation structure with nugget effect to accommodate spatial autocorrelation. Intercept and regression coefficients were extracted from each one of the 2,000 models. Boxplot in the upper margin shows average and 1st and 3rd quartiles of the distribution of aTR.
Fig. 3: Density plots of the intercept (expected mean) of assemblage stasis time aST (millions of years), and regression coefficient (deviation from the mean) of the most important variables. In each plot, the intercept is represented by the gray line and the regression coefficient is represented by the black line. Estimates were extracted from Linear Mixed Models that consider ecoregion-scale variables as fixed effects, ecoregion ID as random effect, and exponential correlation structure with nugget effect to accommodate spatial autocorrelation. Intercept and regression coefficients were extracted from each one of the 2,000 models. Boxplot in the upper margin shows average and 1st and 3rd quartiles of the distribution of aST.
Fig. 4: Density plots of the intercept (expected mean) of assemblage last transition time aLT (millions of years), and regression coefficient (deviation from the mean) of the most important variables. In each plot, the intercept is represented by the gray line and the regression coefficient is represented by the black line. Estimates were extracted from Linear Mixed Models that consider ecoregion-scale variables as fixed effects, ecoregion ID as random effect, and exponential correlation structure with nugget effect to accommodate spatial autocorrelation. Intercept and regression coefficients were extracted from each one of the 2,000 models. Boxplot in the upper margin shows average and 1st and 3rd quartiles of the distribution of aLT.
Fig. 5: Mapped assemblage-level transition rates (aTR), stasis time (aST), and last transition time (aLT) of sigmodontine rodent assemblages at points in ecoregion cores and ecotones. Tip-based metrics in the left maps (A,C,E) were obtained by averaging metrics across 10,000 estimates (100 phylogenies, 100 simulations per phylogeny). Phylogenetic uncertainty on estimates of the tip-based metrics, represented in the right maps (B,D,F), were calculated through the standard deviation of the metrics across 10,000 estimates.
Data accessibility statement
All data we used are already available in online repositories. Range maps are available in the Dryad Digital Repository (http://dx.doi.org/10.5061/dryad.8vt6s95). Phylogenies were published in 2019 by N. Upham and collaborators in PLOS Biology (https://doi.org/10.1371/journal.pbio.3000494). A shapefile with ecoregions is available at https://www.worldwildlife.org/publications/terrestrial-ecoregions-of-the-world. A shapefile with Central Andes boundaries was published by Löwemberg-Neto in 2015, and it is available at http://dx.doi.org/10.11646/zootaxa.3985.4.9. A shapefile with Atlantic Rainforest boundaries was published by Muylaert and collaborators in 2018, and it is available at https://github.com/LEEClab/ATLANTIC-limits. Mammal diet data were published by Wilman and collaborators in 2014 and are available at https://doi.org/10.1890/13-1917.1. Finally, the R codes used to calculate the three new tip-based metrics will be available on the GitHub page of the first author.
Competing interest statement
We declare that there are no competing interests in relation to this study.
Author contributions
ALL, RM, SMH, and LDSD conceived the ideas and designed the methodology. RM and BDP provided occurrence and phylogenetic data. ALL and VJD wrote the R functions. ALL, RM, VJD, BDP, SMH and LDSD contributed to data analysis and wrote the manuscript. All authors contributed critically to the drafts and gave final approval for publication.
Acknowledgements
ALL received a PhD fellowship from the Brazilian Federal Agency for Support and Assessment of Post-Graduate Education (CAPES). LD and SMH received funding from the National Council for Scientific and Technological Development (CNPq; proc. 307527/2018-2 and 304820/2014-8, respectively). We thank Gabriel Nakamura and Arthur Rodrigues (UFRGS) for their suggestions during study development and data analysis. We thank Vinicius G. Bastazini (UFRGS), Maria João Pereira (UFRGS), Augusto Ferrari (FURG), Adriano S. Melo (UFRGS), Fernanda T. Brum (UFPR), Marcus V. Cianciaruso (UFG) and Ricardo Dobrovolski (UFBA) for discussions and suggestions in previous versions of the manuscript.