Predicting genetic biodiversity in salamanders using geographic, climatic, and life history traits

total, eighty-three species remained with an average of approximately 20 sequences per nomi- nal species (see S2 Data for a list of identifiers corresponding to the sequences used in this study). Species delimitation We used three methods of species delimitation to determine the number of genetic lineages present in our samples. The GMYC is a tree-based method that takes a phylogenetic tree as input and finds a point in the tree where branching changes from within to between species [48]. The ABGD [49] and ASAP [50] methods are distance-based delimitation methods that use pairwise genetic distances to establish the threshold between intra- and inter-species diver- gence. Because each method is based on a specific set of assumptions, it is best to use multiple methods and compare their results in order to achieve a more accurate delimitation [51]. By looking for concordance across methods, we can increase our confidence in the identified line- age boundaries and minimize the potential impact of bias introduced by any single method. While we report delimitation results from the genes COI and cytb for all methods, we used a consensus of these results—reflecting agreement among the GMYC, ABGD, and ASAP delimi- tation methods for both COI and cytb—for assessing the influence of geography, environment, and life history traits on predicting salamander genetic diversity. To estimate a species tree for input into the GMYC, we used BEAST v2.5.1 [52]. We used the default parameters except for conducting 100,000,000 million generations, sampling every

5,000, and setting the model of sequence evolution to GTR+I+G [53]. The log files were checked by eye using Tracer v1.7.2 [54]. ESS values were all over 1000 for both cytb and COI. We removed 10% as burnin and retained the maximum clade credibility tree using TreeAnno- tator. After checking that the tree was binary and ultrametric, we used the R package splits [55] to conduct GMYC analyses. In each case we used the single threshold model and all other default settings. We conducted both ABGD and ASAP delimitation analyses via their web por- tals (https://bioinfo.mnhn.fr/abi/public/abgd/abgdweb.html and https://bioinfo.mnhn.fr/abi/ public/asap/asapweb.html, respectively) using the default parameter settings. Predictor variables For each nominal salamander species, we examined numerous geographic, environmental, morphological, and life history variables to identify traits predictive of undescribed salamander diversity using a classification model based on our species delimitation results. A variety of predictor variables were collected, including geographic and environmental values derived from georeferenced locality data (see S3 Data). In addition, three life history traits were avail- able from AmphiBIO, a global database for amphibian ecological traits [56], for most of the species in our study: reproductive strategy (direct developing, larval phase), habitat (terrestrial, fossorial, aquatic, or some combination of these), and body size (total length). To supplement this dataset and fill in any missing trait values, we used AmphibiaWeb [57] and other online

sources (S4 Data). To extract species specific data related to its environmental distribution, we utilized 42 GIS data layers (see S4 Data for data layer details), meant to capture various aspects of the ecology and habitat of each species. These include all 19 BIOCLIM layers from the CHELSA database [58,59] at 1 km resolution, elevation [60], terrestrial habitat heterogeneity [61], global land cover classification [62], global river classification [63], disaster risk [64], and various indica- tors of seasonal growth 58–59]. In addition to traits meant to capture various ecological factors we also gathered data for several traits relating to human impact, in order to measure levels of human disturbance and activity to the species environment. While ecological factors directly PLOS ONE Predicting hidden lineages in salamanders PLOS ONE | https://doi.org/10.1371/journal.pone.0310932 October 18, 2024 4 / 20