Statistical delimitation of biological species based on genetic and spatial data

The delimitation of biological species, i.e., deciding which individuals belong to the same species and whether and how many different species are represented in a data set, is key to the conservation of biodiversity. In the presence of spatial patterns of genetic differentiation, delimitation methods based on genetic data might overestimate the number of species in a dataset. This work tackles this problem in two settings. When individuals are divided into two putative groups, methods that model the relationship between genetic and geographic dissimilarity are used to test whether the two groups belong to the same species. Existing approaches based on partial Mantel testing and regression on distances are explored and new ones are proposed. A modelling challenge is connected to the fact that dissimilarities are not independent. All methodologies are compared through an extensive simulation study involving SLiM and GSpace, two different software packages that can simulate spatially-explicit genetic data at an individual level. A proposed version of the partial Mantel test that uses jackknife instead of permutations is found to provide fairly good power while controlling for the type I error rate in all simulated scenarios. In a setting where no putative grouping is available, existing model-based clustering algorithms (sNMF and TESS3) are integrated with distance-based approaches for the estimation of the number of species in the dataset. Further considered approaches use null models to calibrate tests for the presence of more than one species in the dataset. In particular, a weighted null model is developed that can capture spatial patterns of genetic differentiation. When calibrated with this null model, a test statistic proposed to adapt the ΔK method to TESS3 is found to display promising type I error and power properties on SLiM data.