A multi‐method approach for analyzing hierarchical genetic structures: a case study with cougars Puma concolor

Abstract

Genetic data are increasingly used to describe the structure of wildlife populations and to infer landscape influences on functional connectivity. To accomplish this, genetic structure can be described with a multitude of methods that vary in their assumptions, advantages and limitations. While some methods discriminate distinct subpopulations separated by sharp genetic boundaries (i.e. barrier detection or clustering methods), other methods estimate gradient genetic structures using individual-based genetic distances. We present an analytical framework based on individual ancestry values that combines these different approaches and can be used to a) test for local barriers to gene flow and b) evaluate effects of landscape gradients through individual-based genetic distances that account for hierarchical genetic structure. We illustrate the approach with a data set of 371 cougars Puma concolor from a 217 000 km 2 study area in Idaho and western Montana (USA) that were genotyped at 12 microsatellite loci. Results suggest that cougars in the region show a complex, hierarchical genetic structure that is influenced by a local barrier to gene flow (an urban population cluster connected by high traffic volumes), different landscape features (the Snake River Plain, forested habitat), and geographic distance. Our novel approach helped to elucidate the relative influence of these factors on different hierarchical levels of population structure, which was not possible when using either clustering methods or standard genetic distances. Results obtained with our analytical framework highlight the need for multi-scale management of cougars in the region and show that landscape heterogeneity can create complex genetic structures, even in generalist species with high dispersal capabilities.

Genetic Diversity and Population Structure of Mesoamerican Jaguars (Panthera onca): Implications for Conservation and Management

“…As recommended by previous empirical studies (e.g., [51]), we employed hierarchical genetic structure analysis to detect cryptic patterns of genetic divergence. However, larger and more even sample sizes for some sites and more continuous sampling in general would be beneficial to confirm moderate levels of genetic subdivision detected for jaguars across Mesoamerica.…”

“…We chose the most likely K value by calculating the mean posterior probability, mean L( K ) [47] and delta K (Δ K ) statistic [49] for each K using POPHELPER [50] in R, version 3.2.4 [41]. To examine hierarchical genetic structure within the genetic clusters identified through STRUCTURE, we repeated Bayesian clustering analysis until no additional genetic subdivision was detected (e.g., [51, 52]). Since the Bayesian clustering algorithm in STRUCTURE assumes unrelatedness among sampled individuals [47], we also identified closely related jaguars (parent-offspring, full siblings) with ML-RELATE [53], and repeated the STRUCTURE analysis without including closely related individuals.…”

Bobcats (Lynx rufus) as a Model Organism to Investigate the Effects of Roads on Wide-Ranging Carnivores

“…This suggests that the magnitude of road effects may depend not only on time since construction, but also on other local landscape and environmental factors (including traffic volume). As we obtain additional tissue samples, we plan a hierarchical analysis (Balkenhol et al 2014) to examine such interactions. The spatial structuring of bobcat populations also revealed potential dispersal routes (Fig.…”

Range-Wide Snow Leopard Phylogeography Supports Three Subspecies

“…The Q for each individual was averaged across all ten STRUCTURE runs. To examine hierarchical genetic structure, we conducted additional Bayesian analysis within identified genetic clusters until no further genetic subdivision was detected, or inference was impossible due to low sample sizes (Balkenhol et al 2014;Wultsch et al 2016a). We also used a second Bayesian clustering approach that incorporates a spatially explicit model to generate priors as implemented in GENELAND 4.0.6 (Guillot et al 2005).…”