Defining fine‐scaled population structure among continuously distributed populations

Abstract

Understanding wildlife population structure and connectivity can help managers identify conservation strategies, as structure can facilitate the study of population changes and habitat connectivity can provide information on dispersal and biodiversity. To facilitate the use of wildlife monitoring data for improved adaptive management, we developed a novel approach to define hierarchical tiers (multiple scales) of population structure. We defined population structure by combining graph theory with biological inference about dispersal capability (based on movement, gene flow, and habitat condition) and functional processes affecting movement (e.g. habitat selection across scales of landscape preferences). First, we developed least‐cost paths between high fidelity sites (habitat patches) using a cost surface, informed from functional processes of habitat characteristics to account for resistance of inter‐patch movements. Second, we combined the paths into a multi‐path graph construct. Third, we used information on potential connectivity (dispersal distances) and functional connectivity (permeability of fragmented landscapes based on selection preferences) to decompose the graph into hierarchical tiers of connected subpopulations, denoting the degree that dispersal affected population structure. As a case study, we applied our approach across the greater sage‐grouse (Centrocercus urophasianus) range, a species of conservation concern in western United States. We described the relative importance of local populations and where to potentially avoid landscape disturbances that may negatively affect population connectivity using centrality measures supported by graph theory, and we demonstrated close alignment of the resulting population structure with population densities. This method can be adapted for other species with site fidelity and used as a management tool to evaluate population trends and responses to landscape changes across different temporal and spatial scales.

Defining biologically relevant and hierarchically nested population units to inform wildlife management

“…A comparison of hierarchical population levels (cluster levels 8 and 11) of greater sage‐grouse ( Centrocercus urophasianus ) in the western United States informed from sage‐grouse population structure (O'Donnell et al, 2022a) and environment covariates (panels (a) and (b)) or no environment covariates (panels (c) and (d)) using the Spatial “K”luster Analysis by Tree Edge Removal (SKATER) clustering algorithm.…”

“…We used the hierarchical tiers of population structure, denoting degrees of connectivity, as constraint‐based rules during the clustering. SKATER natively uses a graph of Euclidean distances (least‐cost minimum spanning tree; LC‐MST) that may lack representation of habitat connectivity (O'Donnell et al, 2022a). Therefore, we substituted the LC‐MST used by SKATER with our LCP‐MSTs, representing the hierarchical tiers of sage‐grouse population structure.…”

“…Here, we employed important and extensive enhancements (supplemental S1 in Appendix S1) to the data and workflow for a range‐wide application across 11 western states (United States). Broadly, these included a newly standardized lek database (O'Donnell et al, 2021), methods to define population structure graphs (hierarchical population tiers; O'Donnell et al, 2022a), additional candidate habitat covariates, modifications to clustering workflows, and a significant increase in marked birds for evaluation. Because we transitioned from two disparate and non‐contiguous states to the complete sage‐grouse range, we eliminated problems with confining population units to jurisdictional state boundaries (O'Donnell et al, 2019a).…”

“…The resistance surface reflected costs of movements between neighboring leks based on habitat conditions preferred by sage‐grouse. For example, the resistance surface captured how sage‐grouse generally avoid traversing large mountain ranges and rugged terrain, dense forests, large water bodies, and salt flats but select for greater sagebrush cover (O'Donnell et al, 2022a: see Figure S1). Using the resistance surface and sage‐grouse lek locations, we generated least‐cost paths (edges) between neighboring leks (nodes) to define a fully connected graph (population structure; Figure 3) with a distance‐weighted cost value assigned to each edge (O'Donnell et al, 2022a).…”

“…We used active sage‐grouse leks to define the current population structure (breeding display grounds) and biological inferences. These inferences encompassed dispersal capabilities, genetic flow, seasonal habitat conditions, and functional processes (e.g., selection of habitat at multiple scales) affecting movements (O'Donnell et al, 2022a).…”