An approximation approach to spatial connectivity for a data-limited endangered species with implications for habitat restoration

Among numerous concerns, restoration ecologists are routinely plagued with the problem of where to implement conservation efforts to best maintain spatial connectivity and population structure. Knowledge about connectivity within a metapopulation could offer valuable insight to address this issue and could help with the allocation of limited resources more effectively. However, direct estimation of dispersal is challenging because species can disperse widely within a landscape. Here, we developed a novel hierarchical Bayesian model to estimate spatial connectivity from occurrence data of an endangered stream fish, Topeka shiner (Notropis topeka). Our goal was to identify dispersal corridors that are centrally connected to the metapopulation that could be beneficial in decision making about future habitat restorations aimed at maintaining population structure. Connectivity modeling is data intensive and resource managers may not have the necessary data requirements; thus, we also examined the usefulness of graph theory (i.e. network centrality) as a proxy for connectivity. Model selection identified an upstream biased asymmetric dispersal pattern for the species. We were able to quantify and map connectivity and identified over 68 km of stream reaches as highly connected to the metapopulation. Probability of occurrence in dispersal corridors (i.e. streams) increased with connectivity and decreased with drainage area, highlighting the importance of conserving dispersal corridors and preferred habitat patches. Restorations in connected locations would provide critical habitat near important dispersal corridors. Betweenness centrality was positively correlated with connectivity and occurrence in restored habitat. Modeling of metapopulation connectivity and its correlation with graph theory demonstrated the usefulness of these techniques to guide conservation actions, especially in countries where data collection efforts are not common and conservation funding is limited.