Using spatial demographic network models to optimize habitat management decisions

Conservation and management activities are always constrained by finite resources. Therefore, decisions such as which sites to protect, whether existing habitat should be restored or whether new habitat should be created, and where on the landscape management efforts should be focused present difficult challenges. An overarching goal of many conservation or management plans is the long‐term persistence of populations, which is often dependent on functional connectivity and the maintenance of metapopulation dynamics. Graph theory and network approaches are frequently used tools for modeling functional connections between populations and between habitats. Less often, graph models are used to guide conservation or management decisions. We used spatial networks derived for an amphibian population to determine optimal locations to create new habitat, prioritize existing habitat for restoration, and determine the habitat most critical for maintaining connectivity within the existing metapopulation within Fort Leonard Wood, Missouri, USA, 2012–2014. Using data collected at 206 breeding ponds over 3 years, we constructed demographic networks representing the functional connectivity between ponds by dispersing ringed salamanders (Ambystoma annulatum). We incorporated uncertainty in key model parameters through Monte Carlo simulation, and used a graph‐theoretical parameterization of the metapopulation mean lifetime model to assess how changes in network structure affect persistence of the network. We conducted addition and removal experiments within our Monte Carlo simulations to rank and prioritize locations for pond creation, ponds for restoration, and ponds for preservation. Salamanders bred in 106 ponds in ≥1 year; 2.4% of ponds functioned predominantly as sources and 51% of occupied ponds functioned predominantly as sinks. The importance of a pond to the network was correlated with the number of emigrants dispersing from a pond, the number of ponds reached by these dispersers, and the frequency a pond functioned as a source. Creating new ponds at optimal locations increased the persistence of the network an average of 15.4% compared to randomly selected locations, whereas selective restoration of currently unoccupied ponds resulted in an average increase of 31.4% in network persistence, as compared to randomly selected unoccupied ponds. Through Monte Carlo simulation, we constructed biologically informed demographic connectivity networks for use as a spatial conservation or management planning tool. Although our approach was implemented with an amphibian and a breeding pond network, it is generalizable to any species occupying discrete habitat patches. © 2017 The Wildlife Society.