Comparison and Parallel Implementation of Alternative Moving-Window Metrics of the Connectivity of Protected Areas Across Large Landscapes

Abstract

Abstract Context. A variety of metrics can be used to measure connectivity of protected areas. Assumptions about animal movement and mortality vary among metrics. There is a need to better understand what to use and why, and how much conclusions depend on the choice of metric. Objectives. We compare selected raster-based moving-window metrics for assessing the connectivity of protected areas to natural habitat in the surrounding area, and develop tools to facilitate calculation of these metrics for large landscapes. Methods. We developed parallel implementations of distance-weighted sum and Spatial Absorbing Markov Chain methods in R packages to improve their useability for large landscapes. We investigated correlations among metrics for Canadian protected areas, varying background mortality, cost of movement, mean displacement, dispersal kernel shape, distance measure used, and the treatment of natural barriers such as water, ice, and steep slopes. Results. At smaller spatial scales (2–5 km mean displacement), correlations among metric variants are high, suggesting that any of the metrics we investigated will give similar results and simple metrics will suffice. Differences among metrics are most evident at larger spatial scales (20–40 km mean displacement) in moderately disturbed regions. Assumptions about the impact of natural barriers have a large impact on outcomes. Conclusions. In some circumstances different metrics give similar results, and simple distance-weighted metrics likely suffice. At large spatial scales in moderately disturbed regions there is less agreement among metrics, implying that more detailed information about disperser distribution, behaviour, and mortality risk is required for assessing connectivity.