An evaluation of patch connectivity measures

Measuring connectivity is critical to the study of fragmented populations. The three most commonly used types of patch connectivity measures differ substantially in how they are calculated, but the performance of these measures has not been broadly assessed. Here I compare the ability of nearest neighbor (NN), buffer, and incidence function model (IFM) measures to predict the patch occupancy and colonization patterns of 24 invertebrate, reptile, and amphibian metapopulations. I predicted that NN measures, which have been criticized as being overly simplistic, would be the worst predictors of species occupancy and colonization. I also predicted that buffer measures, which sum the amount of habitat in a radius surrounding the focal patch, would have intermediate performance, and IFM measures, which take into account the areas and distances to all potential source patches, would perform best. As expected, the simplest NN measure (distance to the nearest habitat patch, NHi) was the poorest predictor of patch occupancy and colonization. Contrary to expectations, however, the next-simplest NN measure (distance to the nearest occupied source patch, NSi) was as good a predictor of occupancy and colonization as the best-performing buffer measure and the general IFM measure Si. In contrast to previous studies suggesting that area-based connectivity measures perform better than distance-based ones, my results indicate that the exclusion of vacant habitat patches from calculations is the key to improved measure performance. I highlight several problems with the parameterization and use of IFM measures and suggest that models based on NSi are equally powerful and more practical for many conservation applications.