Habitat loss has led to global declines in biodiversity and ecosystem functioning, and has decreased both the total amount of habitat, as well as altered how that habitat is arranged on the landscape. To combat this fragmentation, organisms must successfully disperse between disparate habitat patches, and this connectivity can allow for the maintenance of both genetic and species diversity in fragmented systems. One way to measure connectivity of fragmented landscapes is to use network modelling, where habitat patches are represented as nodes, and links between nodes denote successful dispersal. While network models have produced useful insights about which habitat patches play critical roles for connectivity, they are often built on the assumption that species disperse a fixed distance. In reality, populations of organisms disperse according to a dispersal kernel, with a higher probability that individuals travel a shorter distance from their natal patch, and a diminishing probability that individuals will travel much longer distances. To comprehensively understand habitat patch connectivity, we must incorporate realistic patterns of dispersal into network models. To do this, we use network models to determine the connectivity of over 37,000 grassland habitat patches in Minnesota, using both a fixed dispersal distance and dispersal kernels.
Results/Conclusions
We find that when we ignore individual variation in dispersal and simply consider dispersal as a fixed distance, we both over- and under-predict estimates of landscape connectivity. When examining network-level metrics that quantify connectivity of the grassland fragments as a whole, using a fixed dispersal distance consistently under-predicts connectivity. We find that fixed dispersal distances produce a smaller number of connected patches overall, and those patches that are connected tend to be less clustered (have fewer neighbors that are also connected to each other), as compared to those where dispersal incorporates individual variation through a dispersal kernel. When examining individual patch-level metrics, we find that in comparison to networks that incorporate a dispersal kernel, using a fixed dispersal distance tends to drastically over-predict the degree centrality and under-predict the closeness of each patch, which are both measures of how many neighboring patches could be reached from the given patch. Our results show it is necessary to incorporate these dispersal kernels into our estimations of connectivity because they incorporate variation in the probability of distances with which individuals might travel – and this can drastically affect which patches are connected to each other.