Dispersal is vital for many aspects of population survival. Yet theories of how dispersal may evolve and influence population ecology have become ever more complex and divorced from the empirical evidence necessary to test their predictions. Dispersal is commonly depicted as a “dispersal kernel”, a graph of the probability density distribution of dispersal distances in the population. Dispersal kernels are characteristically left-skewed and feature a tail to the right that may be long or short, thin or fat. This variation has long been assumed to have biological significance, frequently interpreted to indicate two dispersal phenotypes – “stayers” vs. “goers” – and a qualitative difference between long-distance and short-distance dispersal. This has not been adequately tested against empirical data, largely because lag-time between successive captures of individuals has rarely been considered. In a random model, longer distance movements should more likely occur over increasingly longer lag-times, whereas in a model of distinct dispersal phenotypes, the ratio of long- to short-distance movements should be relatively unaffected by lag-time. We investigated this using an extensive dataset of dispersive movements in an amphibian population. We subdivided this dataset according to lag-time and compared dispersal kernels to test for systematic changes in kernel shape.
Results/Conclusions
In 20 years of studies of Fowler’s Toads at Long Point, Ontario, Canada, we amassed a dataset of 6,288 movement distances performed by 2,593 individuals, with two to 17 captures per individual and lag-times between captures from one to 86 days. We found that these small, beach-dwelling, nocturnal, easily caught amphibians were individually identifiable and moved neither very far nor very fast. The animals moved parallel to the shore a maximum of 830 m in 24 hours. In accordance with the random model, the shapes of dispersal probability distributions plotted for differing lag-times differed systematically. Dispersal kernels displayed progressively lower peaks and longer tails with increasing tag-times. Median distance increased according to the 0.33 power of lag time. A series of power law functions, with parameters varying as the inverse of lag-time, was sufficient to describe movement probabilities of these animals. We conclude that dispersal can be a purely stochastic process and that dispersal kernel shape can be entirely the product of sampling regime with respect to lag-time. Thus the importance of dispersal kernel shape as a presumed fundamental characteristic of the dispersal process has been overestimated.
