The implications of asymmetric dispersal for connectivity and colonization-extinction dynamics

Background/Question/Methods

Variation in movement has broad implications for evolutionary biology, community ecology and population dynamics.  In metapopulation ecology, movement provides the means to promote the persistence of the metapopulation and local populations over time.  Most metapopulation approaches quantify colonization and extinction dynamics based on the area-isolation paradigm.  Under this paradigm, patch area is negatively related to extinction rate, and isolation is negatively related to colonization. 

Oftentimes, isolation measures employ symmetric dispersal kernels in which the effective distance is assumed to be the same from patch i to j and from j to i (e.g. d_ij = d_ji).  Nevertheless, spatial heterogeneity and advection sources may cause effective distance to be asymmetric (e.g. d_ij ≠ d_ji).  Recent metapopulation theory hypothesizes that metapopulation models that assume symmetric dispersal may inaccurately estimate colonization and extinction because they underestimate isolation.  To test this hypothesis, we leverage a long-term time-series on colonization-extinction dynamics in the wind-dispersed orchid Lepanthes rupestris to test for asymmetric dispersal.  We assessed its implications for colonization and extinction dynamics by extending the isolation measure commonly used in incidence function models to account for dispersal asymmetries due to wind advection.  We compared model fit of two kinds of multi-season occupancy models: one that incorporated as a covariate the traditional isolation measure with symmetric dispersal kernel and another that instead incorporated an asymmetric dispersal kernel.

Results/Conclusions

Our results are consistent with the hypothesis that assuming symmetric dispersal underestimates isolation.  We found a general lack of fit of a model with the symmetric dispersal kernel parameterization (ΔAIC=1080.47) and a much better fit of a metapopulation model that incorporated asymmetric dispersal kernel (ΔAIC=0).  We also found that mean patch isolation was greater when using the asymmetric dispersal kernel.  Asymmetric dispersal may be the rule more than the exception in nature.  Therefore, our generalized approach has the potential to improve management and conservation practices by accounting for these asymmetries.