COS 79-7
Inferring dispersal processes from occupancy patterns: Validation with parentage assignments in a rail metapopulation
Accurate dispersal estimates are crucial for understanding metapopulation connectivity and projecting the response of organisms to habitat changes driven by development and climate change. However, direct dispersal estimates are often difficult and costly to obtain, particularly for rare or secretive species. In metapopulations, dispersal distance can be inferred from patch occupancy patterns, based on presence-absence (detection-nondetection) data, using incidence function model (IFM) or buffer connectivity measures, but the accuracy of these approaches is poorly known. We compared expectations of dispersal distance from occupancy models to estimates derived using genetic data in a threatened and secretive marsh bird, the California Black Rail (Laterallus jamaicensis coturniculus). Multi-season occupancy models that account for imperfect detection were fit using buffer and IFM measures of connectivity, calculated at scales ranging from 1-30 km, as colonization covariates. Models were compared with AIC to determine the most supported scale of dispersal. The dispersal estimate from occupancy modeling was then compared to a dispersal kernel generated from geographic distances between first-order relatives identified with 19 microsatellite loci.
Results/Conclusions
Dispersal estimates for Black Rails using both buffer and IFM measures in occupancy models were similar. The best fit model using buffer measures, with 42% support in the model set, yielded a 6 km dispersal distance estimate for Black Rails. Averaging the top four models (with 90% support) from buffer measures of 3, 4, 5, and 6 km yielded a dispersal estimate of 4.90 km. The best fit model using IFM measures yielded a 2 km dispersal estimate, with 32% support in the model set. The top four models included IFM measures calculated at 2, 3, 4, and 5 km (with 83% support) and yielded a model-averaged dispersal distance of 3.02 km. Dispersal distances observed between 19 first-order relatives identified from genetic data were strikingly similar to estimates from occupancy models. The average distance between first-order relatives was 11.07 km (range: 0-80.07 km) and the median distance was 1.28 km. Our preliminary results indicate these methods are useful for estimating dispersal using easily collected occupancy information, even for species that are secretive or difficult to track, but further testing in additional species is needed.