Indirect connectivity estimates of amphibian breeding wetlands from spatially explicit occupancy models

• Published in: Aquatic conservation 2019-11, Vol.29 (11), p.1815-1825
• Main Authors: Brooks, George C., Smith, Jennifer A., Frimpong, Emmanuel A., Gorman, Thomas A., Chandler, Houston C., Haas, Carola A.
• Format: Article
• Language: English
• Subjects: Ambystoma bishopi; Amphibians; Aquatic reptiles; boosted regression trees; Breeding; Connectivity; conservation; Dispersal; Dynamic models; Dynamics; eigenvector mapping; Eigenvectors; Empirical analysis; Endangered & extinct species; Endangered animals; Endangered species; Evaluation; Freshwater fishes; habitat restoration; Habitat selection; Habitats; Landscape; Larvae; Management; Mapping; metapopulation; Metapopulations; Occupancy; Ova; Ponds; Rare species; Refugia; Reptiles & amphibians; Restoration; Spatial distribution; Threatened species; Vegetation; Wetlands; Wildlife conservation
• ISSN: 1052-7613; 1099-0755
• DOI: 10.1002/aqc.3190

In conservation science, metapopulation models are frequently used to explore the spatial dynamics of organisms, and ultimately guide management decisions for threatened and endangered species. Pond‐breeding amphibians are often touted as a classic example of metapopulation structure. However, empirical assessment of this assumption is lacking for many species owing to the prohibitively high costs of movement studies. Our aim was to use spatially explicit occupancy models to evaluate metapopulation dynamics in an endangered amphibian. First, dynamic, spatially explicit occupancy‐based metapopulation models were fitted to 8 years of presence–absence data for the endangered reticulated flatwoods salamander, Ambystoma bishopi. Models were constructed to evaluate the spatial distribution of salamanders across the landscape, rate of patch turnover, and connectivity between patches. Second, the same data were analysed adopting an eigenvector mapping approach and spatial correlograms to investigate habitat characteristics associated with occupancy state. The second approach also yields an independent measure of patch connectivity to corroborate findings from the dynamic model. Both models predicted a steep drop‐off in connectivity with increasing isolation, and no evidence was found of interactions between wetlands separated by distances greater than 1.5 km. The best predictor of salamander occupancy status was the amount of understorey vegetation in wetland basins. Taken together, the two modelling approaches provide complementary information regarding the key determinants of occupancy in the focal species. Furthermore, several results directly translate to management recommendations. Connectivity between distant sites is low, and thus management units should be demarcated accordingly. Understorey vegetation represents egg‐laying habitat and refugia for larvae, and hence should be targeted in future restoration efforts. The methodologies employed in this paper have broad applicability to studies of pond‐breeding amphibians, particularly in situations where direct measures of dispersal are unobtainable.