Identifying factors that shape ecological communities leads to better predictions about how communities respond to extreme events, yields empirical fodder to test theories (e.g., theory of island biogeography), and aids in the development of conservation plans that maximize ecosystem resilience. However, enumerating the richness and composition of communities is a difficult task, made more difficult by observation errors. Methods have been developed to account for false negative detection errors in the estimation of community metrics, but less attention has been paid to false positive detections. We investigate a special case of misidentification where pairs of species present in a community are likely to be confused with one another. This scenario is common when species with similar morphology (e.g., cryptic life stages) or vocalizations (e.g., similar echolocation patterns) are present at a single site. Accordingly, we developed a multi-species occupancy model that accounts for pairwise misclassification of species by utilizing an auxiliary data source where species can be identified without error. We evaluated the model using both simulated data and an amphibian community case study where certain pairs of species are difficult to differentiate in the larval life stage.
Results/Conclusions
When we failed to account for pairwise misidentification of species, species richness was overestimated and covariates influencing community size were masked. After accounting for misidentification of species pairs using our newly developed model, we found extreme variation in local amphibian community size (range = 1 to 12 species). In this system, both connectivity and conductivity were primary drivers of species richness. Our connectivity findings matched patterns expected based on the theory of island biogeography, where richness was associated with higher connectivity. Practically, our work suggests that maintaining the connectivity and water quality of wetlands will maximize amphibian species richness in our study area. Given that biodiversity is a benchmark many management agencies use to measure ecosystem health, our framework will be useful for a variety of settings and taxa. An extension of this model for multi-season data can be used to estimate dynamic processes of communities including species turnover rates.