On stability measures and effects of data structure in the recognition of areas of endemism

Abstract

Incomplete data sampling, bias, and like properties of distribution datasets that potentially introduce uncertainty in biogeographical analyses and blur biogeographical patterns; therefore, it is important to understand their influence. Despite their relevance, these problems have been largely overlooked in biogeography, where concepts such as ambiguity, stability or support have not even been defined. Here, we propose two stability measures for hypotheses of areas of endemism (AEs) and use them to explore the degree to which different structural qualities of data affect the results of analyses of endemism. Our findings suggest that different types of data incompleteness have different effects on the recovery of the species composition and the geographical or spatial structure of AEs, showing that distinct levels of sampling coverage affect the stability of results in different ways. We show that a small proportion of poorly sampled species may have a stronger impact on AEs stability than many species with medium sampling and that excluding poorly sampled species from the analyses does not guarantee more stable results. These results highlight the importance of planning data collection and indicate that, in order to obtain more stable results, focusing on completing the distribution of strongly undersampled species might be preferable to adding records of any species randomly.