Distinguishing effects of area per se and isolation from the sample-area effect for true islands and habitat fragments

The island species area relationship (ISAR) is an important tool for measuring variation in species diversity in variety of insular systems, from true-island archipelagoes to fragmented terrestrial landscapes. However, it suffers from several limitations. For example, due to the sample-area effect, positive relationships between species and area cannot be directly interpreted as evidence for deterministic effects of area per se. Additionally, richness-based analyses may obscure species-level responses to area and isolation that may better inform conservation practice. Here, we use random placement models to control for variation in abundance, occupancy and richness associated with the sample-area effect, allowing deterministic effects of area and isolation, and how they vary with species' functional traits, to be resolved using linear mixed effects models. We demonstrate the utility of this approach using a butterfly assemblage persisting on a naturally fragmented landscape of lake islands. The ISAR did not significantly deviate from random placement in relation to island area, isolation or habitat diversity, supporting stochastic assembly consistent with the sample-area effect. Such inferences support the habitat amount hypothesis, which prioritizes preserving the maximum amount of habitat irrespective of its degree of fragmentation. However, species-level analyses demonstrated that species' abundances were significantly lower on both smaller and more isolated islands than what is predicted by the sample-area effect. Moreover, effects of area per se were significantly greater for smaller, less mobile and rare species. Species' occurrences also significantly deviated from predictions of the sample-area effect in relation to island isolation. Thus, our approach illustrates that richness-based analyses not only result in incorrect inferences on mechanisms underlying ISARs, but also obscure important effects of area per se and isolation on individual species that vary with functional traits. We therefore suggest that these effects should not be solely inferred from richness-based analyses, but rather evaluated on a species-by-species basis.