The small‐island effect: empty islands, temporal variability and the importance of species composition

Abstract

AbstractAimTo evaluate the evidence for a small‐island effect (SIE) on insular floras, considering the effects of including islands without the species of interest, temporal variability and species composition.LocationSmall islands in the Andros, Exuma Cays and Abaco archipelagos of the Bahamas.MethodsAn information‐theoretic multimodel inference approach was used, comparing the fit of a linear model versus two breakpoint models to species–area data. Corrected Akaike's information criterion (AICc) scores were compared to determine the model(s) with the most support.ResultsStrong evidence for an SIE was found in the Andros and Exuma Cays archipelagos. At Andros, an SIE was always apparent, regardless of whether islands with no species were included or which time period was considered. In the Exuma Cays, the inclusion of islands with no species changed the inference of the existence of an SIE in one comparison, and the shape of the SIE (i.e. the best‐fitting model) in two comparisons. Data from the Exuma Cays indicated that the existence of an SIE, as well as the shape of the SIE, may vary over time. Evidence for an SIE for the Abaco archipelago was equivocal, although fewer data were available.Main conclusionsEven though it has been suggested that the exclusion of islands with no species may result in the erroneous detection of an SIE, the inclusion of islands with no species had little effect on the evidence for an SIE in this system. The finding that evidence for, and the shape of, the SIE can change over time suggests caution in interpreting SIE studies for systems that are not in equilibrium. Patterns of species composition as a function of island size support the hypothesis that disturbance events shape the habitat found on these islands, which limits the species pool that can survive on the smaller islands, resulting in an SIE.