The diverse nature of island isolation and its effect on land bridge insular faunas

Aim Isolation is a key factor in island biology. It is usually defined as the distance to the geographically nearest mainland, but many other definitions exist. We explored how testing different isolation indices affects the inference of impacts of isolation on faunal characteristics. We focused on land bridge islands and compared the relationships of many spatial and temporal (i.e., through time) isolation indices with community-, population- and individual-level characteristics (species richness, population density and body size, respectively). Location Aegean Sea islands, Greece. Time period Current. Taxon Many animal taxa. Methods We estimated 21 isolation indices for 205 islands and recorded species richness data for 15 taxa (invertebrates and vertebrates). We obtained body size data for seven lizard species and population density data for three. We explored how well indices predict each characteristic, in each taxon, by conducting a series of ordinary least squares regressions (controlling for island area when needed) and a meta-analysis. Results Isolation was significantly (and negatively) associated with species richness in 10 of 15 taxa. It was significantly (and positively) associated with body size in only one of seven species and was not associated with population density. The effect of isolation on species richness was much weaker than that of island area, regardless of the index tested. Spatial indices generally out-performed temporal indices, and indices directly related to the mainland out-performed those related mainly to neighbouring islands. No index was universally superior to others, including the distance to the geographically nearest mainland. Main conclusions The choice of index can alter our perception of the impacts of isolation on biological patterns. The nearly automatic, ubiquitous use of distance to the geographically nearest mainland misrepresents the complexity of the effects of isolation. We recommend the simultaneous testing of several indices that represent different aspects of isolation, in order to produce more constructive and thorough investigations and avoid imprecise inference.