Evaluating accuracy and precision of species-area relationships for multiple estimators and different marine...

This paper evaluates the accuracy and precision of six nonparametric estimators and six regression models for predicting the number of species in an area using random subsamples of that area. Such extrapolations are important for increased understanding of species-area relationships in different assemblages of species to inform ecological theories, and because they have direct implications in many conservation issues. Often, such extrapolations are evaluated by considering how well they fit existing data, rather than the actual number of species in the larger area. Also, many studies have focused on larger species in terrestrial habitats, where numbers of species are quite well known. Assemblages of invertebrates and smaller plants, especially in marine habitats, are less well known, and their distributions are dominated by small-scale variability. We examined species-area relations for two assemblages on boulders from six locations, for rocky shore assemblages in three locations, and for gastropods colonizing artificial habitat. Changing the focus of subsampling relative to the extent of the study and changing the size of the subsamples was also evaluated for subsets of data. Most estimators fitted the number of species poorly, with the second-order Jackknife consistently the best of the nonparametric measures, and the Negative Exponential the best of the regression models. Increasing the number of replicates improved most models, but some only slightly, and others considerably. Changing focus had little effect, probably because marine assemblages such as these vary at small scales as much as at many larger scales. These extensive analyses provide clear evidence about which estimators should not be used for measuring species-area relationships in assemblages such as these, and which will consistently over- or underestimate the number of species. The findings are applicable to many assemblages dominated by small organisms with strong stochastic variation.