Estimating diversity in unsampled habitats of a biogeographical province

Abstract

Estimating the number of species in a biogeographical province can be problematic. A number of methods have been developed to overcome sample-size limits within a single habitat. We evaluated six of these methods to see whether they could also compensate for incomplete habitat samples. We applied them to the butterfly species of the 110 ecoregions of Canada and the United States. Two of the methods use the frequency of species that occur in a few of the sampled ecoregions. These two methods did not work. The other four methods estimate the asymptote of the species-accumulation curve (the graph of "number of species in a set of samples" versus "number of species occurrences in those samples"). The asymptote of this curve is the actual number of species in the system. Three of these extrapolation estimators produced good estimates of total diversity even when limited to 10% of the ecoregions. Good estimates depend on sampling ecoregions that are hyperdispersed in space. Clustered sampling designs ruin the usefulness of the three successful methods. To ascertain their generality, our results must be duplicated at other scales and for other taxa and in other provinces.