D.P. Larsen, and D. White. 2004. Estimating regional species richness using a limited number of survey units. Ecoscience 11(1)23-35. WED-02-038
Estimation of total species richness at large spatial scales is fundamental for both theoretical and applied ecology. Statistical prediction of regional species richness using a limited number of survey units can provide information about the current status of species diversity and about recent changes. The present study tested a recently developed method based on a relationship between species richness and the average Jaccard Coefficient between replicates. We compared the accuracy and precision of this new approach and a modified version with several established non-parametric estimators (Jackknife-1 and -2, Chao-2, ICE and Bootstrap) using four large regional data sets compiled by the Nature Conservancy for native fish and birds. Re-sampling techniques were used to generate samples of different sizes to examine the effect of sampling effort on accuracy and precision. We focused on the performance of the estimators at low sampling levels, which may be used in field surveys. Bootstrap method consistently underestimated regional species richness the most, although it was the most precise. Chao-2 yielded low precision and intermediate accuracy and ICE performed similarly. The modified version of the new approach was ranked first overall, followed by Jackknife-2. It was also found that the estimation of fish species richness was less accurate and less precise than the estimation of bird species in both regions because of more rare species. We addressed the potentials and limitations of the different estimators and the reason why overestimation can occur at high level of sampling effort.