ON THE USE OF SOME METHODS FOR THE ESTIMATION OF SPECIES RICHNESS

The performance of three methods for the estimation of species richness are tested using computer simulations. Each method relies on different theoretical background: the method of Stout and Vandermeer is based on the species-area relation; Cohen's method assumes the log-normal distribution of species abundances, whereas the jackknife procedure is a non-parametric technique with no distributional assumptions. These methods were applied on a wide range of simulated community types. Factors considered in the generation of both communities and samples were: total species richness, community equitability, density and aggregation level of individuals on samples. Reliability of the estimators was measured as the ratio of estimated to actual richness. The results provide evidence for a consistent negative bias in all the estimators. Reliable estimations depend on the proportion of the whole community included in the sample, i.e. sample representativeness. Stout and Vandermeer's method shows the worst performance as it is strongly affected by aggregation and total species richness. Care is recommended when comparing different community types, but communities with similar attributes can be successfully compared using these methods.