XVI International Botanical Congess

We simulated how the four non-parametric richness estimators Jack1, Jack2, Chao2 and ICE are affected by spatial heterogeneity and by different abundance distributions. We fitted three abundance models to data from four agricultural habitat types, and explicitly modeled spatial autocorrelation, a gradient and an edge effect. For reasonably complete samples, Jack1 and Jack2 were positively biased and ICE was negatively biased. The estimators differed markedly in their variance. All estimators failed for the geometric model and were considerably affected by an edge effect. The estimators assume a random sample from a discrete, spatially homogeneous community that occupies a discrete area. Spatial structures and abundance distribution affect sample representativeness, which may be increased by adapting the sampling design. We recommend using Jack1 or ICE for samples containing at least 80 % of the species.