Extrapolating inventory results into biodiversity estimates and the importance of stopping rules

Seven methods for predicting species diversity from inventory data were tested based on two model data sets. These data sets, derived from state automobile license plates observed in Mexico City and Lawrence, Kansas, had the advantage of providing known 'communities' to be sampled, allowing evaluation of different inference methods. Of the seven methods, those of Chao (1984 ), Clench ( Soberón & LLorente, 1993 ), and model M_th of CAPTURE ( Otis et al ., 1978 ) were the most robust. Error inherent in calculations based on raw data was reduced substantially using a series of bootstrap manipulations. We recommend that optimal design of inventories should include stopping rules based on precision of results rather than on effort expended, an approach that offers considerable advantages, in terms of both accuracy of results and efficiency of sampling efforts.