Species-area curves: The need to include zeroes

The species-area curve relates the number of species (S) present within a sample to the area (A) of the sample, usually by the power function S = CA(z). Values of S may include zero, since this is within the normal range of variation of S. It seems likely that because logarithmic transformation is frequently used to estimate the species-area relationship, samples with S=0 have often been excluded from analysis because log(0) is undefined. Such censorship of the data can be expected to bias both C and z. In this paper I show that in calculating the parameters of the species-area relation, some authors have erred by excluding islands for which S=0. I also draw attention to an alternative model of the species-area relation that is suitable for data containing samples with S=0.