Pairwise measures of species co-occurrence for choosing indicator species and quantifying overlap

Abstract

One of the most important ecological relationships between any two species is the degree of overlap in their distributions, i.e., their co-occurrence. Quantifying this relationship is a key step in the selection of indicator species and many other analyses in conservation biology and ecology. We derived a measure of the co-occurrence of two species based on the relative mutual information (RMI) of their distributions, and then compared its performance to three existing statistics: bivariate or binary covariance (BC), mean pairwise index (MPI), and proportional similarity (PS). To make this comparison, we measured co-occurrence values for all pairwise combinations of species collected from three communities (ground-dwelling beetles, moths, and vascular plants) in the Jerusalem Mountains and Judean Foothills, central Israel. We then used these co-occurrence values to address two different ecological problems: the challenge of identifying good indicator species, and the question of whether congeneric species co-occur more than species from different genera. We found that PS and RMI were the most reliable basis for choosing indicator species, but these two statistics differed in their error structures: PS had lower rates of type I errors (false positives), while RMI had lower rates of type II errors (false negatives). We also found that congeneric species co-occurred more often than species from different genera, but this pattern was statistically significant for only some of the measures of co-occurrence. In our analysis, then, the conclusion that we reached regarding the co-occurrence of congeneric species depended on which co-occurrence statistic was used. We therefore caution that available co-occurrence statistics should not be used interchangeably, because the ecological inferences drawn from a study may depend on the choice of co-occurrence statistic. In summary, we recommend PS and RMI as pairwise measures of species co-occurrence for investigating the reliability of biodiversity indicators and other applications in conservation biology and ecology.