An empirical evaluation of the insurance hypothesis in diversity-stability models

An important stabilizing mechanism in most diversity stability models is the insurance hypothesis, which involves correlation/covariance relationships among species. These models require that species do not fluctuate synchronously over time: that is, the correlation between pairs of species does not equal 1.0. However, the strength of this stabilizing mechanism increases as correlations decline away from 1.0, especially as they become more negative and also as the summed covariance across all species pairs becomes more negative. We evaluated the importance of the insurance hypothesis as a stabilizing mechanism by examining a variety of terrestrial assemblages using long-term data from the Global Population Dynamics Database, the Breeding Bird Survey, and a long-term site in southeastern Arizona, USA. We identified co-occurring assemblages of species and calculated the Spearman rank correlations of all pairs of species and the summed covariance of the entire assemblage. We found that, in most assemblages, positive correlations were two to three times more common than negative and that the magnitude of the positive correlations tended to be stronger than the negative correlations. For all but three assemblages, the summed covariance was positive. Data from larger spatial scales tended to exhibit more positive correlations, but even at the smallest spatial scales, positive correlations outnumbered negative. We suggest that species often covary positively because coexisting species respond similarly to fluctuations in their resource base driven by climatic fluctuations. As such, our review suggests that the insurance hypothesis may not be a strong mechanism stabilizing fluctuations in natural terrestrial communities.