Print PageA simple one-dimensional niche model has been shown to explain the structure of food webs, especially smaller ones. One might also expect a similar approach to explain the structure of antagonistic and mutualistic plant – animal bipartite networks, although nobody has yet tested such a model on these networks. We use a probabilistic niche model to predict the structure of 44 antagonistic, 40 parasitic, and 67 mutualistic bipartite networks. In addition to assessing model fit using standard statistical measures such as AIC and maximum likelihood, for the first time we ask whether the niche model correctly predicts the niche structure observed in empirical data, an idea that borrows from the study of “nestedness”.
Results/Conclusions
In all classes of networks the one-dimensional model predicted similar fractions of interactions between species, and the fraction of interactions predicted correctly decreases with the size of the network. Yet in a novel approach, we show that mean overlap of the one-dimensional niches is strongly correlated with empirical niche overlap and nestedness, even in large networks where pair-wise interactions are less well predicted by the one-dimensional niche model. That this is so shows there to be a fundamental difference between the ability to predict ecologically important features of the overall structure of a network such as niche overlap and the ability to predict individual links. These findings point to exciting applications of the niche model as a component of larger ecosystem models.
