Do unmeasured biotic interactions bias species distribution models?

Background/Question/Methods

One of the most promising developments in ecological methods in the last few years has been the development of species distribution models. These methods combine several types of easily obtained data with sophisticated statistical techniques to estimate the probability that an organism will be found in a given environment. While distribution models seem to hold a great deal of promise, their usefulness is limited by their incomplete characterization of natural environments. In particular, currently used models focus on climate and topography, but only rarely include information on biotic interactions between species. As it is often impossible to obtain enough data to explicitly model biotic interactions, it is important for us to develop an intuitive understanding of what we can learn from studying species distributions using only abiotic variables.
Results/Conclusions

Here we extend one of the most popular mathematical frameworks for studying biotic interactions –consumer resource models- to consider when we can accurately model the probability a species is present and when we can model the probability that an environment is suitable. We show that even if competition is important, statistical models that ignore this phenomena and focus on aspects of the abiotic environment may still provide accurate predictions for the probability that an organism will be present. When we ignore population sinks and stochasticity it is also possible to model the probability that an environment is a part of a species’ niche using only abiotic variables. These results provide a theoretical justification for using climate data to model species distributions, but indicate that models using climate data are not a reliable surrogate for an organism’s abiotic environmental requirements (its fundamental niche).