The influence of scale on the importance of biotic interactions in species distribution models
Species distribution models (SDMs) are a primary tool used by ecologists in predicting species-specific distributional changes in the face of the ongoing biodiversity crisis. There is growing interest in improving the predictions of SDMs based on simple occurrence/environment correlations by incorporating ecologically relevant processes (e.g., dispersal, biotic interactions) into the models. However, the exact scales over which specific ecological processes operate have rarely been tested empirically and are still under intense controversy. We combine species occurrence data from the North American Breeding Bird Survey (BBS) with independently collected data on species interactions (e.g., competition, facilitation, predation). We focus on woodpeckers because these species interact in both positive and negative ways with each other and other bird species. We create a large pair-wise interaction matrix describing all known interactions among woodpeckers and their interacting species. We then establish the relative importance of biotic interactions as predictors of species occurrence in relation to other predictors such as environment and geographic proximity to other occurrence sites in a logistic mixed modeling framework. This framework allows us to assess how the relative importance of: (1) biotic interactions; (2) environmental affinity; and (3) geographic proximity to known species occurrences changes across both grain and extent.
Both environment and geographic proximity display clear scaling patterns. When study extent is large, we find the coefficients of these variables become higher as grain becomes coarser. This means that there is a higher chance of seeing a specific woodpecker species in a site if: (1) the environment is suitable; (2) the site is close to other sites in which the woodpecker was seen. Both positive and negative interactions display positive coefficients meaning that occurrence probability is higher when other species, which are known to interact with it, are recorded in the site. As opposed to environment and geography, positive biotic interactions show a clear decrease in coefficient estimates with increased grain. This means positive interactions have a smaller (but still positive) influence of species probability of occurrence as grain increases. In contrast, negative interactions do not show clear grain-dependence. In this study, we use detailed interaction data for North American woodpeckers to show that (1) biotic interactions (both positive and negative) are important predictors of species occurrences, although typically less important than environmental variables or geographic proximity, and (2) biotic interactions may remain important even at large grain sizes.