In a first classical approach, we include the lowland species probabilities of occurrence as covariates in the alpine species landscape scale models (covariate models). In a second novel approach, we first used SDMs to predict the distribution of the two alpine plants at the landscape scale. We then searched for interactive effects with the lowland species, and used this information to re-predict the landscape parts where alpine and lowland species were previously predicted to co-occur (abiotic + biotic models).
Our 鈥榓biotic + biotic鈥?model improved model precision for both alpine species; but statistically significantly for Viola biflora only. In contrast, the classical covariate approach did not affect the prediction accuracy of Viola biflora and decreased the prediction accuracy for Veronica alpina. This seemed to be caused by collinearity between abiotic and biotic predictors, highlighting potential problems with the conventional method used to account for biotic interactions in SDM.
Including potential effects of biotic interactions can improve predictions of alpine species鈥?ranges at the landscape scale. Ignoring biotic interactions in SDM may lead to biased predictions that are likely to overestimate realized climatic niches and so species distributions. The abiotic + biotic approach can constitute a robust method to account for biotic interactions in SDM.