Associations between species can influence the goodness of fit of species distribution models: The case of two passerine birds

• Red backed-shrike (RBS) and corn bunting (CB) are typical farmland bird species.
• We performed species distribution models for RBS and CB on environmental variables.
• Correlation matrix among presence of bird show significant values between both species.
• The AUC of best model for RBS was increased adding occurrence of CB as predictor.
• Associations between species can affect the goodness of fit of species distribution models.

Species distribution models (SDMs) are numerical tools that combine observations of species presence or abundance with environmental data, in order to develop predictive estimates about species distribution. The main variables used as regressors on SDMs are environmental parameters (climate variables, land-uses typologies, landscape metrics, etc.). However, there are several aspects that can affect the goodness of fit of species distribution models. For example, species traits, presence of associated species and anti-predator behaviors may cause differential responses to the processes that control their distribution. Considering these kinds of factors should improve the performance of models.In this work, SDMs were performed for two bird species found associated during initial exploration of data, by mean of a correlation matrix among bird occurrence. Models were performed first on environmental variables and then on environmental variables plus the occurrence of associated species. The goodness of fit of SDMs was compared using the area under the curve (AUC) and the likelihood ratio test.Our results showed how the associations between bird species can affect the goodness of fit of species distribution models. Specifically, we documented a significant increase in the AUC of the best model for red-backed shrike when adding the occurrence of corn bunting as a predictor. Our findings suggest how species-specific models in applied ecology can be developed to improve the predictive power of SDMs.