Combining multiple models to predict the geographical distribution of the Baru tree (Dipteryx alata Vogel) in the Brazilian Cerrado.

The Brazilian Cerrado is a biome of great biodiversity, but detailed information about the diversity and distribution of species in this region is still insufficient for both testing ecological hypotheses and for conservation purposes. Among native plants in the Cerrado, Dipteryx alata Vogel (commonly known as the "Baru" tree), has a high potential for exploitation. The aims of this paper were to predict the potential spatial distribution of D. alata in the Brazilian Cerrado utilising five different niche modelling techniques. These techniques usually provide distinct results, so it may be difficult to choose amongst them. To adjust for this uncertainty, we employ an ensemble forecasting approach to predict the spatial distribution of the Baru tree. We accumulated a total of 448 occurrence points and modelled the subsequent predicted occurrences using seven climatic variables. Five different presence-only ecological niche modelling techniques (GARP, Maxent, BIOCLIM, Mahalanobis Distance and Euclidean Distance) were used and the performance of these models was compared using Receiver Operating Characteristics (ROC) and the Area Under the Curve (AUC). All models presented AUC values higher than 0.68, and GARP presented the highest AUC value, whereas Euclidean Distance presented the lowest. The ensemble forecasting approach suggested a high suitability for the occurrence of the Baru tree in the Central-Western region of the Brazilian Cerrado. Our study demonstrated that modelling species distribution using ensemble forecasting can be an important computational tool for better establishing sampling strategies and for improving our biodiversity knowledge to better identify priority areas for conservation. For the Baru tree, we recommend priority actions for conservation in the central region of the Cerrado Biome.