As part of a predictive modeling project of the “Quercus ilex” species, the generation of a habitat model has been carried out using the RStudio biomod2 package. Biomod has been configured for the calibration and evaluation of the applied models. These models correspond to the initials “GLM”, “GBM”, “GAM”, “CTA”, “ANN”, “SRE”, “FDA”, “MARS”, “RF”, “MAXENT”, “MAXNET”, and “XGBOOST”.

The graphical result of the dispersion of random points of absences generated by biomod in 10 replicates can be seen in the image, taking into account that it distributes points in areas that were not initially included in the presence records, that is, outside the Iberian Peninsula.

Next, a calibration of the models that have been configured in biomod is performed. A series of configurations are established, such as the percentage of training data (in this case, 70%), the number of replications (20) or the evaluation metrics, which will be True Skill Statistics (TSS) and Area Under the Curve (AUC-ROC). Once this process is finished, an evaluation of the models and their goodness of fit is performed, being the RF (Random Forest) and XGBOOST (Extreme Gradient Boosting) models the most suitable for the project requirements.