Example of using ENMevaluate (library: ENMeval; v. 0.2.0) with models to predict several bumblebee distributions in Europe, modifying prevalence (MaxEnt parameter).
Created on: 6th October, 2015
Contact: Xavier Rotllan-Puig (xavi@rotllanpuig.cat)
Description: The aim of this script is to build several SDMs (with Maxent) with different combinations of those parameters that affect models' performance (i.e. regularization and features). The R library ENMeval allows evaluate which is the optimal combination of these parameters in order to improve model performance while limiting overfitting. The Akaike Information Criterion corrected for small samples sizes reflects both model goodness-of-fit and complexity and it is independent of the partitioning method because it is computed with the full set of presences. The model with the lowest AICc value (i.e. Delta_AICc = 0) is considered the best model out of the current suite of models. Big AUC_diff, equal to overfitted models.
Original data:
-presences_91_12_maxent_meters.csv <- presences of bumblebee species with more than 25 unique
records in Europe between 1991-2012 (Rasmont et al., 2015)
-table of prevalences
-explanatory variables to run Maxent
References:
(1) Muscarella, R., Galante, P. J., Soley-Guardia, M., Boria, R. A., Kass, J. M., Uriarte, M., Anderson, R. P. (2014) ENMeval: An R package for conducting spatially independent evaluations and estimating optimal model complexity for Maxent ecological niche models. Methods in Ecology and Evolution, 5: 1198–1205. doi: 10.1111/2041-210X.12261
(2) Rasmont P., Franzén M., Lecocq T., Harpke A., Roberts S.P.M., Biesmeijer J.C., Castro L., Cederberg B., Dvorák L., Fitzpatrick Ú., Gonseth Y., Haubruge E., Mahé G., Manino A., Michez D., Neumayer J., Ødegaard F., Paukkunen J., Pawlikowski T., Potts S.G., Reemer M., J. Settele, J. Straka, Schweiger O. (2015) Climatic Risk and Distribution Atlas of European Bumblebees. Biorisk 10 (Special Issue), 246 pp.
(3) Radosavljevic, A., Anderson, R. P. (2014), Making better Maxent models of species distributions: complexity, overfitting and evaluation. Journal of Biogeography, 41: 629–643. doi: 10.1111/jbi.12227