SPEDInstabR: An algorithm based on a fluctuation index for selecting predictors in species distribution modeling

Here, we present SPEDInstabR, available as an R package on CRAN and as an RWizard application on http://www.ipez.es/RWizard, which provides tools for the identification of the environmental factors that better discriminate between the conditions prevailing in the area of a species and those existing in the geographical background over which the study is carried out. This could include the world, countries, regions, river basins, etc. or the extent of occurrence of the species estimated by using convex hull, α-shape or Kernel density distributions. The procedure consists of dividing each factor into a number of intervals or bins decided by the user, calculating the number of records in each bin, separately considering the cells where the species occur and those of the selected geographical background. A peak of instability is observed when there are important differences in the factor comparing the bins of presence with the corresponding ones of extent. We consider that those factors with a higher percentage contribution to the Instability index better discriminate between the cells of presence and the extent. We tested the algorithm using virtual species, comparing the generated selections with those produced by MaxEnt.