SDM/ENM in R

Geum radiatum (zoomed prediction)
Habitat suitability for the endangered plant Geum radiatum

The enmSdm package for modeling niches and distributions in R is growing in utility and popularity! As of now it’s only available on GitHub (see the GitHub page for installation instructions), but eventually will be put onto CRAN.

Key features include:

  • A series of “train” functions for calibrating popular ENM/SDM algorithms, including:
    • Maxent (old and new versions)
    • Boosted regression trees (BRTs)
    • Generalized linear models (GLMs)
    • Generalized additive models (GAMs)
    • Natural splines (NS)
    • Random forests (RFs)
    • Conditional random forests (CRFs)
    • Least angle regression models (LARS)
    • A wrapper function to train any of these using cross-validation
  • Weighted versions of model evaluation statistics:
    • AUC
    • Continuous Boyce Index (CBI)
    • True Skill Statistic (TSS)
    • Fpb
    • Threshold calculation and threshold-dependent statistics like ORSS, SEDI, sensitivity, and specificity
  • Functions to thin points geographically and divide them deterministically into distinct folds/groups
  • Niche overlap functions using adaptively-contrained null model tests
  • Functions to calculate and make useful measures of spatial autocorrelation
  • Geographic utility functions (like getCRS, bioticVelocity, and coordPrecision)