Table 1 Nine categories of methods defined in this review that were used to build species distribution models
Modeling method | Definition (example) |
|---|---|
CTA | Classification tree analysis methods, including classification and regression trees, boosted regression trees, and random forest |
Ensemble | A weighted or unweighted average, or combination, of models built with different methods |
Envelope | Models that identify the boundaries of species’ ecological tolerance directly from data, without the use of machine learning (e.g., BIOCLIM, CLIMEX, CliMond, DOMAIN) |
GARP | Genetic algorithm for rule-set production, generates mathematical rules for estimating species presence |
MaxEnt | Maximum entropy, for expressing probability distributions |
Mechanistic | Process-based models, often using parameters of physiological limits to estimate distributions (e.g., species thermal limits) |
Mixed | Uses two or more methods to estimate species distributions, but does not average or combine output into a model ensemble |
Regression | Non-machine learning regression models (e.g., logistic regression, generalized linear models, generalized additive models, etc.) |
Other | Less commonly used methods that did not fit into another category (e.g., ecological niche factor analysis, environmental suitability thresholds, logic thresholds) |