Do bioclimate variables improve performance of climate envelope models?

Ecological Modelling
By: , and 



Climate envelope models are widely used to forecast potential effects of climate change on species distributions. A key issue in climate envelope modeling is the selection of predictor variables that most directly influence species. To determine whether model performance and spatial predictions were related to the selection of predictor variables, we compared models using bioclimate variables with models constructed from monthly climate data for twelve terrestrial vertebrate species in the southeastern USA using two different algorithms (random forests or generalized linear models), and two model selection techniques (using uncorrelated predictors or a subset of user-defined biologically relevant predictor variables). There were no differences in performance between models created with bioclimate or monthly variables, but one metric of model performance was significantly greater using the random forest algorithm compared with generalized linear models. Spatial predictions between maps using bioclimate and monthly variables were very consistent using the random forest algorithm with uncorrelated predictors, whereas we observed greater variability in predictions using generalized linear models.
Publication type Article
Publication Subtype Journal Article
Title Do bioclimate variables improve performance of climate envelope models?
Series title Ecological Modelling
DOI 10.1016/j.ecolmodel.2012.07.018
Volume 246
Year Published 2012
Language English
Publisher Elsevier
Publisher location Amsterdam, Netherlands
Contributing office(s) Southeast Ecological Science Center
Description 7 p.
Larger Work Type Article
Larger Work Subtype Journal Article
Larger Work Title Ecological Modelling
First page 79
Last page 85
Google Analytic Metrics Metrics page
Additional publication details