Performance metrics and variance partitioning reveal sources of uncertainty in species distribution models

Ecological Modelling
By: , and 



Species distribution models (SDMs) are widely used in basic and applied ecology, making it important to understand sources and magnitudes of uncertainty in SDM performance and predictions. We analyzed SDM performance and partitioned variance among prediction maps for 15 rare vertebrate species in the southeastern USA using all possible combinations of seven potential sources of uncertainty in SDMs: algorithms, climate datasets, model domain, species presences, variable collinearity, CO2 emissions scenarios, and general circulation models. The choice of modeling algorithm was the greatest source of uncertainty in SDM performance and prediction maps, with some additional variation in performance associated with the comprehensiveness of the species presences used for modeling. Other sources of uncertainty that have received attention in the SDM literature such as variable collinearity and model domain contributed little to differences in SDM performance or predictions in this study. Predictions from different algorithms tended to be more variable at northern range margins for species with more northern distributions, which may complicate conservation planning at the leading edge of species' geographic ranges. The clear message emerging from this work is that researchers should use multiple algorithms for modeling rather than relying on predictions from a single algorithm, invest resources in compiling a comprehensive set of species presences, and explicitly evaluate uncertainty in SDM predictions at leading range margins.

Publication type Article
Publication Subtype Journal Article
Title Performance metrics and variance partitioning reveal sources of uncertainty in species distribution models
Series title Ecological Modelling
DOI 10.1016/j.ecolmodel.2015.03.017
Volume 309-310
Year Published 2015
Language English
Publisher Elsevier Science B.V.
Publisher location Amsterdam
Contributing office(s) Southeast Ecological Science Center
Description 12 p.
First page 48
Last page 59
Online Only (Y/N) N
Additional Online Files (Y/N) N
Google Analytics Metrics Metrics page
Additional publication details