Evaluation of consensus methods in predictive species distribution modelling

Abstract

International audienceSpatial modelling techniques are increasingly used in species distribution modelling. However, the implemented techniques differ in their modelling performance, and some consensus methods are needed to reduce the uncertainty of predictions. In this study, we tested the predictive accuracies of five consensus methods, namely Weighted Average (WA), Mean(All), Median(All), Median(PCA), and Best, for 28 threatened plant species. North-eastern Finland, Europe. The spatial distributions of the plant species were forecasted using eight state-of-the-art single-modelling techniques providing an ensemble of predictions. The probability values of occurrence were then combined using five consensus algorithms. The predictive accuracies of the single-model and consensus methods were assessed by computing the area under the curve (AUC) of the receiver-operating characteristic plot. The mean AUC values varied between 0.697 (classification tree analysis) and 0.813 (random forest) for the single-models, and from 0.757 to 0.850 for the consensus methods. WA and Mean(All) consensus methods provided significantly more robust predictions than all the single-models and the other consensus methods. Consensus methods based on average function algorithms may increase significantly the accuracy of species distribution forecasts, and thus they show considerable promise for different conservation biological and biogeographical applications

An improved species distribution model for Scots pine and downy oak under future climate change in the NW Italian Alps

“…The use of ensemble modeling is justified by the need to reduce model uncertainty due to different modeling approaches (Marmion et al 2003). Ensemble models in biomod2 are obtained by averaging model predictions and excluding models with low predictive power (AUC <0.75); model predictions are weighted by the AUC of their respective modeling algorithm.…”

Climate change threatens giant panda protection in the 21st century

“…A total of five GCMs were used to produce probability outputs for each scenario. We used the average predicted probability of occurrence across the five GCMs for each grid as our consensus forecast, which was one of best methods for developing an ensemble forecast (Hole et al, 2009;Marmion et al, 2009). Subsequently, we applied the average predicted probability as the threshold to define the presence-absence distribution of giant panda habitats, as this method has been found to be a robust approach (Liu et al, 2005).…”

Potential impacts of climate change on the winter distribution of Afro-Palaearctic migrant passerines

“…The consensus distribution was obtained with an ensemble forecast approach, by selecting the outputs of the five modelling techniques with the best AUC scores, and by further calculating the unweighted mean distributions (Marmion et al 2009; appendix 1, electronic supplementary material) for the corresponding 25 (5 models!5 GCMs) present or 60 (5 models!12 (GCMs! SRES)) future distributions.…”