Climate-based ensemble predictions and uncertainty metrics (TAI, TSD) for plant species in Victoria, southeast Australia (38) and the Himalayan Kingdom of Bhutan (12)

This collection includes each of the climate variables (across multiple source datasets), associated plant species distributions in Victoria (n=38) and Bhutan (n=12), and ensemble metrics generated as part of the manuscript titled 'Building robust predictions of plant species distributions using alternative climate datasets and an ensemble modelling approach supported by complementary measures of uncertainty' (Stewart et al. under review). Ensembles are calculated using modelled predictions of species distributions that have been fitted using alternative source climate datasets (n = 10). Each contributing surface represents an ensemble of 7 algorithms (artificial neural networks, boosted regression trees, random forests, generalised additive models, multivariate adaptive regression splines, classification tree analysis, and flexible discriminant analysis). The ensemble metrics presented include the Threshold Agreement Index (TAI; describes the degree to which model predictions agree with respect to an optimal classification threshold) and the Threshold-scaled Standard Deviation (TSD; the standard deviation across all predictions in the ensemble, penalised in regions where contributing models agree with respect to an optimal classification threshold). These metrics describe the congruence and uncertainty associated with the selection of alternative climate datasets; however, could be applied to other applications (e.g. niche overlap under climate change, alternative algorthms, etc.). These metrics are each calculated, per-pixel, using 10 contributing prediction surfaces.