Pseudoabsence generation
As the occurrence data available was presence-only, pseudoabsences were generated for use in the ensemble modelling. Broadly following recommendations from Barbet-Massin et al. (2012), 1000 pseudoabsences were specified as the minimum amount to be generated for each species. To balance sample sizes of presences and pseudoabsences for species with few occurrence localities, we generated multiple replicates of pseudoabsences where each replicate had equal number of presences and pseudoabsences. The number of replicates per species was calculated by dividing 1000 by the number of presences rounded up to the nearest integer. This ensured a minimum of 1000 pseudoabsence points for all species and integrated uncertainty in model results via replication for species with fewer presence localities. Pseudoabsences were generated randomly requiring a minimum distance of 2.5 km and a maximum distance of 1000 km away from occurrence localities. Because data availability varied among species, analyses included from 3 to 200 replicates (Supplementary Material 2).