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).