Distribution data
Global distribution data for GAS was downloaded from GBIF (ww.gbif.org).
Only records which were labelled as “human observation” or
“occurrence” were retained. These records were then filtered to remove
any coordinates with high levels of uncertainty. Data on the presence of
GAS collected during this current study were then added to the cleaned
GBIF records. This represented our overall dataset, which consisted of
3219 records. This dataset was filtered so that only one presence was
recorded in each climatic grid-cell, resulting in a working dataset of
730 distributional records.
Statistical Species Distribution Models (SDMs) require information on
where a species is absent. Often there is insufficient verified data,
and thus “pseudo-absences” must be used. To allow us to use, and test
the predictive accuracy, of statistical SDM methods, ten sets of
pseudo-absences were sampled. Each of the sets of pseudo-absences were
restricted so that they were always within 500km of a verified GAS
location, but were outside of a grid cell occupied by a presence
location. An upper distance for the pseudoabsences was specified as this
has been shown to prevent models from contrasting completely different
climate conditions, e.g. temperate vs. tropical (VanDerWal et al.,
2009). The prevalence of the pseudo-absences was always equal to the
number of presence points (i.e. 730)