Niche construction
To construct the bioclimatic niche hypervolumes for each species we
first extracted observations from the Global Biodiversity Information
Facility (GBIF; http://www.gbif.org/ipt/) over a 30 year period
(1970-2000) using the ‘rgbif’ R package (Chamberlain et al. 2020)
using the R statistical software (R Core Development Team 2019).
We next utilised 19 ‘bioclimatic’ variables provided by WorldClim 2.1
(Fick & Hijmans 2017) at a resolution of 2.5 arc minutes and clipped
the data to the approximate region of the species observations (i.e.
western Europe; Figure 1). WorldClim’s bioclimatic variables are derived
from temperature and precipitation records and represent both mean and
extreme conditions at a location. We fitted a PCA to the bioclimatic
variables. The first six axes were retained as they represented
>95% of the variance across the 19 bioclimatic variables.
For each species observations, we collected the first six coordinates
projected onto this PCA space and used this to fit a 6-D hypervolume
using multidimensional kernel density estimation (Blonder et al.2014) using the hypervolume package (Blonder et al. 2014, 2018).
Twelve species were removed from all further analyses at this point as
there was an insufficient number of observations to fit the hypervolume
algorithm; thus we retained data for 145 species in total. The removed
species, being very rare, have little impact on community properties. We
also removed data on populations where species occurred at sites only
very infrequently (zero counts for >25% of years) as these
provide unreliable estimates of population stability and may bias
community niche metrics (below).