Considerations on how to optimally include biotic information in
SDMs
Methodological considerations include whether the biotic interaction
should be included as the raw distribution or the predicted suitability
map and whether a classical modelling approach should be used or a Joint
species distribution modelling (JSDM) approach. When the choice is made
to introduce the predicted suitability maps of the interacting species’
instead of the raw distributions, results should be interpreted with
care, as the prediction map may be highly dependent on abiotic factors
(such as climatic variables), making it less likely to explain the
species occurrences which abiotic factors cannot explain (Silva et al.
2014). Another commonly used approach for modelling biotic interactions
is JSDM and it is suitable for situations where the biotic interactions
are not known a priori , and this method helps to understand a
species’ geographical range from a community ecology perspective
(Pollock et al. 2014; Ovaskainen et al. 2017). The risk is, however,
that any detected relationships between species may be due to shared
habitat preferences not accounted for elsewhere in the model instead of
biotic relationships (Wisz et al. 2013; Pollock et al. 2014; Ovaskainen
et al. 2017).
Here, we showed that biotic factors can improve the SDMs of wild bees in
the Netherlands, especially when the distribution of the interacting
species is narrow. Resolution, taxonomic level, the degree of
specificity in the interaction, e.g. specialist species vs. generalist
species, should be taken into account to achieve the most optimal
models. We recommend using single species or genus data as a biotic
variable in the models of specialist species and to use an approximate,
such as flower richness, for more generalist species.