Integrating univariate niche dynamics in species distribution models: a step forward for marine research on biological invasions

Aim The development of approaches to predict the distribution and
potential expansion of invasive species is still an open challenge. Here
our goal is to improve the modelling procedure for marine invaders by
coupling Species Distribution Models (SDMs) with an analysis of their
univariate niche dynamics. In particular, we tested for the first time
whether choosing model predictors among the stable niche dimensions was
effective in improving predictions of invasive species expansion. Location
Mediterranean Sea Taxon Dusky spinefoot, Siganus luridus. Methods We
analysed the univariate niche dynamics for S. luridus across its native
and invaded ranges, by applying a standardized framework that allowed the
identification of cases of niche stability or shift. We compared
inter-range transferability of SDMs fitted with different combinations of
labile or stable predictors. Finally, we evaluated interactions in SDM
settings (calibration area, model technique and predictors set) on models’
predictive ability, using independent data from the most recent phase of
invasion. Results We detected a pattern of niche stability for several
variables, especially salinity and bathymetry, which positively influenced
model inter-ranges transferability: when the models calibrated in the
native range include only stable niche axes, predictive ability is
improved. We also identified a shift toward lower surface temperatures in
the introduced range, which were almost never experienced by the species
before invasion. The model calibrated within the combined ranges was the
most ecologically congruent. Also, models calibrated in the invaded range
allowed a correct prediction of range expansion, with the predicted
suitable areas only slightly underestimated. Main conclusions We provide
the first evidence that using conserved predictors in SDMs improves
inter-range projections of expanding invasive species. Variable selection,
calibration area and modelling technique all matter when modelling
invasive species, with important interaction effects. We provide
guidelines on how to improve SDMs applications in biological invasion
research.