Assistant Professor McGill University, Quebec, Canada
Background/Question/Methods
Describing the spatial distribution of species is vital for successful species, habitat, or biodiversity conservation policies and recovery plans. In this context, species distribution models (SDMs) have become a commonly used numerical tool to estimate species ranges by joining occurrence data with geographic predictors. However, as open-source geographic species data (e.g., GBIF) typically contain substantial systematic spatial and taxonomical biases, the validity of unadjusted SDM projections remains questionable. The novel S2BaK model combines opportunistic species sightings-only data with systematic species surveys to generate computationally efficient “bias-adjustment kernels”, resulting in improved predictive power in comparison to traditional models. As S2BaK can adjust the output of any SDM, it retains the flexibility to use advances in the SDM literature and to be scalable to large systems. While this approach was previously successful in a small country with low data availability (Panama), this project tested and scaled up the uses of S2BaK to a continental context by building integrative models for the data-rich and environmentally heterogeneous European Flora. For 11 826 plant species, we modelled distributions continentally using S2BaK and unadjusted SDM approaches, we validated the models, and we compared resulting range distributions and overall diversity estimates.
Results/Conclusions
We found high levels of spatial biases (R2 = 0.568), which drove a systemic overestimation of estimated species richness in well-sampled countries and regions in unadjusted SDMs, but, interestingly, smaller species-level bias levels in sightings-only records (R2 = 0.146). We show that the S2BaK model successfully removed these biases to allow accurate inferences of individual species distributions on average, including for species not observed in any systematic survey, and downweighed the overestimation of species richness occurring in western European countries.
We provide an optimized continental plant α-diversity information layer, representing the most complete, speciose, and fine-scaled vegetation biodiversity baseline of Europe to date, which will be a key tool in answering ecological questions on plant distributions, their drivers, and their susceptibility to anthropogenic stressors.
