Résumé: Habitat modelling is increasingly relevant in biodiversity and conservation studies. A typical application is to predict potential zones of specific conservation interest. With many environmental covariates, a large number of models can he investigated but multi-model inference may become impractical. Shrinkage regression overcomes this issue by dealing with the identification and accurate estimation of effect size for prediction. In a Bayesian framework we investigated the use of a shrinkage prior, the Horseshoe, for variable selection in spatial generalized linear models (GLM). As study cases, we considered 5 datasets on small pelagic fish abundance in the Gulf of Lion (Mediterranean Sea, France) and 9 environmental inputs. We compared the predictive performances of a simple kriging model, a full spatial GLM model with independent normal priors for regression coefficients, a full spatial GLM model with a Horseshoe prior for regression coefficients and 2 zero-inflated models (spatial and non-spatial) with a Horseshoe prior. Predictive performances were evaluated by cross validation on a hold-out subset of the data: models with a Horseshoe prior performed best, and the full model with independent normal priors worst. With an increasing number of inputs, extrapolation quickly became pervasive as we tried to predict from novel combinations of covariate values. By shrinking regression coefficients with a Horseshoe prior, only one model needed to be fitted to the data in order to obtain reasonable and accurate predictions, including extrapolations.

Résumé: Aim: To assess the degree of overlap between the environmental niches of marine planktonic copepods and test if the distribution of copepod functional groups differs across environmental gradients. Location: The Mediterranean Sea. Methods: Functional groups were defined based on clustering of functional traits in 106 marine copepod species using a multivariate ordination analysis. Functional traits included maximum body length, feeding mode, spawning strategy and trophic group. Simultaneously, the global distribution of the species was used to model their environmental niches with six environmental variables. For each of these predictors, four niche parameters were derived from the univariate response curve of each species to summarise their environmental preferences and ordinate the species in niche space through a PCA. Finally, the differences in the position in niche space of functional groups were tested with variance analysis. Results: We identified seven copepod functional groups with different distributions along the environmental gradients covered by our study. While carnivorous functional groups were affiliated with oligotrophic and tropical conditions, large and small current-feeding herbivores are associated with colder, more seasonally varying and productive conditions. Small cruising detritivores and other small current-feeding herbivores were not affiliated with specific conditions as their constituting species were scattered in niche space. Main conclusions: Since copepod functional groups occupy distinct ecological niches, ecosystem processes related to these groups are expected to vary across environmental gradients. Conditions favouring large current-feeding herbivores should allow for enhanced fluxes of energy and nutrients through Mediterranean Sea ecosystems, while such fluxes should be weakened where large carnivores and small passive ambush-feeding copepods dominate. Our study supports the development of trait-based zooplankton functional groups in marine ecosystem models.