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: Understanding the relationship between environmental variables and pelagic species concentrations and dynamics is helpful to improve fishery management, especially in a changing environment. Drifting fish aggregating device (DFAD)-associated tuna and non-tuna biomass data from the fishers' echo-sounder buoys operating in the Atlantic Ocean have been modelled as functions of oceanographic (Sea Surface Temperature, Chlorophyll-a, Salinity, Sea Level Anomaly, Thermocline depth and gradient, Geostrophic current, Total Current, Depth) and DFAD variables (DFAD speed, bearing and soak time) using Generalized Additive Mixed Models (GAMMs). Biological interaction (presence of non-tuna species at DFADs) was also included in the tuna model, and found to be significant at this time scale. All variables were included in the analyses but only some of them were highly significant, and variable significance differed among fish groups. In general, most of the fish biomass distribution was explained by the ocean productivity and DFAD-variables. Indeed, this study revealed different environmental preferences for tunas and non-tuna species and suggested the existence of active habitat selection. This improved assessment of environmental and DFAD effects on tuna and non-tuna catchability in the purse seine tuna fishery will contribute to transfer of better scientific advice to regional tuna commissions for the management and conservation of exploited resources.

Résumé: Knowing the feeding ecology of marine predators is pivotal to developing an understanding of their ecological role in the ecosystem and determining the trophic relationships between them. Despite the ecological importance of predatory pelagic fish species, research on these species in the Mediterranean Sea is limited. Here, by combining analyses of stomach contents and stable isotope values, we examined the feeding strategies of swordfish, Xiphias gladius, little tunny, Euthynnus alletteratus and Atlantic bonito, Sarda sarda, in the western Mediterranean Sea. We also compared the trophic niche and trophic level of these species with published information of other sympatric pelagic predators present in the ecosystem. Results indicated that, although the diet of the three species was composed mainly by fin-fish species, a clear segregation in their main feeding strategies was found. Swordfish showed a generalist diet including demersal species such as blue whiting, Micromesistius poutassou, and European hake, Merluccius merluccius, and pelagic fin-fish such as barracudina species (Arctozenus risso and Lestidiops jayakari) or small pelagic fish species. Little tunny and Atlantic bonito were segregated isotopically between them and showed a diet basically composed of anchovy, Engraulis encrasicolus, and round sardinella, Sardinella aurita, and sardines, Sardina pilchardus, respectively. This trophic segregation, in addition to potential segregation by depth, is likely a mechanism that allows their potential coexistence within the same pelagic habitat. When the trophic position of these three predatory pelagic fish species is compared with other pelagic predators such as bluefin tuna, Thunnus thynnus, and dolphinfish, Coryphaena hippurus, present in the western Mediterranean Sea, we found that they show similar intermediate trophic position in the ecosystem. In conclusion, the combined stomach and isotopic results highlight, especially for little tunny and Atlantic bonito, the trophic importance of Clupeoid species in their diet. In addition, the importance of demersal resources for swordfish provides evidence for the pelagic-demersal coupling of the ecosystem and the need to manage marine resources in an integrated way. (C) 2016 Elsevier Ltd. All rights reserved.