Abstract
Integrative approaches that investigate trophic ecology drivers provide knowledge to explore and predict changes in food-web dynamics under contrasting scenarios of global change. However, there are few studies that analyse the relationship between environmental factors and trophic interactions and that additionally consider other human stressors such as fisheries. Here, we use Bayesian Stable isotope mixing models to study the trophic ecology of a widespread pelagic predatory fish, the swordfish (Xiphias gladius), in the western Mediterranean Sea and the adjacent Atlantic waters. We explore the relationships between dietary estimates and biological, environmental and anthropogenic drivers using generalized additive models (GAMs). GAMs are used to develop spatial predictions of present prey consumption and, as a prospective exercise, to project changes in prey consumption under different future climate change scenarios. Overall, we found that swordfish diet varied as a response to changing environmental conditions, particularly to varying sea surface temperature (SST), mixed layer depth (MLD) and chlorophyll-a concentration (Chl); and to fishing pressure. Fish consumption was related to SST and MLD. Squid consumption was related to SST, with the greatest contributions observed in swordfish of intermediate lengths. Squid had a higher contribution to swordfish diet around the Canary Islands and the western Mediterranean Sea, while gelatinous organisms were more consumed around the Gulf of Cádiz. The consumption of gelatinous organisms was higher in smaller swordfish and in areas with lower productivity. Our prospective exercise suggested different diet alterations under contrasting future global change scenarios. For the first time, we provide quantitative evidence of how large-scale, spatial-temporal patterns in fishing pressure and environmental conditions can shape the diet of swordfish. Our study presents useful results to assess the diet of this predator and highlight how incorporating trophic interactions into projections can improve our understanding of future distributions.