EcoDiet model: Incorporating multiple data types into the estimation of diet matrices

ICES CM/J:176 Ecosystem models are key-tools in ecosystem assessments: they allow to better understand their functioning and to monitor their response to human-induced pressures. To be robust, such models need to be based on local and thus pertinent data, which is seldom the case. This notably applies to trophic models in which definition of prey-predators relationships usually rests too much on literature and expert knowledge. To facilitate integration of local surveys data while taking into account literature information in qualitative/quantitative assessment of food-webs, we developed a novel Bayesian integrated model called EcoDiet. This model allows for (i) considering a priori knowledge from the literature on both food-web structure and dietary proportions (ii) determining the trophic structure using preys occurrences from stomach content analysis (SCA) (iii) estimating diets of all the ecosystem’s consumers through a mixing model using stable isotopes analysis (SIA). As a test case, this model was applied to the Celtic Sea ecosystem using SCA and SIA from fish collected during the EVHOE survey. We showed how combining multiple sources of data within an integrated approach provided more accurate and realistic diet estimates than literature alone. Local data on feeding ecology are fundamental in building trophic models and EcoDiet proves to be a key step within this approach of field data integration. Sampling a large diversity of compartments and usually covering the whole ecosystem spatial extent, scientific surveys appear ideal for collecting such data. Regular sampling programs would enable periodic updates of trophic models used in ecosystem assessment.

We present here a novel Bayesian integrated model, EcoDiet, which integrates multiple sources of data in order to simultaneously estimate a probabilistic food-web topology and the diets of all the consumers in the ecosystem.