Abstract
Systematic analyses that examine uncertainty in models are essential for assessing their credibility. In this study, we implemented an uncertainty analysis that quantifies the effect of parameter uncertainty on a set of ecological indicators in outputs of the marine ecosystem OSMOSE model applied to the northern Peru Current ecosystem (NPCE OSMOSE). We worked under simple uncertainty assumptions corresponding to ranges of 10, 20, and 30% variability around the reference values of the parameters describing the dynamics of the species modelled in NPCE OSMOSE. The results based on nearly 1.5 million simulations help to identify the main sources of uncertainty that could be of use to focus future research and point to the most reliable indicators in the face of uncertainty. First, uncertainty in the parameters of some species, in particular a key zooplankton species and Humboldt squid, have far-reaching impacts on the modelled biomass of other key species. Second, a set of ecological indicators appear to be relatively insensitive to input uncertainty and may therefore be useful in supporting ecosystem-based management. Furthermore, our findings underline the need for better species representation in terms of data quality but also bottom-up and top-down processes in trophic models. We highlight the difficulties of studying uncertainty in complex models while presenting an approach that can serve as a template for addressing uncertainty analysis in other ecosystem models. Finally, although this approach focuses on parameter uncertainty, it could also serve as a guide to address structural, initial conditions and model forcing uncertainties.