Global sensitivity and uncertainty analyses of an ecosystem model for simulating food web dynamics in the Cooperation Sea, Southern Ocean

There is growing interest in developing and using ecosystem simulation models to advise fisheries management in the Southern Ocean. However, poor understanding of the impacts of uncertainty in ecosystem model parameters slows down progress towards operational ecosystem models. To address this issue, we explored uncertainty in the parameters estimated during the calibration of an OSMOSE ecosystem model for the Cooperation Sea (“OSMOSE-CooperationSea”) and the impacts of this uncertainty. Our investigations pertained to four types of calibrated parameters: (1) Plank.access, the proportion of the biomass of background species groups available to focal species groups; (2) Bio_flux, the parameter controlling the flux of migratory species group biomasses across the modelled domain boundaries; (3) M_larval, the instantaneous larval mortality of the focal species groups; and (4) M_natural, the additional natural mortality of the focal species groups. Results with the Morris method suggested that the community in the Cooperation Sea was most sensitive to changes in the M_larval parameter of mesopelagic fishes. The biomasses of large-size, long-lived species such as toothfishes, Adélie penguin (Pygoscelis adeliae), seals, and whales were most sensitive to the parameters specific to these species groups. By contrast, the biomasses of small-sized, short-lived species such as mesopelagic fishes and krill species were most sensitive to changes in the parameters specific to the predators of these species groups. Monte Carlo simulations indicated that community dynamics were more sensitive to the M_larval and M_natural parameters than to the Plank.access and Bio_flux parameters. After gradually increasing the M_larval or M_natural parameter, the biomasses of Adélie penguin, seals and whales decreased, while the biomasses of mesopelagic fishes and Antarctic krill increased. By providing a comprehensive analysis of uncertainty in the parameters estimated during the calibration process, the present study represents an important step towards an operational ecosystem model for supporting ecosystem-based management in the Cooperation Sea. The present study will serve as a valuable basis for similar ecosystem modelling efforts in the Southern Ocean.