Evaluating impacts of imprecise parameters on the performance of an ecosystem model OSMOSE-JZB

Ecosystem models are widely advocated for ecosystem-based fisheries management (EBFM). However, the use of ecosystem models is still limited in tactical fisheries management due to large uncertainties that undermine the confidence of stakeholders. Using a Monte Carlo simulation approach, this study quantified the impact of imprecise parameters on performance of a multi-species individual-based model “OSMOSE” (Object-oriented Simulator of Marine Ecosystems Exploitation) in evaluating the status of fish community, population, and individual. Uncertainties associated with the following five parameters were considered: fishing mortality (Mfishing), additional natural mortality (Mnatural), larval mortality (Mlarval), species-specific relative fecundity (RF), and the proportion of available low trophic level (LTL) biomass to high trophic level (HTL) groups (Plank.access). The imprecise value of Mlarval had the highest relative influence on model outputs. The imprecision of model outputs were sharply increased when the imprecisions of RF and Plank.access were high. The “compound sources” resulted in a greater imprecision of model predictions, confirming that the hierarchical model structure of OSMOSE facilitated the error diffusion. We also investigated the synergetic and antagonistic effects of aforementioned parameter uncertainty on the total biomass of the community (Biocom) and mean trophic level of the community (mTLcom). Both Biocom and mTLcom seemed not sensitive to the changes in Mnatural. The Mfishing and Mlarval had negative effects on mTLcom, and there was a positive relationship between two ecological indicators and last two parameters (RF and Plank.access). The Mlarval had a dominant impact on Biocom and mTLcom. Increasing Mfishing and Mlarval could result in a larger decline of Biocom and mTLcom compared to that caused by a single parameter. Similarly, there was a larger increase of Biocom and mTLcom after increasing RF and Plank.access. There was a small variation of Biocom and mTLcom when both Mfishing and RF changed, suggesting Mfishing could offset the effect of RF on Biocom and mTLcom. This study suggests that the imprecision of model parameters can greatly reduce utility of an ecosystem model. We recommend to develop research priority to reduce identified key sources of uncertainty and better understand error diffusion resulting from hierarchical model structures.