Ecosystem models of Lake Victoria (East Africa): exploring the sensitivity of ecosystem effects of fishing to model choice

Abstract

Ecosystem simulation models are valuable tools for strengthening and promoting ecosystem-based fisheries management (EBFM). However, utility of these models in practical fisheries management is often undermined by lack of simple means to test the effect of uncertainty on model outputs. Recently, the use of multiple ecosystem models has been recommended as an insurance against effects of uncertainty that comes with modelling complex systems. The assumption is that if models with different structure and formulation give consistent results, then, policy prescriptions are robust (i.e. less sensitive to model choice). However, information on the behaviour of trends from structurally-distinct ecosystem models with respect to changes in fishing conditions is limited, especially for freshwater systems. In this study, we compared outputs of two ecosystem models, Ecopath with Ecosim (EwE) and Atlantis, for Lake Victoria under different fishing pressure scenarios. We compared model behaviour at the ecosystem level, and also at a level of functional groups. At functional group level, we determined two questions: what is the change in the targeted group, and what are the consequent effects in other parts of the system? Overall results suggest that different model formulations can provide similar qualitative predictions (direction of change), especially for targeted groups with similar trophic interactions and adequate data for parameterization and calibration. However, considerable variations in predictions (where models predict opposite trends) may also occur due to inconsistencies in the strength of the aggregate multispecies interactions between species and models, and not necessarily due to model detail and complexity. Therefore, with more information and data, especially on diet, and comparable representation of feeding interactions across models, ecosystem models with distinct structure and formulation can give consistent policy evaluations for most biological groups.