Importance of trophic information, simplification and aggregation error in ecosystem models

Ecosystem models are becoming increasingly important as pressure from fisheries intensifies and ecosystem-based fisheries management becomes more widely used. Trophic webs often form the basis of ecosystem models and ecosystem-specific dietary information is crucial for optimal model performance. This is particularly the case if model predictions are used in management decisions. The Tasmanian live fish fishery for banded morwong was used as a case study to investigate the importance of trophic information, model simplification and aggregation error on ecosystem model results. Dietary analysis of 6 commonly captured reef fish was undertaken. Significant trophic overlap was found between blue throat wrasse Notolabrus tetricus and purple wrasse N. fucicola, and banded morwong Cheilodactylus spectabilis and bastard trumpeter Latridopsis forsteri. Marblefish Aplodactylus arctidens and long-snouted boarfish Pentaceropsis recurvirostris had significantly different diets from other species studied. Using this information, a detailed qualitative model was produced and then simplified through the aggregation of variables. Variables were aggregated using 3 methods: Euclidean distance, Bray-Curtis similarity, and regular equivalence for inclusion in 3 simplified models. Variable aggregation is undertaken in many studies and may create aggregation error. Each aggregation method produced a different proportion of incorrect model predictions as a result of aggregation error. The model simplified using regular equivalence produced the least aggregation error and a web structure aligned with the dietary analysis. More widespread use of these methods in fisheries management should be considered.