Food-web aggregation, methodological and functional issues

Trophic species in food webs are often aggregated into fewer groups, using theoretical and empirical approaches, either for modelling tractability or because of the lack of data resolution. Heterogeneities in the resolution of food webs used in the literature have led to question their use to establish general topological rules. Despite an increasing number of studies relating topology to ecosystem functioning, we still have no idea on how species’ aggregation affects our perception of network functionalities. Therefore, we re-examined the conclusions drawn from an experimental manipulation relating top-predator foraging behaviour and biomass to food-web topology (Lazzaro et al. 2009) by aggregating a 74-species network according to different criteria (taxonomy, trophic similarity, size, expertise). We found that initial significant effects and functional properties were preserved over a large portion of the aggregation gradient (2/3) despite strong variations in the topological descriptor values along the gradient. Aggregation tended to produce more type II errors (false positive) than type I errors, advocating that most effects in aggregated networks are not methodological artefacts. Aggregation by taxonomy, trophic similarity and expertise better preserved functional properties (down to 38, 30 and 17 nodes, respectively) than aggregation by size (down to 40 nodes). Our results suggest that it is possible to relate the structure of aggregated networks to ecosystem properties provided that the methodological approaches are standardized and the level of lumping does not a exceed a reasonable threshold.