A key assertion of the CEG model is that a paleocommunity’s trophic network can never be specified by a single topology (Roopnarine, 2006; Roopnarine, 2009). There is uncertainty associated with the biotic interactions of a fossil species because no one was there to observe them. Preserved evidence of interactions such as bite marks, gut contents or leaf damage record a subset of the possible range of interactions. Moreover, the topology specified for a single community is expected to vary spatially and temporally. The strength and direction of interspecific interactions of extant species are known to vary according to physical conditions, the presence or absence of other species in the community, relative population sizes, and the incumbency of species when addition to the community is asynchronous. These uncertainties must be incorporated into any realistically complex model of a community food web. In the CEG model, species are therefore grouped into trophic guilds based on the most accurate trophic interpretations available, into trophic guilds. Trophic guilds are defined as the trophic habits and habitats of member species, for example, the “very small carnivorous amniotes” of a Late Permian terrestrial community.
The resulting guild structure represents a species aggregation scheme. The most common aggregation scheme is to assemble species into groups called “trophic species”. Trophic species group species that are assumed to have the same prey and predators. The motivation for this grouping is unclear in cases where link data are available at the species level. One advantage, however, may be to avoid biases introduced by an undersampling of poorly resolved links. Patterns of connection among trophic species may also illuminate patterns of energy and nutrient flow among major species ecotypes in the community. There is no guarantee, though, and in fact no expectation for the preservation of network topology in the conversion of species-level data to a trophic species network. It is always preferable to use species-level data to represent true community complexity. Furthermore, aggregation into trophic species is an inference the strength of which cannot be justified for fossil taxa, and the scheme should be avoided in paleo-food webs. Given that species-level data are rarely available for fossil species, however, and are basically never complete, aggregation is necessary. Dunne et al. (2008) therefore converted species-level data to trophic species in their study of Cambrian food webs. The CEG model aggregates species into trophic guilds, groups of species that cannot be distinguished trophically on the basis of available data. An example would be “epifaunal, seagrass-dwelling suspension feeding bivalves”. Those species, in a particular community, potentially share the same predators and prey. Trophic guilds are similar but not equivalent to trophic species, yet it is clear that if a trophic species is an accurate representation of the species which it comprises, then the composition of a similar trophic guild will approach the composition of the trophic species as the species data become more precise. A network of trophic guilds is termed a metanetwork, and is an hierarchically higher level representation of a species-level network. Two trophic guilds linked in a metanetwork contain species that are potentially trophic interactors. A metanetwork therefore summarizes the most accurate and precise data available for a paleocommunity’s food web.
The contrast between the two aggregation schemes is reduced to one of accuracy and precision. The trophic species scheme assumes a high level of precision, thereby justifying an assumption of trophic neutrality among species within the trophic species. This level of precision is unlikely to be available for fossil taxa, and in any case can never be tested. The metanetwork and trophic guild scheme assumes that the understanding of a species trophic habit is accurate, even though its precise interspecific interactions may be unknown or known incompletely. These uncertainties, stemming from incomplete data and temporal-spatial variance of the data, will be addressed in a future post by exploring the range of species-level food webs implied by the metanetwork.