Node centrality indices in food webs: Rank orders versus distributions

Network analysis examines the role of species in ecological communities. The most common approach involves measurement of centrality of species or other groups of individuals based on their topological positions in food webs, followed by establishing the rank order of importance of these groups. However, ranking may differ considerably with indices of centrality and therefore comparison of rank orders is essential to obtain more meaningful results on species performance. Since ranking ignores absolute differences between centrality values, species orders may neglect important structural information in food webs. Consequently, simultaneous examination of the distribution of index values is inevitable. Hierarchical clustering and consensus generation revealed that rank orders of centrality exhibit a similar pattern over six example food webs, while distributions differ not only with indices because their relationships are largely inconsistent with food webs as well. Therefore, optimal analysis of networks and the selection of keystone species in any ecological study should rely upon both of these procedures. Similar conclusions are drawn from the detailed evaluation of a sample food web from the Florida Bay.