In search of operational trophospecies in a tropical aquatic food web
Additional Document Info
Aggregations of biological species on the basis of trophic similarity (trophospecies) are the basic units of study in food web and ecosystem research, yet little attention has been devoted to articulating objective protocols for defining such aggregations. This study formulates several possible definitions based on alternative measures of similarity and hierarchical clustering. Twenty-four alternative methods were applied to a food web consisting of 116 original trophic entities (OTEs) from a tropical floodplain, for which the relative magnitude of each trophic link was estimated based on dietary data measured as volumetric proportions. The resulting 24 trophic hierarchies were compared based on cophenetic correlation arid matrices of OTE pairwise similarity, and patterns were interpreted based on additional ecological analyses for this system. Similarity measures based on topological food web (presence/absence) data yielded slightly greater cophenetic correlations than did measures based on dietary proportions (flow webs), but, overall, OTE pairwise correlations were not greater for one method relative to the other. The difference between these two approaches is driven by the treatment of weak feeding links; at least for the system considered here, distinctions among dietary generalists were obscured when weak links were weighted lightly. Additively combining the predator and resource aspects of each OTE's trophic role performed better than combining them multiplicatively. In general, there was little correspondence between OTE overlap in resource use and the extent to which predators were shared. Two measures of cluster similarity, one designed by us specifically for food webs (maximum linkage) and the other a standard method (average similarity between clusters), yielded more consistent and ecologically interpretable patterns of aggregation than other measures of cluster similarity considered. In deciding whether two trophospecies should be assigned a trophic link, the maximum linkage convention (in which a link is included if any pair of OTEs, one in each trophospecies, are linked) produced more aggregation than the minimum linkage convention (in which a link is included only if every pair of OTEs, pair members in each of the two trophospecies, is linked). The choice of similarity level for defining trophospecies remained an unresolved issue based on our analysis of this dataset. Perhaps the greatest challenge is posed by sampling bias within empirical datasets, and we ultimately conclude that it is difficult to identify trophospecies in this dataset by strictly objective criteria.