Trophic niche-space imaging, using resource and consumer traits

Trophic (feeding) interaction strength between two species depends on the traits of both the consumer and the resource. But how many traits of consumer and resource have to be measured to predict interaction strengths, and which? A theoretical framework for systematically determining trophic traits from empirical data was recently proposed. Here we demonstrate this approach for a group of 14 consumer fish species (Labeobarbus spp., Cyprinidae) and 11 aquatic resource categories coexisting in Lake Tana in northern Ethiopia. We analysed consumer and resource traits with known roles in feeding ecology: a matrix of 19 phenotypic traits relating to all aspects of feeding, from detection to digestion, measured on more than 1,300 fish specimens was used as input for the consumers, while resource categories were characterized using 11 traits. To estimate empirical interaction strengths we used diet data of more than 4,700 fish. We systematically investigated structure and geometry of trophic niche space, obtaining an image of trophic niche space, which we called a Trophic Trait Model (TTM). Interaction strengths are predicted by the distances between species in this space. The TTM was constructed by first estimating the parameters of a generic model, based on observed diets, empirical trait variables, and a prescribed number of dimensions of trophic niche space. Secondly, the model was cross-validated to assess its predictive power and finally the model was optimised for its predictive power by an exhaustive search through all combinations of up to two resource traits and two consumer traits, retaining the combinations that scored best in cross- validation. The procedure was carried out separately for one to four dimensions. Our results show trophic niche to be multi-dimensional: the positions of consumers and resources in trophic niche space are characterized by at least two resource and two consumer traits. Results further suggest that trophic niche space has a pseudo-Euclidean geometry. Our analysis not only informs theory and modelling, but may also be helpful for predicting trophic interaction strengths for pairs of species from models fitted to other, similar species.