Predicting community outcomes from pairwise interactions: integrating density- and trait-mediated effects

Understanding how species interactions shape the structure of ecological communities based on pairwise comparisons has been a difficult undertaking for ecologists because effects in reassembled communities can be different than simple density-mediated interactions would suggest. Part of this complexity occurs because many species change their behavior and morphology with different predators and competitors and, thus, change their per-capita interaction rates (i.e. trait-mediated interactions). Our objective was to use a simple experimental community of two predators (larval dragonflies, Anax longipes, and larval salamanders, Ambystoma tigrinum), two prey (larval green frogs, Rana clamitans, and larval bullfrogs, R. catesbeiana), and a shared prey resource to determine whether we can predict interactions in a reassembled community by combining our knowledge of density- and trait-mediated interactions,. We combined pairwise laboratory experiments on predation rates and predator-induced behaviors with a mesocosm experiment to examine density- and trait-mediated effects. We used a factorial combination of no predators, caged Anax (to induce anti-predator traits without changing prey density), and lethal Anax crossed with no predators, caged Ambystoma, and lethal Ambystoma. The species interactions in the reassembled community were qualitatively predictable based on the pairwise experiments. Lethal Anax preyed upon Ambystoma and green frogs while lethal Ambystoma only preyed upon green frogs. Anax also reduced the activity of the green frogs; this caused a decrease in salamander predation on green frogs, a decrease in green frog acquisition of resources, and an increase in bullfrog acquisition of resources. Ambystoma had no effect on green frog activity, no effect on resource acquisition by green frogs, and no effect on resource acquisition by bullfrogs. These results suggest that we can better understand how species interact in natural communities if we have a more detailed understanding of trait-mediated mechanisms. However, if predicting the structure of large communities requires identifying how each species alters its traits in the presence of all other species along with altering density, improving our predictive ability may be a prohibitively large undertaking.