Impacts of major predators on tropical agroforest arthropods: comparisons within and across taxa.

In food web studies, taxonomically unrelated predators are often grouped into trophic levels regardless of their relative importance on prey assemblages, multiple predator effects, or interactions such as omnivory. Ants and birds are important predators likely to differentially shape arthropod assemblages, but no studies have compared their effects on a shared prey base. In two separate studies, we excluded birds and ants from branches of a canopy tree ( Inga micheliana) in a coffee farm in Mexico for 2 months in the dry and wet seasons of 2002. We investigated changes in arthropod densities with and without predation pressure from (1) birds and (2) ant assemblages dominated by one of two ant species ( Azteca instabilis and Camponotus senex). We first analyzed individual effects of each predator (birds, Azteca instabilis, and C. senex) then used a per day effect metric to compare differences in effects across (birds vs ants) and within predator taxa (the two ant species). Individually, birds reduced densities of total and large arthropods and some arthropod orders (e.g., spiders, beetles, roaches) in both seasons. Azteca instabilis did not significantly affect arthropods (total, small, large or specific orders). Camponotus senex, however, tended to remove arthropods (total, small), especially in the dry season, and affected arthropod densities of some orders both positively and negatively. Predators greatly differed in their effects on Inga arthropods (for all, small, large, and individual orders of arthropods) both in sign (+/-) and magnitudes of effects. Birds had stronger negative effects on arthropods than ants and the two dominant ant species had stronger effects on arthropods in different seasons. Our results show that aggregating taxonomically related and unrelated predators into trophic levels without prior experimental data quantifying the sign and strengths of effects may lead to a misrepresentation of food web interactions.