A terrestrial field experiment showing the impact of eliminating top predators on foliage damage

Certain ecological models predict that the impact of top predators on plants (producers) usually depends on the number of trophic levels in the food web: with three levels the impact is positive, whereas with four it is negative. These models assume that each of the nonbasal levels eats only the next level down. Indeed, freshwater pelagic systems show the predicted response when a fourth trophic level (piscivorous fish) is experimentally added to a three-level food web. In terrestrial systems, however, the top predator, if insectivorous, typically feeds on more than one trophic level. This 'closed-loop omnivory', which may also characterize certain aquatic systems, necessitates a different model. Anolis lizards are often such top predators on small islands in the Bahamas, eating both carnivorous and herbivorous arthropods. Numbers of web-spiders on islands without lizards are about 10 times those on islands with lizards and in experimental enclosures with lizards removed are 3 times those in enclosures with lizards at natural densities. Total leaf damage (by area) to buttonwood (Conocarpus erectus) on islands without lizards is 1.5 times that on islands with lizards. These results indicate that lizards reduce the overall damage from herbivorous arthropods, even though they also reduce the numbers of web spiders which eat some herbivorous arthropods. Here we report on an experiment with lizards, spiders, herbivorous arthropods and sea grape, that indicates that the impact of top predators on producers depends on the relative strengths of interactions among the upper tropic levels. The impact is predicted to be positive if herbivorous arthropods are more prone to be eaten by lizards than by spiders (Fig. 1, model A) and negative if vice versa (model B). Experimental removal of lizards both increased scar damage on sea grape leaves, produced by homopterans and other organisms, and decreased damage produced by gall midges, indicating that the food web is a composite of models A and B (Fig. 1, model C). Contrary to the experiments in freshwater pelagic systems with four trophic levels, the net effect of top predators on producers was positive, apparently because the model A pathway predominates.