Chapter 18

EcosystemDecayinClosed

ForestFragments

John Terborgh and Kenneth Feeley

OVERVIEW

We summarize a long-term study of forest fragments isolated since 1986 as islands in Lago Guri, a vast hydroelectric
impoundment in Venezuela. We studied replicate islands of two classes: small (≥0.5 ha,<2 ha) and medium (≥4 ha,
<15 ha). Islands of both classes lacked predators of vertebrates and were deficient in pollinators and seed dispersers,
but supported hyperdense populations of predators of invertebrates (birds, lizards, anurans, spiders), rodents, and
especially herbivores (howler monkeys, common iguanas, and leaf-cutter ants). Medium islands differed from small
islands in supporting a predator of leaf-cutter ants (armadillo), a scatterhoarder (agouti), and often a mesopredator
(capuchin monkey), while supporting herbivores at densities intermediate between those of small islands and the large
landmasses that served as controls. Large landmasses supported intact faunas and served as controls.
Our results strongly supported the hypothesis of Hairston, Smith, and Slobodkin that consumers will increase in the
absence of predators to levels that result in damage to vegetation. Tree and sapling mortality was higher, and sapling
recruitment dramatically lower, on small islands supporting hyperabundant herbivores. However, trophic cascades on
predator-free islets were more complex than predicted by contemporary food-web theory because consumers belonged
to various functional groups (e.g., pollinators, seed predators, folivores). Moreover, hyperabundant herbivores gener-
ated indirect bottom-u peffects mediated via nutrient cycling. Indirect effects included enhanced tree growth, density
overcompensation in bird communities, and decreased rates of avian extinction on islands supporting hyperabundant
howler monkeys.
On small islands, sapling mortality exceeded recruitment for all species. This suggests that herbivore defenses of
common plant species are most effective in the presence of top-down regulation. Hyperabundant herbivores appeared
to be under bottom-u pregulation. For exam ple, tortoises on islands exhibited reduced growth rates and howler mon-
keys were below normal weight and reproduced slowly with respect to controls. We did not find any direct impact of
edge exposure on tree demography.
Overall, we conclude that terrestrial trophic cascades are far more complex than implied by simple, three-level
trophic models. The multiplicity of pathways suggests a complex interaction web. This web can be distorted in myriad
ways with consequences that most would regard as undesirable.

INTRODUCTION

An initial effect of habitat fragmentation is
to distort or disrupt many landscape-scale
processes, including both biological processes,
such as predation, pollination, and seed dis-
persal, and physical processes, such as fire
and the moderating effects of a continuous

habitat on microclimate (Kapos 1989, Aizen and

Feinsinger 1994, Asquithet al. 1999, Cochrane

and Laurance 2002, Laurance et al. 2002,

Chapman and Chapman 2003). The multiplic-

ity of processes that are altered by fragmentation

has impeded efforts to understand the conse-

quences of fragmentation from a mechanistic

standpoint.