Tropical tree communities: a test of the nonequilibrium hypothesis

Ecology, March, 1996 by John Terborgh, Robin B. Foster, Percy V. Nunez

INTRODUCTION

In 1979 Hubbell proposed a "nonequilibrium" model of tropical forest tree communities in which all species lack density dependence and are, by implication, adaptively equivalent. Over the ensuing 15 yr, Hubbell and several coworkers have searched diligently for evidence of density dependence in both a 13-ha dry forest plot in Costa Rica (Hubbell 1979), and a 50-ha moist forest plot in Panama (Hubbell and Foster 1990, Hubbell et al. 1990, Condit et el. 1992). These efforts have so far yielded only modest success. A few of the most abundant species in the 50-ha Panama plot showed reduced growth and survivorship as juveniles when they were the nearest neighbors of conspecific adults. A large majority of the species tested failed to show evidence of density dependence, and, consequently, Hubbell and his coworkers have continued to reiterate their support of the nonequilibrium model (above references).

The avenue of density dependence is not the only approach to testing the nonequilibrium model, however. Hubbell and Foster (1986a:322) described another:

The critical distinction between equilibrium and nonequilibrium communities, to our way of thinking, is whether a particular species assemblage is stabilized by intrinsic biotic interactions under the given environmental regime of the site, and whether, if the species mixture is altered, the community tends to return to its former composition.

In short, a nonequilibrium community would not be expected to return to its former composition after disturbance.

Here we show that a highly diverse tropical forest in the floodplain of the Manu River in southeastern Peru consistently returns to its former composition after major disturbance, thereby contradicting Hubbell's second criterion for a nonequilibrium community.

On testing the nonequilibrium model

The nonequilibrium model specifies that the composition of a tropical tree community is subject to random errors of sampling so that it "drifts" through time at a rate inversely proportional to the size of the population under consideration. An unfortunate consequence of emphasizing the time dimension is that a test of the model is empirically out of reach, because one does not have the leisure to observe a plot over the span of several tree generations.

However, it should be possible in principle to conduct a test of the model in space rather than in time. Independent but environmentally equivalent sites are, by the nonequilibrium hypothesis, expected to show no correlation in species abundances. The challenge here is that of selecting independent sites, because the distance over which communities are expected to exhibit spatial correlation in composition cannot be deduced from the structure of the model, and Hubbell does not comment on the matter. The problem of spatial auto-correlation in composition would be avoided, however, if one selected environmentally similar sites that were isolated from one another by forests of distinct species composition.

MATERIALS AND METHODS

Environmental setting

We conducted the research within the Manu River basin in southeastern Peru, based at the Cocha Cashu Biological Station (11 [degrees] 52 [minutes] S, 71 [degrees] 21 [minutes] W). The Manu River meander belt offers an ideal setting in which to study environmentally equivalent, but independently constituted, stands of forest. The river meanders actively, cutting into banks on the outside of bends and depositing sediment on point bars at the tips of elongating meander loops (Kalliola et al. 1991). Between the Cocha Cashu Biological Station and the mouth of the Manu River, a linear distance of about 90 km, [approximately equal to]100 ha of forest are eroded away by the river each year (Kalliola et al. 1987).

Point bars advance on elongating meander loops in a series of incremental levee-backwater units (Foster et al. 1986, Kalliola et al. 1987). The levees are promptly colonized by pioneer vegetation, initiating a primary succession, the first stage of which consists of essentially monotypic stands of a composite tree, Tessaria integrifolia (Kalliola et al. 1988). The succession subsequently progresses through a regular sequence of compositional stages until, eventually, the forest attains a characteristic vertical organization comprised of several strata of superimposed crowns (Terborgh and Petren 1991). From rough estimates of the duration of the pioneer and intermediate stages, Terborgh and Petren (1991) estimated that at least 300 yr are required for this succession to reach the structurally "mature" stage.

Every meander loop recreates the primary succession. We do not assume, however, that spatial correlation in species composition is absent in nearby loops, because mature forest at the base of loops is frequently continuous with the mature forest of adjacent loops. However, continuity of the floodplain forest is interrupted, often by a kilometre or more, wherever the river cuts into the uplands that bound the floodplain. Upland forests are distinct from floodplain forests with respect to tree species composition (see Results: Floodplain vs. upland forests, below), soil type (ultisols vs. entisols), and chemical properties (acidic vs. neutral soils; M. Riley and J. Terborgh, unpublished data). Here we assume that meander loops isolated by one or more intervening stretches of upland forest are independent from the standpoint of spatial correlation of species composition.