2032Ecology,79(6), 1998, pp. 2032–2040q1998 by the Ecological Society of AmericaRAIN FOREST FRAGMENTATION AND THE DYNAMICS OFAMAZONIAN TREE COMMUNITIESWILLIAMF. LAURANCE,LEANDROV. FERREIRA,JUDYM. RANKIN-DEMERONA,1ANDSUSANG. LAURANCEBiological Dynamics of Forest Fragments Project, National Institute for Research in the Amazon (INPA),C.P. 478, Manaus, AM 69011-970, BrazilAbstract.Few studies have assessed effects of habitat fragmentation on tropical forestdynamics. We describe results from an 18-yr experimental study of the effects of rain forestfragmentation on tree-community dynamics in central Amazonia. Tree communities wereassessed in 39 permanent, 1-ha plots in forest fragments of 1, 10, or 100 ha in area, andin 27 plots in nearby continuous forest. Repeated censuses of.56 000 marked trees ($10cm diameter at breast height) were used to generate annualized estimates of tree mortality,damage, and turnover in fragmented and continuous forest.On average, forest fragments exhibited markedly elevated dynamics, apparently as aresult of increased windthrow and microclimatic changes near forest edges. Mean mortality,damage, and turnover rates were much higher within 60 m of edges (4.01, 4.10, and 3.16%,respectively) and moderately higher within 60–100 m of edges (2.40, 1.96, and 2.05%)than in forest interiors (1.27, 1.48, and 1.15%). Less-pronounced changes in mortality andturnover rates were apparently detectable up to;300 m from forest edges. Edge aspecthad no significant effect on forest dynamics. Tree mortality and damage rates did not varysignificantly with fragment age, suggesting that increased dynamics are not merely transitoryeffects that occur immediately after fragmentation, while turnover rates increased with agein most (8/9) fragments.These findings reveal that fragmentation causes important changes in the dynamics ofAmazonian forests, especially within;100 m of habitat edges. A mathematical ‘‘core-areamodel’’ incorporating these data predicted that edge effects will increase rapidly in im-portance once fragments fall below;100–400 ha in area, depending on fragment shape.Accelerated dynamics in fragments will alter forest structure, floristic composition, biomass,and microclimate and are likely to exacerbate effects of fragmentation on disturbance-sensitive species.Key words: Amazon rain forest; deforestation; edge effects; forest dynamics; habitat fragmen-tation; tree mortality and turnover; wind disturbance.INTRODUCTIONThe rate of tropical deforestation exceeds 153106ha annually, resulting in extensive fragmentation offorest landscapes (Whitmore 1997). Although researchon fragmented tropical forests is increasing (Schelhasand Greenberg 1996, Laurance and Bierregaard 1997),few studies have assessed effects of fragmentation onthe dynamics of tree communities. Trees largely de-termine the architecture and microclimatic conditionsof the forest, and hence changes in tree-communitydynamics may strongly affect other forest species (Bro-kaw 1985, Levey 1988, Clark 1990) and ecologicalprocesses (Powell and Powell 1987, Klein 1989, Did-ham et al. 1996).In recently fragmented forests, edge effects appearto be a major cause of change in plant communities.In rain forests the harsh external climate is buffered byManuscript received 25 February 1997; revised 2 Septem-ber 1997; accepted 17 September 1997.1Present address: Station de Recherches Forestieres,I.N.R.A.-Groupe Regional de Guyane, B.P. 709, 97387 Kour-ou Cedex, France.dense canopy cover, but this buffering breaks downnear forest edges (Kapos 1989, Williams-Linera 1990)and may lead to higher mortality of desiccation-sen-sitive plants (Lovejoy et al. 1986, Sizer 1992, Kaposet al. 1997). Strong turbulence can result when windsstrike abrupt forest edges, increasing rates of wind-throw and forest structural damage (Laurance 1991,1997, Ferreira and Laurance 1997). Fragmented rainforests often exhibit a proliferation of vines, lianas, andsecondary vegetation near edges (Laurance 1991, Tur-ner and Tan 1992, Malcolm 1994, Viana et al. 1997),and some seasonal forests appear highly prone to in-vasions of exotic plants (Fensham et al. 1994, Laurance1997).In central Amazonia, a long-term study of rain forestfragments experimentally isolated in the early 1980s(Lovejoy et al. 1986, Bierregaard et al. 1992) has begunto yield insights into the dynamics of fragmented treecommunities. Working with a single tree family (Myr-taceae), Ferreira and Laurance (1997) found sharplyelevated rates of tree mortality and damage within;100 m of forest-pasture edges. Laurance et al.September 1998 2033FRAGMENTATION AND FOREST DYNAMICSFIG. 1. Study area in central Amazonia, showing locations of forest fragments and controls (shaded blocks) used in thestudy. Stippled areas are cattle pastures or regrowth forest, while unstippled areas are rain forest. Thick, solid lines are roads.(1998a) detected a marked increase in tree recruitmentrates within;100 m of edges, and found that regen-erating trees in study plots with high recruitment werebiased toward families with many pioneer and second-ary species, and against families of old-growth, forest-interior species.Here we present a large-scale analysis of tree-com-munity dynamics in recently fragmented rain forests,using data collected over an 18-yr period. We focus onthree vital rates, tree mortality, tree damage, and treeturnover, which we compare in fragments of varioussizes to adjacent continuous forest.METHODSStudy areaThe study area is located;80 km N of Manaus,Brazil in central Amazonia (28309S, 608W), at 100–150 m elevation. Local soils are nutrient poor. Rainfallranges from 1900–2500 mm annually with a dry seasonfrom June to October. The forest canopy is 30–37 mtall, with emergents to 55 m. Species richness of treesis very high and can exceed 250 species ($10 cm dbh)per hectare (Rankin-de Merona et al. 1992).The study area is surrounded by large expanses(.200 km) of continuous forest to the west, north, andeast. In the early 1980s, a series of 1-, 10-, and 100-ha fragments (Fig. 1) were isolated by distances of 70–1000 m from surrounding forest by clearing the inter-vening vegetation to establish cattle pastures. Frag-ments were fenced to prevent encroachment by cattle.Reserves ranging from 1–1000 ha in area were delin-eated in nearby continuous forest to serve as experi-mental controls. Regrowth forests have regenerated insome cleared areas and are dominated