Forest fragmentation effects on palm diversity in central Amazonia


*Present address and correspondence: CENARGEN/ EMBRAPA, C. P. 02372, 70.770–900 Brasília, DF, Brasil (fax + 55 61 340–3624; e-mail scariot@cenargen.


1 The effects of fragmentation on quantitative measures of floristic diversity in a palm community were examined in the Biological Dynamics of Forest Fragments study area in central Amazonia. Three 1-ha, three 10-ha, two 100-ha and three continuous forest reserves, distributed among three sites, were surveyed. In each reserve, 10 20 × 20 m plots were sampled, resulting in a total of 110 plots representing 4.4 sampled hectares.

2 The taxon composition of this palm community was dominated by stemmed, understorey palms. A total of 23 225 individuals from 36 taxa was recorded; five of the taxa were not sampled in continuous forest.

3 Taxa richness did not vary across reserve size or sites unless taxa not sampled in the continuous forest were removed from the analysis. Smaller forest fragments then harboured fewer taxa in the seedling stage than large forest fragments or continuous forest, despite the short time since isolation (10–15 years). There was a significant effect of location on the number of taxa per plot for all life stages, but only seedling and total were significantly affected by reserve size.

4 Reserve size did not affect the Shannon and Evenness indices. Reserves of similar sizes were floristically more similar than reserves of very different sizes.

5 Palms are important for the structure and composition of the forest. Their conservation may require the establishment of a number of large reserves.


The primary cause of the decline in diversity of rain forest taxa is habitat loss (Ehrlich 1988). Habitat destruction results in habitat fragmentation, thus causing further loss of original habitat, reduces the size of habitat fragments and increases their isolation (Andrén 1994). Habitat fragmentation may influence local and regional patterns of biological diversity in several ways, including the loss of unique microhabitats, habitat insularization and associated changes in dispersal and migration patterns and small- and large-scale soil erosion (Soulé & Kohm 1989). Despite the debate concerning the optimum size and spatial arrangements of viable nature reserves (Quinn & Hastings 1987, 1988; Gilpin 1988), it is clear that the effects of forest fragmentation will be greatest in relatively small fragments, which are likely to contain far fewer forest taxa than larger ones. It is important to understand the differences between the dynamics of forest fragments and continuous forest in order to provide guidance for the design and management of nature reserves and for conservation policy.

The effects of forest fragmentation on the diversity and abundance of plants and animals have been documented in several forest types (Galli et al. 1976; Gottfried 1979; Klein 1989; Aizen & Feinsinger 1994; Scariot 1996; Laurance & Bierregaard 1997; Benitez-Malvido 1998). Moreover, the potential influence of edge effects on the distribution, behaviour, and survival of plant and animal taxa has been studied in fragmented forest areas (Kapos 1989; Sizer 1992; Murcia 1995; Didhan 1997; Ferreira & Laurance 1997; Kapos et al. 1997). However, such studies are scarce in the Amazonian rain forest, although several examples show the importance of forest fragmentation on the composition and structure of populations and communities. For instance, the community of carrion and dung beetles in 1- and 10-ha forest fragments in central Amazonia has fewer taxa, sparser populations and smaller individuals than the continuous forest (Klein 1989), and the leaf litter invertebrates are influenced by edge effects and by forest fragment size (Didhan 1997). Forest fragmentation also reduces some bird guilds (Bierregaard & Lovejoy 1988; Stouffer & Bierregaard 1995) and primate taxa richness is lower in smaller forest fragments than in large ones (Rylands & Keuroghlian 1988). The sedentary habit and patchy distribution of plants (Levin & Kerster 1974; Platt & Weiss 1985) make them particularly susceptible to habitat destruction and to reductions in the size of reserves, which may cause changes in taxa composition and population sizes (Schemske et al. 1994). Despite the importance of plants in forest structure and function their response to habitat fragmentation has been studied less than for animals (Laurance & Bierregaard 1997), although in central Amazonia tree seedling abundance has been shown to decrease when reserve size decreases (Scariot 1996; Benitez-Malvido 1998).

This study was restricted to species from the palm family (Palmae) for three reasons. First, palms are among the dominant vascular plants in many tropical forests. Secondly, they are important components of the forest structure in Amazonian forests. Thirdly, in the Amazon basin palms are used by humans and serve as key resources for animals during periods of food scarcity (Terborgh 1986). Consequently, it is important to identify the effects of forest fragmentation on the taxa composition and diversity of palms. Negative effects of forest fragmentation on these parameters will influence the perceived and real value of forest fragments managed for conservation.

In this paper, the effects of short-term isolation (10–15 years) of forest patches on components and measures of floristic diversity in a set of 11 reserves representing four size classes (1-ha, 10-ha, 100-ha and continuous forest) at three sites in central Amazonia are reported. The questions addressed in this study are: (i) What are the floristic composition, taxa richness and diversity of palm communities occurring in fragmented and continuous forests of central Amazonia? (ii) Do reserves of similar size have more in common floristically than those of differing size? (iii) Are there any significant effects of reserve size and/or site on palm taxa composition, richness and diversity? (iv) What ecological processes underlie the effects of forest fragment size on these community parameters? The answers to these questions are discussed in terms of their implications for the design of reserves and biological conservation in Amazonia.


Study area

This study was conducted on reserves held by the Brazilian Institute of Amazonian Research (INPA) and the Smithsonian Institution, in the central Amazon basin, approximately 80 km north of the city of Manaus, Brazil (2°25′ S, 59°50′ W). These reserves are sites for the Biological Dynamics of Forest Fragments Project (BDFFP). Development of a series of cattle ranches in this region in the late 1970s and early 1980s in previously uncut lowland rain forest, created a set of protected forest fragments isolated from the continuous forest. A description of the development, ecology and climate of the reserves is given in Lovejoy et al. (1986) and Lovejoy & Bierregaard (1990).

Annual rainfall in Manaus is approximately 2200 mm and is strongly seasonal (RADAMBRASIL 1978). During the dry season, June–October, the mean monthly rainfall is commonly less than 100 mm. The rainy season peaks in February and March, with monthly rainfall of about 300 mm. Soils are nutrient-poor, yellow, alic latosols with a high clay content (Chauvel 1983). Upland soils are dominated by clay, while soils of low wet areas have a high sand content (Rankin et al. 1992). Intact forest has a closed canopy averaging 35 m in height, with occasional emergents reaching 50 m. The heavily shaded understorey is dominated by stemless palms.

Experimental design and field sampling

The reserves (isolated forest fragments and intact continuous forests) used for this study are located in the Dimona, Porto Alegre and Esteio (also known as Colosso) ranches. In Dimona and Porto Alegre, four reserves were established, one each of 1, 10 and 100 ha and continuous adjacent forest. At Esteio, where no 100-ha reserve was established, 1-ha, 10-ha and continuous forest reserves were monitored. The 1-, 10- and 100-ha forest fragments were created between 1980 and 1984, and isolated from continuous forests by 100–350 m of cleared land. When the forest is cut and left unburned, the remaining forest fragments are surrounded by secondary growth forest dominated by Cecropia spp. (Moraceae). When the cut forest is burned, the emerging vegetation is made up of a mixture of woody species dominated by Vismia spp. (Clusiaceae). The reserves are distributed across a 40-km east–west swath of what was previously intact, unbroken forest (Rankin-de-Merona et al. 1992), and the forest fragments are approximately square. In Porto Alegre the 1-ha reserve (reserve number 3114) is closer to the continuous forest and 100-ha reserve (3304) than to the 10-ha reserve (3209); in Esteio the 1-ha reserve (1104) is located between the 10-ha reserve (1202) and the continuous forest. Only in Dimona is the 1-ha reserve (2108) closer to the 10-ha reserve (2206) than to 100-ha and continuous forest.

A field survey of the palms was carried out from August 1993 to April 1994. In each reserve, 10 20 × 20 m plots (each 400 m2) were positioned randomly, but restricted to topographically flat areas, resulting in a total of 110 plots (4.4 hectares in aggregate). By restricting plots to flat areas, sampling biases were avoided that would have resulted if the specialist palm taxa that occupy either the low-lying or steep areas of this region were included.

Every palm individual in each plot was classified by its reproductive stage. For arborescent palm taxa (those that develop an aerial stem), seedlings were defined as plants lacking a well-defined aerial stem at the time of census; juveniles were individuals with a defined aerial stem but no signs of current or previous reproduction; and adults were plants that either were reproducing at the time of the census or showed evidence of having previously reproduced. For non-arborescent palm taxa (those with a subterranean stem), seedlings were defined as individuals that still had undivided leaves; juveniles were individuals with incipient or well-developed division of the leaves into pinnae; and adults were those individuals with evidence of current or previous reproduction.

In each 20 × 20 m plot, all palms were identified and recorded to the variety level when possible, following Henderson (1995). For the analyses described below, each taxon, including varieties, was treated independently. Unidentified taxa were classified as morphotaxa, and voucher specimens were prepared to compare with material deposited at the herbaria of INPA and CENARGEN (The National Research Center of Genetic Resources and Biotechnology).

Data analysis

Taxa–area curves

The shape of the taxa–area curve is a good indicator of the degree to which sampling effort captures the taxa present in the area sampled. An asymptote that is maintained as sampled area increases indicates that even the rare taxa have been detected. The slope of the taxa–area curve itself also indicates the degree to which the sampling effort captured the total taxa pool. The cumulative number of taxa in each life stage regressed on the log-transformed sampled area was used to estimate the slope of the taxa–area curves in each of the 11 reserves. To detect possible differences among reserve sizes and among sites (blocks) with respect to the slope of the taxa–area curve, the slopes of the regressions were used as response variables to perform randomized incomplete block design analysis of variance (anova), with type III sum of squares, available on the MGLH procedure of systat (1992). Differences among reserve sizes in the slope of the taxa–area curves would indicate that the reserve sizes differed in their ability to sample all of the taxa present.


Taxa richness was estimated, and the Shannon index of diversity (H′) and the Evenness index (J′) calculated in all sampled reserves for each life stage and for all palms (all three life stages pooled). Using a randomized incomplete block design anova, possible differences were searched for among reserve sizes and among sites (blocks) on the above indices. Pairwise comparisons for sites (blocks) are not allowed in a random block design (Sokal & Rohlf 1995), but means and results of the anova test are presented.

Taxa richness (S) was defined as the total taxa number present in a reserve or site. The Shannon diversity index (Magurran 1988) was calculated as:

H′ = –Σpi lnpi where pi is the proportion of the sample represented by the ith taxa (pi = ni/N) and N is the total number of individuals. Possible values of H′ range between 1.5 and 3.5 and only rarely surpass 4.5, where high values indicate high diversity. Evenness (J′) was estimated by J′ = H′/ln S (Pianka 1975), and possible values range between 0 and 1.0, with high values representing high evenness.

Number of taxa per plot

To test whether the number of taxa per plot varied with reserve size within site, the taxa number per plot in each life stage was used, rather than taxa richness. Given two reserves of equal taxa richness, if one reserve maintains a higher taxa number per plot than the other reserve, it can be concluded that a random plot sampled may capture more taxa when drawn from the former than the latter reserve. A nested anova (type III sum of squares) available in the Proc GLM (SAS 1996) was used to search for effects of reserve size and site (both factors were fixed) on the number of taxa per plot in each life stage. When reserve size within site was significant, anova was used to estimate effects within each site separately. Results of these tests are applied only to the particular condition where the experiment was conducted (Federer 1955). When anova results showed differences among means, t-tests with adjusted P-values (produced by permutation available in the Proc MULTTEST; SAS 1996) were conducted (2000 permutation resamples) to detect the linear and quadratic trends using the logarithm of the reserve size.

Community similarity

To describe changes in taxa composition across reserve sizes and sites, the relative frequency of each taxa was recorded for each reserve size class and for each site. The index of similarity between two reserves or two sites (beta diversity) was estimated using the percentage similarity coefficient (Renkonen's PS) for each life stage. This coefficient varies from 0 (completely dissimilar) to 100 (completely similar). It is considered one of the best indices of similarity (Wolda 1981) and is defined as:

PS = Σmin (p1i, p2i)where pi j is the proportion of taxon i in sample jnji/Nj, and Nj is the number of individuals in sample j.


Community composition of reserves and sites

A total of 36 taxa from 11 genera was found in the 23 225 individuals recorded in the sampled areas (Table 1), with most taxa present in all sampled areas. Taxa not recorded in the sampled area of a given reserve size class (Table 1) did not appear to be present in the unsampled portions of that reserve size class (A. Scariot, personal observation). Of the 36 taxa, 11 occurred in only some reserve sizes, but there was no tendency for the smaller forest fragments to be missing taxa relative to the larger reserves. Of these 11 taxa, only three, Astrocaryum acaule, Bactris killipii and Bactris oligocarpa, did not occur in 1-ha forest fragments. Five taxa, Attalea maripa, Bactris balanophora, Bactris sp2., Geonoma maxima var. chelidonura and Geonoma sp., were absent from 10-ha forest fragments, while seven taxa, Astrocaryum acaule, Attalea maripa, Bactris maraja var. maraja, Bactris oligocarpa, Geonoma sp., Geonoma stricta var. stricta and Lepidocaryum tenue, were not present in the 100-ha forest fragments. Five taxa, Astrocaryum acaule, Bactris maraja var. maraja, Bactris sp2, Bactris oligocarpa and Lepidocaryum tenue, did not occur in the continuous forest but they were represented by only 206 individuals (0.009% of the total number sampled) and occurred mainly in 1-ha (172 individuals) and 10-ha (33) forest fragments, being almost absent from the 100-ha (one) forest fragments. The natural history of these taxa is poorly known and without information on their life spans it is difficult to know if any of these taxa had had enough time to establish and reproduce in the fragments since isolation. Astrocaryum acaule and Bactris maraja var. maraja are secondary taxa that can occur in naturally disturbed patches of continuous forest (Henderson 1995), as also seems to be the case for Bactris sp2. Immature individuals were found for all three (one juvenile of Astrocaryum acaule, four seedlings of Bactris sp2 and two seedlings and seven juveniles of Bactris maraja var. maraja). Bactris oligocarpa, although being characteristic of lowland rain forest (Henderson 1995), was not found in the continuous forest and had all its four individuals (two adults and two seedlings across the three sites) in the 10-ha forest fragments. Either this taxon was in the fragments before isolation or it had had enough time to establish and reach maturity since isolation. Lepidocaryum tenue is the only invasive taxon, and all individuals (188 across all life stages) were restricted to the 1- and 10-ha fragments in Dimona.

Table 1.  List of palm taxa (n = 36) occurring in plateaux in different reserve sizes and sites studied in central Amazonia. Presence of taxa are shown separately for reserve sizes and for sites. Taxa not occurring in continuous forests are either secondary (*), invasive (**) or of unclear status (***)
Taxa1 ha10 ha100 haContinuous forestEsteioDimonaPorto Alegre
Astrocaryum acaule* x    x
Astrocaryum gynacanthumxxxxxxx
Astyrocaryum sciophilumxxxxxxx
Attalea attaleoidesxxxxxxx
Attalea maripax  xxx 
Bactris acanthocarpa var. acanthocarpaxxxxxxx
Bactris acanthocarpa var. intermediaxxxxxxx
Bactris balanophorax xx x 
Bactris constanciaexxxxxxx
Bactris elegansxxxxxxx
Bactris gastonianaxxxxxxx
Bactris hirta var. hirtaxxxxxxx
Bactris hirta var. pulchraxxxxxxx
Bactris killipii xxxxxx
Bactris maraja var. maraja*xx  xx 
Bactris oligocarpa*** x  xxx
Bactris simplicifronsxxxxxxx
Bactris sp1xxxxxxx
Bactris sp2*x x  xx
Bactris tomentosa var. tomentosaxxxxxxx
Desmoncus phoenicocarpusxxxxxxx
Euterpe precatoria var. precatoriaxxxxxxx
Geonoma aspidiifoliaxxxxxxx
Geonoma deversaxxxxxxx
Geonoma maxima var. chelidonurax xxxxx
Geonoma maxima var. maximaxxxxxxx
Geonoma maxima var. spixianaxxxxxxx
Geonoma sp.x  xx x
Geonoma stricta var. strictaxx xxx 
Iriartella setigeraxxxxxxx
Lepidocaryum tenue**xx   x 
Oenocarpus bacabaxxxxxxx
Oenocarpus bataua var. batauaxxxxxxx
Oenocarpus minorxxxxxxx
Socratea exorrhizaxxxxxxx
Syagrus inajaixxxxxxx

Relatively few taxa (eight) were not found in all three sites (Table 1), and three were restricted to a single site. Bactris balanophora (five individuals) and Lepidocaryum tenue (188) were found only at Dimona, and Astrocaryum acaule (one individual) was found only at Porto Alegre. Although few taxa were restricted to a single site, none of the three sites included all of the 36 taxa observed in this study (Table 1).

Six palm taxa had subterranean trunks, 29 had aerial trunks and only one (Desmoncus phoenicocarpus) was a liana. There was no association between the number of taxa of each major habit (i.e. aerial or subterranean trunk) and reserve size (chi-square, d.f. = 3, P = 0.87) or site (chi-square, d.f. = 2, P = 0.89). Although most taxa had aerial trunks, only five (Attalea maripa, Euterpe precatoria var. precatoria, Oenocarpus bacaba, Oenocarpus bataua var. bataua and Socratea exorrhiza) were large arborescent taxa (up to 20 m tall). Six taxa were midstorey palms, reaching no more than 12–15 m tall: Astrocaryum gynacanthum, Bactris balanophora, Bactris constanciae, Iriartella setigera, Oenocarpus minor and Syagrus inajai. All other 25 taxa were confined to the understorey, which therefore dominated the composition of the community.

Most of the taxa–area curves reached an asymptote after sampling seven plots, indicating that data from 10 plots per reserve were enough to include all palm taxa (data not shown). Also, no taxon was found in the unsampled area of a reserve or site that had not been recorded in the sampled plots. Curves of the log abundance–taxa sequence were quite similar among sites and reserve sizes (Scariot 1996). The anovas performed on each life stage (in which the dependent variable is the slope of the cumulative taxa count regressed on the logarithm of the sampled area) revealed no significant differences among reserve sizes and among sites in terms of the slope of the curve (for seedlings, juveniles, adults and the total sampled, all values of F were between 0.008 and 2.6; P = 0.17–0.99). In all sites and reserves, the curves followed the log-normal taxa–abundance distribution that is characteristic of speciose communities (Preston 1962; May 1975), with very few highly abundant taxa and with most taxa occurring at low densities.

Palm community diversity in forest fragments

A comparison of richness (S), Shannon (H′) and Evenness (J′) indices (Table 2) showed that similar values were obtained for all reserve sizes (total community, all taxa included). Total numbers of individuals/0.4 ha were similar for all three sites (1994–2214).

Table 2.  Comparison of number of individuals per 0.4 ha, taxa richness, diversity (Shannon) and Evenness indices (mean and 1 SE) of the palm community in different reserve sizes in central Amazonia. Means are of the three sites for each reserve size, except for the 100-ha reserves which are represented by only two sites
   1 ha
 mean ± SE
 10 ha
 mean ± SE
 100 ha
 mean ± SE
 Continuous forest
 mean ± SE
Individuals per 0.4 ha 1699.33 ± 204.321593.33 ± 124.602360.50 ± 36.502875.33 ± 323.99
Taxa richness (S)Adult
15.67 ± 2.02
20.67 ± 1.20
21.00 ± 0.58
24.00 ± 1.73
15.00 ± 3.21
21.33 ± 2.18
23.00 ± 1.00
26.00 ± 1.53
15.50 ± 0.50
21.00 ± 0.00
25.50 ± 1.50
26.00 ± 1.00
17.00 ± 1.52
22.67 ± 0.33
24.33 ± 1.20
26.67 ± 0.33
Diversity (H′)Adult
2.08 ± 0.20
2.21 ± 0.09
1.91 ± 0.30
2.15 ± 0.22
1.99 ± 0.18
2.25 ± 0.07
2.20 ± 0.15
2.39± 0.11
2.04 ± 0.06
2.35 ± 0.17
2.29 ± 0.29
2.45 ± 0.22
2.10 ± 0.15
2.16 ± 0.15
2.17 ± 0.13
2.32 ± 0.09
Evenness (J′)Adult
0.76 ± 0.04
0.73 ± 0.02
0.63 ± 0.09
0.68 ± 0.06
0.75 ± 0.02
0.74 ± 0.01
0.70 ± 0.05
0.74 ± 0.03
0.74 ± 0.01
0.77 ± 0.05
0.71 ± 0.10
0.75 ± 0.07
0.74 ± 0.03
0.69 ± 0.05
0.68 ± 0.03
0.70 ± 0.03

Taxa richness

anova did not detect any significant differences in taxa richness using data from among reserve sizes (for seedlings, juveniles, adults and the total, F-values were 0.12–2.6; P = 0.16–0.94, d.f. = 3). Taxa richness was similar across sites (adult, 15–17; juvenile, 20–21.7; seedling, 23–23.5; total, 24.5–26.7). There were no significant differences among sites (F = 0.24–1.7; P = 0.28–0.79, d.f. = 2). This suggests that the taxa richness of the palm community has not been affected by the size or the location (block) of the reserve since the fragments were established. When taxa not sampled in the continuous forest (Table 1) were removed from the data, the results changed for the youngest life stage. Reserve size still did not affect adults (F = 0.24; d.f. = 3, P = 0.87) or juveniles (F = 1.35; d.f. = 3, P = 0.36), but there were nearly statistically significant differences between reserve sizes with respect to the taxa richness of seedlings (F = 4.46; d.f. = 3, P = 0.07) and, as a result, for the total (F = 4.29; d.f. = 3, P = 0.07). When Tukey tests were applied taxa richness was not statistically different between 100-ha and 1-ha forest fragments (P = 0.085) or between continuous forest and 1-ha forest fragments (P = 0.073). of five taxa not sampled in the continuous forest, only Lepidocaryum tenue is an invasive species and it had already established reproductive populations in the 1- and 10-ha fragments. There were no significant effects of site (F = 0.19–1.25; P = 0.36–0.83, d.f. = 2, anova).

Shannon diversity and Evenness

The Shannon index of diversity (Table 2) did not differ significantly among reserve size categories (F = 0.09–2.08; P = 0.22–0.96, d.f. = 3). The range between sites was also small (adult, 1.96–2.15; juvenile, 2.08–2.43; seedling, 1.87–2.49; total, 2.13–2.59) but anova detected significant differences among the three sites (blocks) for seedlings, juveniles and the total (F = 12.85–13.61; P = 0.009–0.011, d.f. = 2), but not for the adult stage (F = 0.37; P = 0.7, d.f. = 2).

The Evenness index (Table 2) also showed no significant differences among reserve sizes (F = 0.10–1.0; P = 0.46–0.95, d.f. = 3). Among sites J′ varied between adults (0.72–0.76), juveniles (0.67–0.78), seedlings (0.59–0.79) and the total (0.65–0.78). Differences in evenness among sites (blocks) were significant for seedlings (F = 12.67; P = 0.01, d.f. = 2) and for the total (F = 8.37; P = 0.02, d.f. = 2) but not for adults (F = 0.54; P = 0.61, d.f. = 2) or juveniles (F = 4.13; P = 0.08, d.f. = 2).

Shannon and Evenness indices are affected by the number of taxa and the number of individuals sampled (Magurran 1988). Dimona was the site with both the highest numbers of taxa and individuals and this contributed to its increased Evenness indices.

Taxa number per plot

Number of taxa per plot differed significantly among sites at all life stages. Plots at Dimona had significantly higher numbers of adults, juveniles, seedlings and total taxa (P < 0.05 for all comparisons, Tukey test) than plots at Esteio and Porto Alegre (Fig. 1).

Figure 1.

Mean number of taxa per plot in each life stage (mean and 1 SE) by site. ‘Total’ represents all life stages pooled. Bars with different superscript letters for the same life stage differ significantly at P < 0.05.

The effect of reserve size within site on the number of taxa per plot was statistically significant except for the juvenile stage (Table 3), and the effect of reserve size was therefore compared within each site for all other life stages individually. Statistically significant differences among reserve sizes in the number of adult taxa per plot were detected by anova in Esteio (P = 0.04, F = 3.29, d.f. = 2), and Porto Alegre (P = 0.04, F = 2.83, d.f. = 3), but not in Dimona (P = 0.60, F = 0.61, d.f. = 3). Differences among reserve sizes were significant at all sites for seedlings (P = 0.0001, F = 9.8–14.2; d.f. = 2 for Esteio, d.f. = 3 for other sites) and total taxa (P < 0.005, F = 6.95–9.97). Despite anova being significant, the t-test did not detect linear or quadratic trends in the regression of number of adult taxa on the log reserve size (adjusted P-values = 0.49–1.0; data in Table 4). Linear trends were detected for seedlings in all sites (adjusted P-value < 0.05) but a quadratic trend was detected only in Esteio (adjusted P-value < 0.05, at other sites P = 0.99–1.0). Linear trends were detected for the total taxa at all sites (adjusted P-value < 0.05) but no quadratic trend was detected (adjusted P-values = 0.45–1.0).

Table 3.  Nested anova for the effect of reserve size on mean number of taxa per plot within three sites. A separate anova is presented for each life stage. ‘Total’ represents all three life stages pooled. Reserves sizes are 1 ha, 10 ha, 100 ha and continuous forest. Sites are Esteio, Dimona and Porto Alegre
Dependent variableEffectsd.f.MSFP
 Site/reserve size
 Site/reserve size
 Site/reserve size
 Site/reserve size
Table 4.  Mean and SE of number of taxa per plot within the three sites. ‘Total’ represents all three life stages pooled
Life stageReserve sizeEsteio mean ± SEDimona mean ± SEPorto Alegre mean ± SE
Adult1 ha 3 ± 0.57 6.9 ± 1.02 5.6 ± 0.54
 10 ha 5 ± 0.33 6.0 ± 0.71 3.5 ± 0.45
 100 ha 6.5 ± 0.30 4.6 ± 0.61
 Continuous forest 4.7 ± 0.44 5.9 ± 0.62 5.6 ± 0.56
Juvenile1 ha 9.7 ± 0.5311.5 ± 0.8310.4 ± 0.63
 10 ha10 ± 0.66 9.8 ± 1.19 9.5 ± 0.54
 100 ha12.9 ± 0.7310.6 ± 0.65
 Continuous forest10.1 ± 0.5012.4 ± 0.5410.8 ± 0.64
Seedling1 ha12 ± 0.8113.7 ± 0.3913.9 ± 0.58
 10 ha14.5 ± 0.7114.3 ± 0.5512.3 ± 0.61
 100 ha16.8 ± 0.5116.6 ± 0.60
 Continuous forest16.1 ± 0.4518.5 ± 0.5015.3 ± 0.61
Total1 ha15.4 ± 0.6017.4 ± 0.4916.4 ± 0.49
 10 ha17.6 ± 0.4916.6 ± 0.6014.6 ± 0.60
 100-ha18.7 ± 0.5518.6 ± 0.58
 Continuous forest18.1 ± 0.6420.2 ± 0.2917.6 ± 0.49

Community similarity

For all life stages, there was greater similarity between reserves of similar size than between reserves of extremely different sizes, as measured by Renkonen's PS index (Table 5). Thus, the 1-ha reserve was most similar to the 10-ha reserve, the 10-ha was more similar to the 1- and 100-ha than the continuous forest, the 100-ha was most similar to the 10-ha and continuous forest, which in turn was most similar to the 100-ha reserve. The continuous forest and 1-ha reserves had the lowest similarity in all life stages (Table 5). For all life stages, Esteio and Porto Alegre was the most similar pair of sites (Table 6). The extremely high value obtained for seedlings and for the total in this comparison may be a consequence of the high percentage of Oenocarpus bacaba seedlings that occurred at both these sites (Scariot 1996).

Table 5.  Percentage similarity index estimated for pairwise comparisons of reserve size classes (all taxa included). A separate analysis is presented for each life stage. ‘Total’ represents all three life stages pooled
Reserve size
Life stage 1 ha10 ha100 ha
Adult1 ha
10 ha
100 ha
 Continuous forest69.673.473.1
Juvenile1 ha
10 ha
100 ha
 Continuous forest78.780.281.1
Seedling1 ha
10 ha
100 ha
 Continuous forest64.674.181.5
Total1 ha
10 ha
100 ha
 Continuous forest69.176.081.8
Table 6.  Percentage similarity index estimated for pairwise comparisons of sites (all taxa included). A separate analysis is presented for each life stage. ‘Total’ represents all three life stages pooled
Life stage EsteioDimona
 Porto Alegre76.064.5
 Porto Alegre87.469.2
 Porto Alegre91.053.5
 Porto Alegre89.958.8


Coarse and fine grain richness of the palm community

The comparison between coarse and fine grain levels of taxa richness is often neglected in biogeographic studies but is important to the analysis of local community structure (Holt 1992, 1993; Holt et al. 1995). Studies of the whole community may fail to detect severe consequences of fragmentation for specific populations and taxa that were not uniformly distributed throughout a prospective nature reserve (Robinson et al. 1992; Holt et al. 1995). Fine grain measurement of taxa richness (number of taxa per plot) may uncover patterns of diversity that may not be detected by coarse grain measures of taxa diversity, and is thus pertinent to understanding how diversity is partitioned at fine spatial scales (Kohn & Walsh 1994).

Taxa with population densities that allow for high resilience and whose individuals have a long generation time should have lower extinction rates than taxa with low resilience and short generation times (Pimm 1991). Palms have quite long life cycles (Astrocaryum mexicanum only reproduces at ages of above 39 years, Sarukhán 1980; Euterpe globosa (= E. precatoria) at over 51 years, Van Valen 1975), although nothing is known of their resilience (which should vary among taxa and populations). Some of the most common taxa recorded here (e.g. Astrocaryum sciophilum and Attalea attaleoides) also occur and reproduce in the secondary growth and pasture land that surrounds the forest fragments, although they do not reach the same population densities as in the fragments or in continuous forest (A. Scariot, personal observation). The community of adult and juvenile palms may not have had enough time to respond to the fragmentation that occurred 10–15 years ago. How long it will take for effects to be expressed in the juvenile and adult life stage classes will depend on the length of individual life spans and on sensitivity to disturbance, both of which are taxa-specific.

The higher numbers of palm taxa per plot in all life stages in Dimona than in Esteio and Porto Alegre (Fig. 1) may be a result of environmental heterogeneity across sites or even due to biotic effects. For adults, the variation among reserve sizes within sites (Table 4) may be caused by either factors present pre-isolation or by differential mortality. It seems that the rate or spatial pattern of recruitment, or both, were affected by fragmentation, resulting in a general trend towards lower numbers of seedling taxa per plot in the smaller (1- and 10-ha) fragments than in the continuous forest in all the three sites. The consistently lower values for the 10-ha fragment in Porto Alegre were possibly due to the fact that part of this reserve had been flooded during the rainy period, which may have inhibited establishment and survival of some palm taxa even though all of the sample plots were located outside the flooded area. Some palm populations in the forest fragments may become extinct if they fail to reach the levels of recruitment that characterize the continuous forest (Scariot 1996). Decreases in the number of seedling taxa per plot may follow declines in reserve size because of factors such as: (i) decreased reproduction or viability (Allee effect); (ii) the probabilities of survival and reproduction (vital rates) of individuals of a given developmental stage may be directly proportional to population size (demographic stochasticity); (iii) temporal changes in vital rates (as a result of environmental stochasticity) may have affected all the individuals of a given stage similarly; or (iv) edge effects, which may be more prominent in smaller than in larger reserves. Seeds and seedlings are likely to be the first life stages to respond to habitat fragmentation. In fact, in the BDFFP seedling abundances decreased from continuous forest towards smaller forest fragments (Scariot 1996; Benitez-Malvido 1998). Decreases in seed production and seed dispersal, increases in seed and seedling predation, and a decline in the number and quality of suitable habitats can all contribute to a reduced number of seedling recruits in the forest fragments, and may also affect taxa composition.

Forest fragments may have less integrated biota, making them more invasible (Pimm 1984, 1991). The invading taxa are added to the existing taxa pool and may thus mask the reduction in the pool of closed forest taxa caused by forest fragmentation or even increased richness (Huston 1994) through immigrant seeds (Martinez-Ramos & Soto-Castro 1993). Even if the rare taxa were not already present as rarities inside small fragments, the habitat disturbance characteristic of such sites and the presence of edges (Kapos 1989; Murcia 1995) provides a considerable area of suitable habitat for invasive and secondary taxa. Indeed, eliminating taxa not sampled in the continuous forest from the analysis showed that for all life stages pooled (total), and for the seedling stage in particular, reducing reserve size may have a negative effect on taxa richness.

Community diversity and similarity

Although the Shannon index is influenced by the number of taxa and individuals present, no statistically significant effect of reserve size on H′ was detected. Thinning processes were either undetectable or not operating at this time (10–15 years since isolation). Although there was no difference in the Shannon diversity for adults among sites, the fact that the Dimona site differed from Esteio and Porto Alegre with respect to diversity for juveniles, seedlings and total highlights the possible heterogeneity of this ecosystem. In fact, density and taxa richness of tropical forest palms vary with edaphic (Clark et al. 1995) and topographic (Kahn & Castro 1985) factors. Dimona had significantly higher diversity because of its higher taxa richness and density of individuals (particularly seedlings) than the other two sites (Scariot 1996). This, in turn, influences the Evenness measures for seedlings and the total.

Within the BDFFP, both floristic composition and structure of the palm community in reserves of any given size resemble those of similar-sized reserves more closely than those of much smaller or larger reserves. There was a greater range in the index of percentage similarity at younger life stages, suggesting that the seedling stage was most strongly affected by fragmentation. Eventually, with time, this possible lower seedling recruitment may lead to taxa composition in small forest fragments diverging even more from continuous forest. The potential effects of forest fragmentation and habitat disturbance on floristic composition (the identity of surviving taxa) are as important as effects on taxa diversity per se.

If there has been no differential adult mortality among sites since the isolation of the forest fragments, then the observed differences in the similarity index for seedlings may be intrinsic to the sites. Palm densities vary according to habitat heterogeneity (Kahn & Castro 1985; Scariot et al. 1989; Clark et al. 1995) and may not have been fully controlled by only setting plots in plateaux. Higher similarity between Porto Alegre and Esteio may be a result of proximity of these two sites, but the ultimate cause of this similarity is unknown.

If taxa differ in their requirements for successful establishment, survival and reproduction, then the number of taxa present in a given area will be determined more by habitat diversity than by the size of the reserve. However, increasing the geographical area sampled or preserved generally also increases the number of habitats; consequently, area should have a positive influence on taxa diversity (Simberloff & Abele 1982; Kohn & Walsh 1994). Any set of small forest fragments may eventually contain more taxa than one continuous reserve of the same area because of the habitat heterogeneity sampled by the archipelago of forest fragments. For the palm community in central Amazonia, forest fragmentation increased the diversity of habitats that could have been occupied by invasive and secondary taxa. Although forest fragments have only been isolated for a short time (10–15 years), it is already possible to detect invasion of 1- and 10-ha forest fragments by Lepidocaryum tenue. Secondary taxa (Astrocaryum acaule, Bactris maraja var. maraja, Bactris sp2) also seem to be favoured by the habitat disturbance characteristic of forest fragments, although as yet they do not occur in the adult stage. As increasingly more land is disrupted, the rate of immigration from source areas to fragments will decline. Immigrants will be disproportionately represented by invasive taxa, for which there is the lowest threat from habitat destruction (Janzen 1986). Taxa requiring specialized habitats or continuous forest for their survival will have a lower probability of survival in forest fragments than invasive and secondary taxa. Rare and patchily distributed plants in the natural disturbance sites may become common in the surrounding matrix (Janzen 1984).

Implications for conservation

The richness and abundance of palm taxa, as well as their occurrence in all strata of the forest, and their importance as a vital food source for wildlife (Terborgh 1986; Spironelo 1992), make this family a primary target for conservation. Two major impacts of fragmentation can be noted. First, small reserves may have fewer taxa recruiting into them than areas of continuous forest do, contributing to a possible future change in the taxa composition of small compared to large reserves. Secondly, an invasive palm taxon occurs in forest fragments and secondary taxa also seem to be favoured by fragmentation.

Conservation decisions based only on diversity indices are risky when, for example, taxa that are characteristic of disturbed habitats can replace forest taxa or invade disturbed parts or whole forest fragments. Under these conditions, relatively small reserves could maintain (at least temporarily) diversity indices similar to undisturbed forest and, if immigration of invasive taxa is successful, relatively small reserves may even enjoy higher diversity than undisturbed forest. Acceptance of the null hypothesis (that there were no differences among fragment sizes in taxa diversity), even if statistically correct, could lead to the unsound conclusion that conservation measures are not needed. On the other hand, the null hypothesis might be rejected if taxa of disturbed habitats successfully colonize and increase composition in small reserves but this too may be biologically misleading if taxa present in small reserves may, in the long term, replace primary forest taxa through competitive interactions. Forest taxa may also be driven to extinction within small forest fragments if viable populations of all key taxa cannot be maintained in such areas.

Palm taxa of open or disturbed habitats can be conserved elsewhere or even on the edges of the reserves intended to conserve forest taxa. Rare taxa or those whose marginal distributions are included in the geographical area under consideration can only have viable population sizes if conserved in large reserves or throughout an extensive network of reserves. Palms cannot therefore be successfully protected for the long-term simply by setting aside large reserves in one or a few sites.


The comments of S. Mazer, D. A. Clark, J. Endler, J. Melack, C. Sandoval, G. Colli, R. Henriques and two anonymous referees significantly improved the quality of the manuscript. Discussions with J. Connell, P. Raimondi, S. Cooper, M. Pacheco, H. Paz and C. Cordeiro helped with the analysis. Financial support for field work was provided by WWF (Fundo Mundial para Conservação), the National Geographic Society, NSF, EMBRAPA, Fundação Botânica Margaret Mee, and INPA/SI (BDFFP), and is very appreciated. This is publication number 201 of the Biological Dynamics of Forest Fragments Project.

Received 6 January 1998revision accepted 2 June 1998