Diversity and dominance in palm (Arecaceae) communities in terra firme forests in the western Amazon basin

Authors

  • JAANA VORMISTO,

    1. Department of Systematic Botany, Institute of Biological Sciences, University of Aarhus, Herbarium, Build. 137, Universitetsparken, DK-8000 Aarhus C, Denmark
    Search for more papers by this author
  • JENS-CHRISTIAN SVENNING,

    1. Department of Systematic Botany, Institute of Biological Sciences, University of Aarhus, Herbarium, Build. 137, Universitetsparken, DK-8000 Aarhus C, Denmark
    Search for more papers by this author
  • PAMELA HALL,

    1. Department of Systematic Botany, Institute of Biological Sciences, University of Aarhus, Herbarium, Build. 137, Universitetsparken, DK-8000 Aarhus C, Denmark
    Search for more papers by this author
    • *

      Present address: Department of Biology, Florida State University, Tallahassee, FL 32309, USA.

  • HENRIK BALSLEV

    Corresponding author
    1. Department of Systematic Botany, Institute of Biological Sciences, University of Aarhus, Herbarium, Build. 137, Universitetsparken, DK-8000 Aarhus C, Denmark
    Search for more papers by this author

Henrik Balslev (fax +45 8613 9326; e-mail henrik.balslev@biology.au.dk).

Summary

  • 1Current discussions about the structure and functioning of tropical plant communities revolve around whether tropical forests are dominated by sets of widespread and locally abundant species and the relative importance of environmental variation and geographical distance as determinants of variation in species composition. We explored these questions by examining diversity and dominance patterns of palm communities in lowland rain forests in the western Amazon basin (Yasuni National Park, Ecuador and Iquitos-Pebas region, Peru).
  • 2We used multiple regression to analyse floristic similarity between sites as a function of environmental differences (in topographic position, elevation difference, amount of exchangeable cations and soil texture) and geographical distance. We studied dominance patterns by quantifying the correlations between local abundance and landscape frequency of species within and between the regions.
  • 3We found floristic composition to be highly variable and more strongly related to geographical distance than to environmental differences. Together, geographical distance and environmental differences explained 70–85% of the variation in floristic similarity.
  • 4Species abundant in one region were also often abundant in the other, although some dominant species in Iquitos-Pebas did not occur in Yasuni. Local dominance was more pronounced in Yasuni, where the dominant species formed a larger and less variable proportion of all individuals within transects.
  • 5Limited dispersal and, to a lesser extent, local environmental variation appear to act as determinants of floristic variation in the western Amazon basin. Oligarchic dominance patterns were observed, particularly in Yasuni.
  • 6Our results lend support to the hypothesis that dispersal across broader spatial scales plays a strong role in the assembly of local ecological communities, but indicate that such limitation supplements, rather than replaces, local ecological determinism.

Introduction

Neotropical lowland rain forests are among the most species-rich in the world (e.g. Gentry 1988; Balslev et al. 1998) and local species richness (alpha diversity) of trees and other vascular plants reaches one of its global peaks in the western and central Amazon basin with > 300 tree species of ≥ 10 cm diameter-at-breast-height (d.b.h.) per hectare. While the high alpha diversity of some Amazonian forests is well established, the larger-scale patterns of Amazonian plant diversity are much less known (e.g. Terborgh & Andresen 1998; Condit et al. 2002; Tuomisto, Ruokolainen & Yli-Halla 2003). A better understanding of these larger-scale diversity patterns, notably beta diversity (here defined as differences in species composition between local sites (transects) within a single broad habitat type, cf. Condit et al. 2002; Tuomisto, Ruokolainen & Yli-Halla 2003) is crucial not only for understanding how the high alpha diversity is maintained, but also for the development of informed conservation plans for the region (Tuomisto et al. 1995; Terborgh et al. 2002).

In the recent discussions of Amazonian plant distribution and diversity patterns three general views have emerged. The first proposes that terra firme forests in the Amazon basin are dominated by a limited set of ecologically superior species that combine high frequency with high local abundance, and form predictable oligarchies over wide areas; these dominant species are abundant and widespread because they are broadly adapted to suites of environmental conditions that extend over large areas (Pitman et al. 1999, 2001; cf. Poore 1968 for similar conclusions regarding the mixed-dipterocarp forests of Malaysia). Pitman et al. (2001) found that a limited number of common species dominated tree communities in the western Amazon basin, accounting for over 50% of trees at all scales. Low beta diversity of tree species has been found both at meso-scale (1–100 km2) in Colombian Amazonia (Duivenvoorden 1995) and over great distances in the western Amazon basin (Condit et al. 2002; but see Phillips et al. 2003; Tuomisto, Ruokolainen & Yli-Halla 2003; Tuomisto, Ruokolainen, Aquilar & Sarmiento 2003). Oligarchic forests also cover large areas in flooded or disturbed habitats in the Amazon basin (Peters et al. 1989).

Alternatively, species may be ecologically equivalent and local species composition is a result of species immigration and extinction. Species composition is unrelated to environmental factors, but exhibits strong spatial autocorrelation due to dispersal limitation (Hubbell 2001). While some support has been found for this view for tropical forest plant communities (Hubbell et al. 1999; Hubbell 2001), other recent studies have concluded that this theory alone is insufficient to explain the floristic similarity patterns observed over hundreds of kilometres (Condit et al. 2002; Phillips et al. 2003; Tuomisto, Ruokolainen & Yli-Halla 2003).

The third view is that terra firme forests in the Amazon basin grow in a fine-grained patchwork of different environments that create corresponding differences in plant species composition due to ecological differences among species (Gentry 1988; Tuomisto et al. 1995; Tuomisto, Ruokolainen & Yli-Halla 2003; Tuomisto, Ruokolainen, Aquilar & Sarmiento 2003). Several studies have reported that changes in species composition in neotropical forests correspond to variation in soil characteristics or topographic position at both local (Svenning 1999; Vormisto et al. 2000; Tuomisto, Poulsen, Ruokolainen et al. 2003) and larger scales (Tuomisto et al. 1995; Clark et al. 1995; Clark et al. 1999; Phillips et al. 2003; Tuomisto, Ruokolainen & Yli-Halla 2003; Tuomisto, Ruokolainen, Aquilar & Sarmiento 2003). The mechanisms underlying the three views are not necessarily mutually exclusive and finding evidence in favour of one of them does not therefore invalidate the others.

Here, we investigate differences in species composition (especially relevant to the latter two views) and dominance patterns (particularly relevant to the first) of palm communities in terra firme forests in the western Amazon. We assessed the relative importance of stochastic and environmental variables controlling floristic differences in palm species between local sites, and whether all palm communities in the western Amazon basin are dominated by the same, limited set of species? We used multiple linear regressions on similarity matrices to analyse the dependency of floristic composition on differences in environmental conditions and geographical distance (cf. Phillips et al. 2003; Tuomisto, Ruokolainen & Yli-Halla 2003). If floristic differences arise from neutral processes, then floristic similarity should strongly decrease with geographical distance, whereas if environmental site differences are important they should correlate strongly with the floristic data. We also analysed dominance patterns within and between two regions in the western Amazon basin (cf. Pitman et al. 2001).

Methods

study species

Palms are diverse and abundant, often even dominant in the forests of the western Amazon basin (e.g. Kahn & de Granville 1992; Henderson 1995; Terborgh & Andresen 1998) and include a range of habits from small understorey shrubs to tall canopy trees and lianas. Our study is therefore different from most previous studies, which focused on a single growth form, commonly trees (e.g. Duivenvoorden 1995; Pitman et al. 1999, 2001; ter Steege et al. 2000), although non-trees may constitute ≥ 50% of vascular plant alpha diversity in these forests (Duivenvoorden 1994; Balslev et al. 1998). Palms are easily distinguished from other plant families and recent taxonomies (Henderson 1995, 2000) make floristic comparisons among localities more reliable.

study regions

The Plio-Pleistocene dissected relief is composed of weathered Miocene marine-lacustrine deposits, but Quaternary fluvial deposits are also fairly extensive (Räsänen & Irion, in press). Our sampling was done in low rolling hills, 100–300 m above sea level, in the Yasuni National Park at the base of the Andes and, also, 600 km to the south-east on the flat plain near Iquitos and Pebas (Fig. 1). In both regions the climate is perhumid and hot; mean annual precipitation is c. 2800 mm in Yasuni and 3000 mm in the Iquitos-Pebas region, and the mean annual temperature in both regions is 26–28 °C (Marengo 1998; Romero-Saltos et al. 2001). The soils of the Iquitos-Pebas region are quite variable and may be very nutrient poor (Ruokolainen & Vormisto 2000) whereas the Yasuni region soils are less variable and, in general, richer in cations (Table 1; also cf. Tuomisto et al. 2002; Tuomisto, Poulsen, Ruokolainen et al. 2003).

Figure 1.

The location of 21 transects (5 × 500 m) in terra firme forest in two regions in the western Amazon basin, the Iquitos-Pebas in Peru and the Yasuni in Ecuador, used to study diversity and dominance patterns in palm communities.

Table 1.  Palm community and environmental characteristics in 21 transects (5 × 500 m) in terra firme forest in two regions in the western Amazon basin (see Fig. 1)
TransectIndividualsSpeciesFisher's αCation (Ca + K + Mg + Na) concentrationElevation difference (m)Valley/hill subunits% of clay% of silt% of sand
Iquitos-Pebas
 PI 1A1263336.20 1.6715.1   9/1136.345.518.2
 PI 1B1013326.29 5.3814.5  11/921.366.212.5
 PI 21076295.49 1.0219.0  10/1027.763.0 9.3
 PI 3 962264.92 0.47 8.5  18/223.649.926.5
 PI 41024285.32 0.9816.0   9/1121.348.730.0
 PI 5A 988336.57 0.30 9.9  11/923.059.717.3
 PI 5B1162315.85 0.33 6.7  20/0 7.671.720.7
 SA A1478274.69 0.6415.7   6/1428.066.7 5.3
 SA B1511233.85 0.5017.0  13/748.045.0 7.0
 TA A 989223.9914.2421.8   4/16 6.890.2 3.0
 TA B 785244.68 7.3322.8  13/710.362.727.0
 Mean1114285.26 2.9915.211.3/8.723.160.916.1
 SD 2233.970.92 4.40 5.2 4.7/4.712.313.5 9.4
Yasuni
 YA 11360233.93 3.1532.7   5/1532.044.024.0
 YA 41148223.86 8.6929.9   4/1636.748.315.0
 YA 8 878285.52 4.3326.1  12/841.947.810.2
 YA 101143264.74 6.4126.2   3/1722.155.822.1
 YA 141137223.87 5.5329.1   3/1727.849.810.2
 YA 161366213.5219.5161.3   3/1728.245.326.5
 YA 191073264.80 6.7634.9   8/1236.949.613.5
 YA 221311233.9617.8455.1   5/1535.939.924.2
 YA 251607223.6120.5442.5   4/1629.638.731.7
 YA 281505213.45 3.5341.1   5/1533.546.520.0
 Mean1253234.13 9.6337.9 4.5/15.532.546.621.0
 SD 2182.410.67 6.8912.2 2.8/2.8 5.7 5.0 6.5

sampling design

Linear transects, each 5 m by 500 m long, were cut across the topography in old-growth forests on non-inundated ground (terra firme). There were 10 transects in the Yasuni region (total sampling area of 2.5 ha) and 11 transects in the Iquitos-Pebas region (2.75 ha). The geographical distance between transects was 0.5–170 km in the Iquitos-Pebas region and 1–18 km in the Yasuni region. In each transect we recorded all palm individuals, including seedlings. For clonal and multistemmed species each ramet was counted as a separate individual. Specimens from the two regions were compared to ensure a standardized taxonomy. Voucher specimens were collected (with additional vouchers for all individuals that we could not assign with certainty to a taxon represented by an earlier collection) and deposited in herbaria in Turku (TUR), Iquitos (AMAZ), Lima (USM) and Aarhus (AAU) (Iquitos-Pebas taxa) or Aarhus (AAU) and Quito (QCA) (Yasuni taxa). When possible, identification was taken to the variety level following the nomenclature of Henderson (1995, 2000) (Table 2). The conspecific varieties in Table 2 were easily distinguished in the field and, for the purpose of our analysis, were treated as if they were species (see also Svenning 1999, 2001b). Some morphospecies could not be matched definitively to the taxa included in Henderson (1995, 2000), and were given tentative names (Attalea cf. plowmannii, Bactris aff. major var. infesta and Bactris sp2), but were also analysed as species. Bactris acanthocarpa and B. macroacantha, and Geonoma deversa and G. poeppigina, were lumped in the analyses, because young individuals could not always be reliably distinguished. To ensure that our results were robust, we recomputed the similarity matrix regressions and log-log correlation for the abundances of the shared species between the regions (see below) combining varieties within species and excluding complex species groups and morphospecies, and obtained qualitatively similar results. We collected three to six surface soil samples (top 5–20 cm of the A horizon, excluding the organic O horizon) for laboratory analyses at hilltops and valley bottoms in each transect. Soil samples were analysed in the Agricultural Research Centre of Finland, Jokioinen, for exchangeable cations Ca, K, Mg and Na (extraction in 1 m ammonium acetate at pH 7, analysis with atomic absorption spectrophotometer).

Table 2.  Palm species found in 21 transects (5 × 500 m) in terra firme forest in two regions in the western Amazon basin
Yasuni regionIndividualsIquitos-Pebas regionIndividuals
Oenocarpus bataua  3137Lepidocaryum tenue  2160
Prestoea schultzeana  2536Astrocaryum murumuru  1330
Iriartea deltoidea  1999Socratea exorrhiza  1238
Geonoma macrostachys var. macrostachys  1915Geonomoa macrostachys var. macrostachys  1098
Phytelephas tenuicaulis   390Oenocarpus bataua   878
Ammandra dasyneura   281Geonoma stricta var. piscicauda   793
Astrocaryum chambira   279Iriartea deltoidea   793
Euterpe precatoria   246Attalea cf. plowmannii   542
Attalea maripa   237Euterpe precatoria   399
Wettinia maynensis   197Iriartella stenocarpa   370
Astrocaryum murumuru   151Hyospathe elegans   291
Geonoma brongniartii   144Geonoma maxima var. chelidonura   286
Chamaedorea pinnatifrons   141Geonoma deversa   181
Geonoma stricta var. stricta   140Geonoma macrostachys var. acaulis   142
Aiphanes ulei   109Wendlandiella gracilis var. polyclada   129
Geonoma maxima var. maxima   105Attalea maripa   124
Hyospathe elegans    94Phytelephas macrocarpa   120
Geonoma aspiidifolia    92Geonoma poeppigiana   106
Socratea exorrhiza    74Geonoma brongniartii    95
Geonoma stricta var. trailii    47Astrocaryum chambira    93
Bactris corossilla    43Oenocarpus mapora    93
Geonoma stricta var. piscicauda    33Bactris fissifrons    89
Bactris maraja var. juruensis    27Geonoma camana    82
Bactris concinna    26Bactris simplicifrons    69
Chamaedorea pauciflora    20Chelyocarpus repens    68
Geonoma triglochin    15Bactris maraja var. maraja    58
Bactris marara var. maraja    13Geonoma stricta var. trailii    58
Bactris schultesii    10Bactris bifida    52
Desmoncus mitis var. mitis     9Bactris macroacantha    51
Oenocarpus mapora     6Bactris hirta var. lakoi    48
Aphandra natalia     3Geonoma arundinacea    41
Desmoncus giganteus     3Bactris sp2    38
Desmoncus orthacanthos     3Oenocarpus balickii    38
Attalea insignis     1Bactris hirta var. hirta    35
Bactris simplicifrons     1Bactris halmoorei    32
Syagrus sancona     1Bactris aff. major var. infesta    24
  Bactris sphaerocarpa    23
  Geonoma stricta var. stricta    23
  Phytelephas tenuicaulis    19
  Prestoea schultzeana    18
  Bactris acanthocarpa    17
  Bactris maraja var. trichospatha    17
  Chamaedorea pauciflora    17
  Desmoncus mitis var. tenurrimus    16
  Desmoncus polyacanthos    15
  Bactris schultesii    13
  Geonoma maxima var. maxima     7
  Aiphanes ulei     6
  Bactris killipii     5
  Desmoncus giganteus     5
  Desmoncus mitis var. mitis     2
  Geonoma leptospadix     2
  Chamaedorea pinnatifrons     1
  Wettinia maynensis     1
Total12 528Total12 251

data analyses

We analysed beta diversity or differences in species composition with analyses of floristic similarity matrices computed with Sørensen (presence-absence data) and Steinhaus (abundance data, measured as the density of each species) coefficients (Legendre & Legendre 1998). Major patterns among sites were investigated with principal coordinates analyses (PCoA, Legendre & Legendre 1998). Multiple linear regressions on similarity (or distance) matrices were used to analyse the dependency of floristic variation on differences in environmental conditions and geographical distances (Legendre et al. 1994; Legendre & Legendre 1998; cf. Tuomisto, Ruokolainen & Yli-Halla 2003). Elevation difference (between the lowest and highest points within each transect), and a valley-hill index (proportion of 25-m sections ≥ 7 m above the lowest point), were computed for each transect. As palm roots have been reported to grow to maximum depths of ‘several metres’ (Tomlinson 1990), we assumed that roots of palms growing ≥ 7 m above the valley bottom would not penetrate to this level. Mean values for each exchangeable cation were calculated for each transect, and their sums then ln-transformed (cf. Tuomisto, Poulsen, Ruokolainen et al. 2003). Soil texture was represented by percentage weight of clay (particle size < 0.002 mm), silt (0.002–0.063 mm), and sand (0.064–2 mm). Each of the four environmental descriptors was converted to a similarity matrix by computing 1 − (the Euclidean distance/the maximum Euclidean distance) between the values for each pair of transects. Geographic distance in kilometres was logarithmically transformed, and then converted into similarity values. Therefore, all matrices in the regression analysis were similarity matrices. The multiple regression analyses were done using backward elimination, assessing significance using a permutation test based on 999 permutations (Legendre et al. 1994). Backward elimination proceeds by excluding the explanatory factor with the least significant partial regression coefficient and repeating the process until only significant (at P < 0.05, after Bonferroni correction) explanatory factors remain in the model.

Computing multiple regressions against both spatial and environmental, only environmental, or only spatial variables allowed estimation of total variance explained (RT = total R2), total environmental variance (RE), and total spatial variance (RS). Use of similarity matrices means that we partitioned variation in floristic similarity rather than in species frequencies (cf. Dutilleul et al. 2000). Pure spatial (RPS), pure environmental (RPE), mixed spatial-environmental variation (RMX), and unexplained fractions (RUN) could then be calculated as: RPS = RT − RE; RPE = RT − RS; RMX = RE + RS − RT; and RUN = 1 − RT (cf. Borcard et al. 1992). Our full (RT) models were the final models after backward elimination. All variables except interplot distance were treated as environmental.

We used Fisher's α index as the measure of diversity within each transect. Fisher's α provides a species richness measure, which is robust to variation in the number of individuals (Condit et al. 1996). Fisher's α is defined as: S = α ln(1 + N/α), where S is species number, N is number of individuals and α is the sole parameter (Fisher et al. 1943). We calculated α for each pair of S and N using Newton's method (Condit et al. 1996; Leigh & Loo de Lao 2000).

We analysed dominance patterns within and across the regions using the approach of Pitman et al. (2001). When testing for correlation between abundance (number of stems within a transect) and frequency (proportion of transects occupied) within a region, we calculated mean abundance for transects in which a given species had at least one individual.

All resemblance matrices and the PCoA were computed in R-package 4.0, using the Geographic Distances module to calculate interplot distances (Casgrain & Legendre 2001). Multiple linear regressions on similarity matrices were analysed with Permute! 3.4a (Legendre et al. 1994), both programmes available at http://www.fas.umontreal.ca/BIOL/Legendre/indexEnglish.html.

Results

floristic and diversity patterns

A total of 24 779 individuals of 64 palm species were recorded (785–1607 per transect, Tables 1 and 2). There were 36 species in Yasuni and 54 in Iquitos-Pebas with 10 and 28 unique species, respectively. There were 21–33 species per transect, with Iquitos-Pebas being the more diverse region in terms of both species number and Fisher's α (Table 1).

The two regions were floristically distinct from each other, with floristic composition more similar within Yasuni than within the Iquitos-Pebas region (Fig. 2), as Iquitos (but not Pebas) transects were more variable (Table 3). Floristic similarity decreased approximately linearly with log-transformed distance, with higher r2 for Sørensen than for the Steinhaus index (Fig. 3). The similarity of individual environmental variables also decreased with log-transformed geographical distance, albeit more weakly (r2 values: elevation difference, 0.21; cation quantity, 0.19; soil texture, 0.08; valley-hill index, 0.08).

Figure 2.

Ordination diagrams (principal coordinate analysis) of 21 transects (5 × 500 m) in terra firme forest in two regions in the western Amazon basin, the Iquitos-Pebas in Peru and the Yasuni in Ecuador. Ordinations were based on the palm species composition. Similarity was measured by Sørensen (presence-absence data) and Steinhaus (abundance data) indices.

Table 3.  Average (± standard deviation) floristic presence-absence (Sørensen) and abundance (Steinhaus) similarity values. The calculations are based on 10 transects in the Yasuni and 11 transects in Iquitos-Pebas region (four close to Iquitos, seven close to Pebas)
 Sørensen indexSteinhaus index
Between Yasuni and Iquitos-Pebas0.52 (± 0.06)0.25 (± 0.09)
Within Yasuni0.80 (± 0.07)0.55 (± 0.15)
Within Iquitos-Pebas0.68 (± 0.10)0.43 (± 0.17)
Within Iquitos0.63 (± 0.15)0.35 (± 0.22)
Within Pebas0.76 (± 0.07)0.49 (± 0.20)
All transects0.62 (± 0.14)0.36 (± 0.18)
Figure 3.

Similarity of the palm flora in terra firme forest between all unique pairs of the 21 (5 × 500 m) transects in two regions in the western Amazon basin, as a function of distance between the transects. Graph a is based on presence-absence data (Sørensen index) and graph b is based on abundance data (Steinhaus index). T2 values for least-square regressions against linear and logarithmic distance are given in the legend.

Geographic distance and environmental similarity variables together explained 70–85% of the variation in floristic distance (Table 4, Fig. 4). Geographic distance explained the largest proportion of the variation in floristic differences between the transects, especially for species presence-absence. The importance of environmental variables nearly doubled when abundance rather than presence-absence was analysed (Table 4, Fig. 4). Among the environmental similarity variables, only cation concentration and soil texture had significant effects on the variation in floristic differences.

Table 4.  Regression of floristic similarity against five explanatory similarity matrices. The full model shows the initial standardized regression coefficients and P-values, while the final model shows the variables left after the backward elimination procedure (at P < 0.05 level after Bonferroni correction). Significance levels were tested using 999 permutations
VariableSørensen index; regression coefficients (P value)Steinhaus index; regression coefficients (P value)
Full model
 Amount of cations  0.23 (0.002)0.30 (0.001)
 Texture  0.20 (0.001)0.21 (0.001)
 Elevation range−0.03 (0.257)0.09 (0.096)
 Valley-hill index  0.03 (0.216)0.07 (0.104)
 Distance  0.73 (0.001)0.50 (0.001)
 R2  0.85 (0.001)0.71 (0.001)
Final model
 Amount of cations  0.22 (0.002)0.35 (0.001)
 Texture  0.21 (0.001)0.21 (0.001)
 Distance  0.72 (0.001)0.54 (0.001)
 R2  0.85 (0.001)0.70 (0.001)
Figure 4.

Relative importance of different factors explaining variation in palm floristic similarities between sites. The diagrams are based on multiple regression to partition the variation into four independent components. Only explanatory variables that had statistically significant contributions to the final model after backward elimination were included in the partitioning (see Table 4). (a) Presence-absence data (Sørensen index). (b) Abundance data (Steinhaus index).

dominance patterns

Oligarchic dominance was evident in both regions, both separately and in combination (Table 2), but was much more pronounced in Yasuni (Fig. 5). Six species (Astrocaryum murumuru, Attalea maripa, Geonoma macrostachys var. macrostachys, Iriartea deltoidea, Oenocarpus bataua and Socratea exorrhiza) occurred in all 21 transects and constituted 61% and 38% of the individuals in Yasuni and Iquitos-Pebas, respectively. One of these, Oenocarpus bataua, was the most common species in Yasuni, with abundance ranging from 4 to 50% of all individuals in a transect. The most common species in the Iquitos-Pebas region (Lepidocaryum tenue, Table 2), however, occurred in only six of 11 transects (abundance 13–45% of all individuals, Table 5). Within a transect at least 60% of all individuals were accounted for by two to three species from a group of five in Yasuni, but three to five species from a group of 13 in Iquitos-Pebas (Table 5). Species with over 1000 individuals (n = 4 in both cases, Table 2) constituted a smaller proportion of the total number of palms in Iquitos-Pebas than in Yasuni (Table 6, Fig. 5), but species with over 100 individuals accounted for similarly high (> 90%) proportions in both regions, while representing less than half of the species.

Figure 5.

Rank-abundance of palm species with densities of > 100 individuals in Yasuni (16 species, black bars) and Iquitos-Pebas transects.

Table 5.  Relative abundances (% of palm individuals) of the dominant species per transect. Dominant palms were defined here to include species that together made up at least 60% of the individuals in a transect
Iquitos–Pebas regionPI 1API 1BPI 2PI 3PI 4PI 5 API 5BSA ASA BTA ATA B
Astrocaryum murumuru 18.513.4  9.719.318.7   13.320.6
Attalea cf. plowmannii         15.716.4
Euterpe precatoria   12.1       
Geonoma deversa        5.6    
G. macrostachys var. macrostachys17.516.613.426.513.9      
G. maxima var. chelidonura      10.9    
G. stricta var. piscicauda   9.7 8.4 9.3  8.1  15.5  
Iriartea deltoidea  12.8 9.2 9.7      8.6
Iriartella stenocarpa     14.918.7    
Lepidocaryum tenue  25 12.734.623.931.544.7  
Oenocarpus bataua     15.3 21.312.6  
Socratea exorrhixa 8.212.5  8.96.8  6.1  38.912.7
Wendlandiella gracilis var. polyclada          14.5
Total percentage of individuals66.760.167.166.860.264.865.268.365.967.964.2
Yasuni regionYA 1YA 4YA 8YA 10YA 14YA 16YA 19YA 22YA 25YA 28 
Attalea maripa    8.1        
Geonoma macrostachys var. macrostachys22.13520.615.215.9 22.3    
Iriartea deltoidea 20.4 18.31517.7 24.411.817.7 
Oenocarpus bataua50.110.338.838.133.9 17.9  44.1 
Prestoea schultzeana     44.620.639.453.3  
Total percentage of individuals72.265.767.571.664.862.360.863.865.161.8 
Table 6.  Dominance patterns in the palm communities. The mean and (standard deviation) for transects at each site is given for the proportion of individuals for shared species
 Iquitos- PebasYasuni
Species with ≥ 1000 individuals
 Number of species  4  4
 Proportion of all species (%)  711
 Proportion of all individuals (%)4877
Species with ≥ 100 individuals
 Number of species1816
 Proportion of all species (%)3344
 Proportion of all individuals (%)9096
Species shared between regions
 Number of species2626
 Proportion of all species (%)4872
 Proportion of all individuals (%)63 (± 22.9)96 (± 3.9)
Species shared within a region
 Number of species  714
 Proportion of all species (%)1339
 Proportion of all individuals (%)49 (± 19.1)92 (± 3.8)

Twenty-six species were shared between the Yasuni and Iquitos-Pebas regions, and made up the majority of all individuals (Table 6), although values were much lower and more variable in Iquitos-Pebas (Fig. 6). More species occurred in all transects in Yasuni than in the geographically separated transects in Iquitos-Pebas (Table 6). However, in the seven clustered Pebas transects the number of shared species was similar to Yasuni (13 vs. 14 species). Nevertheless, Pebas’ shared species made up a much smaller and more variable proportion of the total number of individuals (average 59% ± 23.6%).

Figure 6.

The relative abundance (% of palm individuals) of the 26 species shared between regions per transect (grey bars Iquitos-Pebas, black bars Yasuni; the order of the transects is the same as in Table 1).

Average abundance was positively correlated with frequency across transects within each separate region (Yasuni, Spearman correlation rs = 0.62, P = 0.0001; Iquitos-Pebas, rs = 0.46, P = 0.0006) as well as for the regions combined (only the shared species, rs = 0.69, P = 0.0001). For the 26 species that occurred in both regions, abundance in the two regions was positively (log-log) correlated (rs = 0.39, P = 0.05, Fig. 7). However, when the non-shared species were also included in the analysis, the correlation between abundance in the two regions disappeared (rS = 0.027, P = 0.83, Fig. 7).

Figure 7.

Correlation of palm species abundances. Open circles show species shared between regions (26 species), for which the correlation is shown (as a linear least-squares regression fit).

Discussion

floristic variation

Geographic distance and environmental differences influence palm species composition, explaining 70–85% of the variation in floristic similarity. Contrary to Tuomisto et al. (2003) and Phillips et al. (2003) we found geographical distance to be the more important, notably for the rarer species, perhaps reflecting the smaller extent of our study and sampling differences, respectively. In NW Borneo, environmental control of tree species similarities was strongest at the regional scale, and geographical control at finer scales (Potts et al. 2002; also cf. Condit et al. 2002; Tuomisto, Ruokolainen & Yli-Halla 2003).

Palm floristic similarity decreased strongly and regularly with increasing logarithmic distance. Tuomisto et al. (2003) also reported a logarithmic distance decay in the similarity of fern and melastome communities up to 600 km, but no obvious distance decay at 600–1600 km, and Condit et al. (2002) showed similarity between tree communities in 1-ha forest plots to decay strongly at 0.01–5 km, more weakly at 5–100 km, but not at 100–1400 km.

The weak logarithmic distance decay in the similarity of the environmental variables indicates that they can only account for a minor part of the floristic effect (as also shown by variance decomposition). According to the neutral theory, spatially limited seed dispersal could cause floristic similarity to decline with logarithmic distance (Hubbell 2001; Chave & Leigh 2002; Condit et al. 2002). Several other studies have found strong indications of dispersal limitation for palms and other plant species at both fine and broad scales in tropical forests (e.g. Condit et al. 2000, 2002; Svenning 2001a; Charles-Dominique et al. 2003). On the other hand, it cannot alone explain distribution patterns of tree species at all scales in forests of Panama and western Amazonia (Condit et al. 2002), nor for the palms in this study.

Interestingly, when floristic distance was based on abundance data (Steinhaus index) rather than on presence-absence data (Sørensen index), total variance explained was lower, mostly due to reduced importance of geographical distance (Table 4, Fig. 4). The Steinhaus index emphasizes quantitative differences in the abundance of common species, which were also more widespread and hence less spatially constrained, and this as well as stochastic noise in local densities, could contribute to the reduction.

Variation in floristic similarity matrices was related to variation in matrices for extractable cations and soil texture, confirming relationships between edaphic factors and palm species distribution and abundance patterns in the Neotropics (Clark et al. 1995; Sollins 1998; Vormisto et al. 2000; Svenning 2001b).

Edaphic differences (a wide variety of poor soils in Iquitos-Pebas region, high concentrations of cations in Yasuni) may contribute to greater regional as well as local diversity (measured as a species number and Fisher's α) in the Iquitos-Pebas region (Tables 1 and 2). Palm diversity could peak at the poorer end of the nutrient gradient (cf. diversity patterns of Melastomataceae; Tuomisto et al. 2002), or greater soil heterogeneity could provide more habitat types, which in turn may increase local diversity through mass effects (Shmida & Wilson 1985). Strong mass effects have been reported from palm communities on the Amazonian slopes of the Andes (Kessler 2000). Pitman et al. (2001) found that Yasuni had higher local and regional tree diversity than the seasonally dry Manu region in south-eastern Peru; probably that diversity decreases with increasing seasonality in tropical forests (cf. Gentry 1988; Pyke et al. 2001).

Pitman et al. (2001) and Terborgh et al. (2002) suggested that the floristic similarity between the Yasuni and Manu regions occurs because both regions lie in the Andean foreland, and that the Iquitos region would be different, because it is much further from the Andes, and contains sediments that are different from those of the Yasuni and Manu regions. However, the floristic relationships of pteridophytes and Melastomataceae across western Amazonia do not show this pattern (Tuomisto, Ruokolainen & Yli-Halla 2003).

Palm diversity was relatively high in both Yasuni (Fisher's α 3.45–5.52) and Iquitos-Pebas (3.85–6.57) in comparison with other sites in western South America. We calculated Fisher's α using data from five study sites in tropical moist forests in western Ecuador (Borchsenius 1997), one site in the Chocó area in Colombia (Galeano et al. 1998) and from two Peruvian sites in the lower Ucayali River valley (Kahn & Mejia 1991) (1.02–3.13, 2.23, and 3.87–5.14, respectively).

do the same palm species dominate everywhere?

Palm communities in both regions were dominated by a limited set of ubiquitous species. Five or fewer species accounted for over 60% of individuals in a transect. Pitman et al. (2001) found that 150 common tree species dominated both in Yasuni and Manu. Species common to the two regions usually formed > 90% of the individuals in a transect in Yasuni, but a smaller (on average 63%) and more variable (25–91%) proportion in Iquitos-Pebas. Pitman et al. (2001) found similar patterns, but with less pronounced dominance differences between Yasuni & Manu. Pitman et al. (2001) found that for the 254 tree species shared between regions, abundances in the two regions were correlated (Spearman rs = 0.40, P = 0.0001) and concluded that a small number of common species (> 1 individual ha−1) dominate both communities at all scales. The 26 palm species common to Yasuni and Iquitos-Pebas were similarly correlated (Spearman rs = 0.39, P = 0.05).

Some species restricted to a single region were important quantitatively and sometimes dominant there, and included the most abundant species (Lepidocaryum tenue) in Iquitos-Pebas. Hence, the correlation in species abundance between Yasuni and Iquitos-Pebas disappeared when those species restricted to a single region were not excluded. Kahn & Mejia (1991) found Lepidocaryum tenue to be dominant at one site and common at another site (dominated by Iriartea deltoidea and Hyospathe elegans), 150 km south-west of Iquitos.

Pitman et al. (2001) suggested that the oligarchic dominants could be ecological generalists, in contrast to the many rarer, ecologically specialized species (cf. Poore 1968 for dipterocarp forest). Five out of the six species occurring in all transects were tall species, which are known to have wider geographical ranges and possibly more generalized ecological demands (Ruokolainen & Vormisto 2000; Ruokolainen et al. 2002). Furthermore, the distribution of woody understorey species in the Colombian portion of the Amazon basin was found to correlate better with soil conditions than the distribution of canopy species (Duque et al. 2002). Alternatively large size could be favoured by reduced seed source limitation (cf. Clark et al. 1998; Nathan & Muller-Landau 2000), with understorey palms (the majority of the species) having lower fecundity and therefore increased recruitment limitation due to heavy shade (cf. Chazdon 1986; Cunningham 1997; Svenning 2002).

diversity and dominance patterns of palms in the western amazon basin

Our results strongly suggest that the three different views that have dominated recent discussions on Amazonian plant community structure must be considered complementary rather than mutually exclusive. Both limited dispersal and local environment seem to control floristic patterns, and palm communities were dominated by limited species oligarchies. However, we also found that regions of the western Amazon basin as little as 600 km apart, with similar climate, exhibit clear differences in diversity and dominance. Our results differ only from previous western Amazonian studies in the relative importance of different factors (Duivenvoorden 1995; Tuomisto et al. 1995; Pitman et al. 1999, 2001; Condit et al. 2002; Duivenvoorden et al. 2002; Phillips et al. 2003; Tuomisto, Ruokolainen & Yli-Halla 2003; Tuomisto, Ruokolainen, Aquilar & Sarmiento 2003). Indeed, most present evidence points to strong spatial structuring of the flora, and knowledge of its scale-dependency and geographical consistency and underlying causes requires more extensive sampling.

Acknowledgements

We thank Guillermo Criollo, Victor Vargas and the villagers in Puerto Izango, Santa Ana and Tarapoto for practical help, Hanna Tuomisto and Kalle Ruokolainen for discussions, and Flemming Nørgaard for making the map. We acknowledge funding from the European Union (Marie Curie Fellowship to JV), the Ministry of Education, Finland (grant to JV), the Mellon Foundation (grant to HB & PH), and the Danish Natural Science Research Council (grants #21-01-0415 to JCS and #21-01-0617 to HB).

Ancillary