BIOTROPICA 38(4): 468–478 2006 10.1111/j.1744-7429.2006.00168.

Edaphic and Floristic Variation within a 1-ha Plot of Lowland Amazonian Rain Forest1

Axel Dalberg Poulsen2

Royal Botanic Garden, 20A Inverleith Row, Edinburgh EH35LR, Scotland

Hanna Tuomisto
Department of Biology, University of Turku, FI-20014 Turku, Finland

and

Henrik Balslev
Institute of Biological Science, University of Aarhus, Universitetsparken, DK-8000 Aarhus C, Denmark

ABSTRACT
Several studies in lowland tropical rain forests have documented effects of local-scale topographic variation on plant species distribution and abundance patterns.
Few studies have compared the distribution patterns of more than one plant group, however, and even fewer have related these to measured physical and chemical
soil characteristics. Here, we document such soil characteristics within a square 1-ha plot in Amazonian Ecuador, and compare them to the distribution patterns of
terrestrial pteridophytes, angiosperm ground herbs, and palms. Substantial variation in soil properties was found within the plot. The three plant groups showed highly
correlated floristic patterns within the 1-ha plot even after the effect of geographical distances had been taken into account. Mantel tests yielded significant correlations
between edaphic patterns, as measured by distances in various soil and topographic characteristics, and floristic patterns. For all three plant groups, differences in
elevation within the plot were highly correlated with floristic distances, and for terrestrial pteridophytes and palms, distances in soil calcium content and sand content
were also important. Our results resembled those obtained at wider spatial scales with the same plant groups, which indicates that soil factors may play an important
role for distribution and beta diversity of plants, even at the local scale.

RESUMEN
Varios estudios en bosques tropicales lluviosos han documentado los efectos de la variación local de la topografı́a sobre los patrones de distribución y abundancia de
especies de plantas. Sin embargo, pocos estudios han comparado los patrones de distribución en más de un grupo de plantas, y aún más escasas son las investigaciones
que relacionan éstas a medidas fı́sicas y quı́micas de las caracterı́sticas de los suelos. En el presente estudio, documentamos estas caracterı́sticas de suelos en una parcela
cuadrada de 1 ha en la Amazonia ecuatoriana, y las comparamos a los patrones de distribución de pteridofitas terrestres, hierbas angiospermas terrestres y palmeras. Se
encontró una gran variación en las propiedades de los suelos dentro de la parcela. Los tres grupos de plantas presentaron patrones florı́sticos altamente correlacionados
dentro de la parcela aún cuando se tomó en cuenta el efecto de distancias geográficas. Pruebas de Mantel resultaron en correlaciones altamente significativas entre
patrones edáficos, medidos como distancias en varias caracterı́sticas de suelos y de topografı́a, y patrones florı́sticos. Para los tres grupos de plantas, diferencias en
elevación dentro de la parcela fueron altamente correlacionadas con distancias florı́sticas, y para pteridofitas terrestres y palmeras, distancias en el contenido de calcio
y de arena en el suelo fueron importantes también. Nuestros resultados son parecidos a aquellos obtenidos en escalas espaciales mas amplias para los mismos grupos
de plantas, indicando que factores edáficos pueden ser importantes para la distribución y la beta diversidad de plantas aun en una escala local.

Key words: Ecuador; ground herbs; local scale variability; Mantel test; multiple regression; palms; soils; species composition; terrestrial pteridophytes, tropical rain
forest.

STUDIES CONDUCTED IN DIFFERENT PARTS OF THE TROPICS et al. 1997; Ruokolainen & Tuomisto 1998; Tuomisto et al. 2002,
at different spatial scales differ in how much importance they place 2003a, 2003b; Vormisto et al. 2004a, 2004b). At the mesoscale
on environmental factors, especially soil properties, for plant species (1–100 km2 ) in the Neotropics, two studies on trees and palms (ter
distributions. At the landscape scale (100–10,000 km2 ), edaphic Steege et al. 1993; Clark et al. 1995, 1998, 1999) and one study on
gradients have been found to be important for the distribution of herbs (Costa et al. 2005) found that distributions were related to
trees in several studies (Baillie et al. 1987, Duivenvoorden 1995, soil and topography.
Ruokolainen et al. 1997, Ruokolainen & Tuomisto 1998, Pyke At the local scale (<1 km2 ), there seems to be general agreement
et al. 2001, Duque et al. 2002, Phillips et al. 2003), but oth- that topographic factors are important for the distributions of at
ers have reported that most tree species occur in more than one least some tree species (Lieberman et al. 1985, Hubbell & Foster
forest type (Pitman et al. 1999). Studies on palms, Melastomat- 1986, Sabatier et al. 1997, Webb & Peart 2000, Harms et al. 2001,
aceae, and pteridophytes have invariably found that landscape-scale Debski et al. 2002, Valencia et al. 2004). Such topographically
edaphic differences are related to floristic patterns (Ruokolainen determined patterns are also common in palms (Svenning 1999,
2001; Vormisto et al. 2004a) and herbs (Poulsen 1996; Poulsen &
1 Received 14 April 2005; revision accepted 8 September 2005. Balslev 1991; Poulsen & Nielsen 1995; Tuomisto et al. 1995, 1998;
2 Corresponding author; e-mail: axel@dalbergpoulsen.com Tuomisto & Poulsen 2000).

468

C 2006 The Author(s)

Journal compilation

C 2006 by The Association for Tropical Biology and Conservation
 Edaphic and Floristic Variation 469

The effects of obvious local soil differences, such as those be- In vegetatively spreading species, such as Geophila and Selaginella,
tween white sand and more fine-textured soils, are generally rec- each rooting stem was counted as one individual. For palms, each
ognized but are rarely quantified (Tuomisto & Ruokolainen 1994, ramet of multistemmed species was counted as one individual.
Vormisto et al. 2000), and documentation of less conspicuous local Several voucher specimens were collected of each species
soil variation is sparse. The study by Johnston (1992) found that and deposited in herbaria in Aarhus (AAU), Quito (QCA), and
soil calcium concentration and pH were correlated with tree species Turku (TUR, pteridophytes only; herbarium acronyms according
composition within one hectare of Tabanuco forest in Puerto Rico. to Holmgren et al. 1990).
This suggests that local floristic patterns may not only be related To document soil conditions within the 1-ha plot, 20 soil
to soil drainage, which topography is usually thought to represent, samples were obtained in 1996. One sample was taken in each of
but also to other soil properties. One study is not sufficient to give the ten subplots along the northern edge of the plot, and another
a general idea of the degree to which soil physical and chemical ten soil samples were taken in a regular manner from the rest of the
properties may vary at the local scale, however. plot (see Fig. 1). Each sample consisted of five pooled subsamples
The present study assesses the variability in soil properties and taken from the top of the mineral soil (0–5 cm), obtained within
floristic composition in a 1-ha study plot in Cuyabeno, Amazonian an area of approximately 5 × 5 m. The samples were air-dried in
Ecuador. Apart from the 0.1-ha plot in Costa Rica studied by Whit- the field and stored in plastic bags until they were analyzed in the
more et al. (1985), the Cuyabeno plot is probably the only tropical laboratory of the Agricultural Research Centre of Finland. Physical
forest plot where an attempt has been made to document the species and chemical analyses were carried out using standard procedures
richness of all plants (Balslev et al. 1998). The tree enumeration in (van Reeuwijk 1993) and included: exchangeable cations (calcium,
the plot set the new World record in single-hectare tree species potassium, magnesium, and sodium in 1 M ammonium acetate at
richness (Valencia et al. 1994), and the total count of more than pH 7), aluminum (in 1 M KCl), soil pH (in 1 M KCl), loss on
900 plant species (Balslev et al. 1998) is unparalleled. These num- ignition at 420◦ C (LOI; a proxy for the content of organic matter),
bers are generally thought of as measures of within-habitat species phosphorus (Bray method), and particle size (% of clay, silt, and
richness (alpha diversity). sand). Most of these factors are known to affect plant growth due
The species count included 96 ground herbs. Many of them to their role as major plant nutrients (Ca, K, Mg, and P), potential
had distribution patterns apparently related to topography, and toxicity (Al), other physiological effects (pH), or as determinants
we suggested that edaphic variation within the plot significantly of soil drainage, water retention capacity, and nutrient fluxes (LOI,
increased the number of species present, although no soil data were texture).
available to test this suggestion (Poulsen & Balslev 1991). We have
since obtained soil samples from the plot to study how much the NUMERICAL ANALYSES.—Most numerical analyses were based on
soils vary in chemical and physical properties within the plot. Here, resemblance matrices. Both the Sørensen index (which is binary,
we present the soil analysis results and relate the edaphic patterns to and hence only takes into account species presence and absence)
the floristic patterns of terrestrial pteridophytes, angiosperm ground and the Steinhaus index (which is mathematically similar, but also
herbs, and palms. takes into account species abundance) were used to measure floristic
similarity between the 0.01-ha subplots. Separate similarity matrices
METHODS were computed for pteridophytes, herbs, and palms, and additional
matrices were computed with all plant groups combined. Both
FIELDWORK.—The 1-ha study plot was situated in Cuyabeno indices are sensitive to sample size and tend to lower values when
(00◦ 00 S, 76◦ 12 W) in lowland (250–300 m elevation), nonin- many species are rare and sample sizes are small (Chao et al. 2005),
undated (terra firme) rain forest in Amazonian Ecuador. The plot but we did not attempt to correct for this error here, as we are only
was 100 × 100 m; divided into 100 square, permanently marked interested in the correlations among the similarities, not in their
subunits, and inventoried for all plants in 1988–1989. The to- absolute values.
pography was measured using a Suunto clinometer taking readings Two sets of similarity matrices were produced. One set con-
between the corners of all subunits. tained all 100 subplots, and the other set contained just the 20 sub-
Three of the plant groups inventoried in the Cuyabeno plot are plots from which soil samples were available. Since soil sampling
considered here: pteridophytes growing on the ground (ferns and was done on a regular grid, some soil samples coincided with the
fern allies; henceforth called pteridophytes), angiosperm ground limits between the original 0.01-ha subplots. For pteridophytes and
herbs (henceforth called herbs), and palms (members of the family herbs, virtual 0.01-ha subplots centered on these soil samples were
Arecaceae). The pteridophyte and herb inventories also included created with the help of the coordinates of each individual plant.
individuals of climbing and epiphytic species, which normally do For palms, individual coordinates were not available; so one of the
not complete their life cycles on the forest floor. neighboring original subplots was used instead. When similarities
All individuals belonging to the target plant groups that were needed to be converted to distances, the formula Distance = 1 −
rooted in the ground, including all size classes, were recorded for Similarity was used.
each subplot. For pteridophyte and herb individuals, but not for Matrices of Euclidean distances based on the measured
palms, coordinates were assigned within the subplot, which allowed soil variables were computed using each soil variable separately.
arbitrary reassignment of subplot limits at the data analysis stage. Concentrations of elements (Ca, K, Mg, Na, Al, and P) were
470 Poulsen, Tuomisto, and Balslev

FIGURE 1. The 1-ha plot in Cuyabeno, Amazonian Ecuador, with 1-m elevational contours. The topographically lowest part of the plot is shaded grey; the lowest
point is in the upper left (northwestern) corner; the highest in the upper right corner (+ c. 19 m). The black circles in the top two rows show both the locations of
the 20 soil samples and the measured values of eight soil variables in them (value being proportional to circle diameter). The black circles in the lowest row indicate
the ordination scores of the 100 subunits along the first axis of a floristic ordination (PCoA).

log-transformed before the distances were computed. Additional Mantel tests were performed to test how much of the correlation
matrices were based on log-transformed geographical distance, dif- between two distance matrices (e.g., floristic and environmental)
ference in elevation, and difference in slope between subunits. The remained after the correlation with a third distance matrix (e.g., ge-
log-transformed geographical distances simulate the effect of ran- ographical) had been taken into account. The statistical significance
dom dispersal (Hubbell 2001), and elevation is usually strongly of each correlation was determined by a Monte Carlo permutation
correlated with drainage. To obtain information of the floristic gra- test using 999 permutations, which allows testing of the statistical
dients within the plot, Principal Coordinates Analysis (PCoA) was significance at the P < 0.001 level.
performed for each plant group. A multiple regression was performed on distance matrices to
Correlations between pairs of distance matrices were computed quantify how much of the variation in each of the floristic distance
using the Mantel test. The standardized form of the Mantel statistic matrices could be explained with a combination of those edaphic
was used, which is mathematically similar to the Pearson correlation and spatial distance matrices that had yielded significant Mantel
coefficient, but is computed between the cell values of two resem- correlations with the floristic distance matrix in question. Backward
blance matrices rather than the measured values of the original vari- elimination was applied to retain, in the final model, only those
ables. Because spatial autocorrelation can be expected to be present variables that had a statistically significant (P < 0.1 after Bonferroni
within a 1-ha plot, and was indeed found in the present case, partial correction) contribution to the proportion of variation explained.
 Edaphic and Floristic Variation 471

and calcium, together constituted between 52 and 84 percent of the

TABLE 1. Soil characteristics of the square, 1-ha plot at Cuyabeno, Amazonian
sum of cations (mean 68%). The highest cation concentrations were
Ecuador (N = 20).
found in the lowest, northwestern corner of the plot (Fig. 1), and
Mean Range SD cation content was negatively correlated with elevation (Table 2).
Soil aluminum content ranged between 6.3 cmol(+)/kg and
Ca (cmol(+)/kg) 0.25 0.10–1.02 0.22 12.8 cmol(+)/kg, and was 3–21 times higher than the content of
K (cmol(+)/kg) 0.18 0.13–0.22 0.02 cations. The pattern of aluminum content in the plot was opposite
Mg (cmol(+)/kg) 0.30 0.17–0.59 0.14 to that of the sum of cations, with the lowest values in the lowest
Na (cmol(+)/kg) 0.06 0.01–0.23 0.06 part of the plot (Fig. 1; Table 2).
Sum of Ca,K,Mg,Na 0.78 0.47–2.06 0.40 The soils were acidic, with pH ranging between 2.99 and 3.68.
Al (cmol(+)/kg) 9.58 6.33–12.80 1.60 The most acidic soils were found on the upper slopes (Fig. 1), but
pH 3.33 2.99–3.68 0.18 this was not a very clear pattern (Table 2). LOI varied from 6 to
LOI (%) 10.06 6.56–14.17 2.11 14 percent, and phosphorus contents ranged from 2 to 11 mg/kg;
P (mg/kg) 6.37 2.03–11.04 2.03 neither showed any clear relationship with topography.
clay (<0.002 mm, %) 31 20–41 5.75 The soils were generally fine-grained, with a mean sand content
silt (0.002–0.063 mm, %) 61 53–68 4.85 of only 8 percent. The most sandy soils were found in the lowest part
sand (0.063–2 mm, %) 8 5–15 2.58 of the plot (Fig. 1; Table 2). The silt fraction was very prominent:
it ranged from more than half to two-thirds of soil weight in all
samples, and averaged 61 percent. The clay fraction varied between
The multiple regressions were conducted using Permute!, and 20 and 40 percent (Table 1). The exchangeable cation content was
all other analyses using Le Progiciel R. Both programs are available positively correlated with sand content (Table 2), and so were the
on the Internet (http://www.bio.umontreal.ca/casgrain/fr/labo/). contents of the individual cations, with the exception of potassium.
Detailed explanations of all the methods can be found in Legendre In general, those soil variables that yielded high Pearson cor-
and Legendre (1998). relations also yielded high Mantel correlations, but the latter were
almost always lower than the former. Mantel correlations are pro-
vided in Table 2, mainly to enable comparisons with the Mantel
RESULTS correlations involving floristic distance matrices (see below).
Some soil variables (sum of cations, calcium content, pH) as
SOILS.—The concentration of exchangeable cations (sum of Ca, K, well as elevation were correlated with position along the East–West
Mg, and Na) in the soil samples from the Cuyabeno plot varied from axis of the plot, but only one (aluminum content) was correlated
0.47 cmol(+)/kg to just over 2 cmol(+)/kg (Table 1). Magnesium with position along the North–South axis. Differences in aluminum

TABLE 2. Pearson correlations between original variables (lower triangle) and Mantel correlations between distance matrices (upper triangle) for different soil variables.
Statistically significant ( P < 0.05) Pearson correlation coefficients are shown in bold. Mantel correlation coefficients larger than 0.2 are shown in bold, but their
statistical significance has not been tested. x and y are the East–West and North–South coordinates of the sample units, respectively.

Cat sum Ca K Mg Na Al pH LOI P Clay Silt Sand Elev Slope Geo distance

Cat sum 0.84 0.24 0.85 0.57 0.64 0.26 0.01 0.17 0.38 −0.01 0.67 0.39 −0.02 0.17
Ca 0.94 0.12 0.55 0.45 0.72 0.29 0.07 0.02 0.25 −0.10 0.65 0.42 −0.08 0.18
K 0.41 0.25 0.27 0.03 0.08 0.06 0.30 0.51 0.07 0.01 0.03 −0.05 −0.04 −0.04
Mg 0.90 0.73 0.50 0.45 0.46 0.26 0.05 0.17 0.38 0.07 0.43 0.26 0.07 0.15
Na 0.79 0.67 0.23 0.69 0.45 0.12 −0.05 0.07 0.26 0.06 0.69 0.22 −0.06 −0.05
Al −0.71 −0.75 0.13 −0.63 −0.45 0.41 0.33 −0.01 0.19 −0.11 0.58 0.29 0.13 0.24
pH 0.41 0.49 −0.34 0.28 0.33 −0.65 0.42 −0.10 0.03 −0.08 0.27 0.14 0.12 0.05
LOI −0.18 −0.27 0.64 −0.10 −0.16 0.67 −0.59 0.01 −0.09 −0.12 0.01 0.02 0.19 0.06
P 0.46 0.41 0.59 0.35 0.40 −0.13 −0.19 0.35 0.13 0.16 0.00 −0.08 −0.10 0.07
Clay −0.58 −0.49 −0.15 −0.59 −0.51 0.37 −0.19 0.08 −0.09 0.73 0.29 0.19 0.00 0.13
Silt 0.26 0.19 0.13 0.34 0.16 −0.10 −0.07 0.03 −0.08 −0.90 −0.04 −0.02 0.05 0.12
Sand 0.80 0.75 0.10 0.67 0.84 −0.63 0.57 −0.22 0.35 −0.55 0.12 0.36 −0.18 −0.01
Elev −0.62 −0.61 −0.15 −0.59 −0.37 0.56 −0.40 0.33 −0.07 0.52 −0.36 −0.48 0.01 0.34
Slope 0.25 0.18 −0.07 0.42 0.02 −0.50 0.08 −0.33 −0.14 −0.36 0.36 0.13 −0.36 0.05
x −0.50 −0.50 −0.14 −0.44 −0.32 0.42 −0.48 0.35 0.06 0.03 0.15 −0.35 0.73 −0.11
y 0.41 0.40 0.08 0.41 0.23 −0.58 0.27 −0.19 0.28 0.10 −0.28 0.30 0.19 0.24
472 Poulsen, Tuomisto, and Balslev

content and elevation yielded the highest Mantel correlations with

TABLE 3. Mantel correlations between distance matrices calculated separately for
geographical distance, which indicates that they were highly spatially
three different plant groups using 100 subunits of the 1-ha Cuyabeno
autocorrelated (Table 2).
plot, Amazonian Ecuador. The correlations shown in parentheses were
obtained from a partial Mantel test and show the correlations between
TOPOGRAPHICAL PATTERNS IN SPECIES RICHNESS, ABUNDANCE, AND
the floristic distance matrices after the effect of log-transformed geo-
FLORISTIC COMPOSITION.—A total of 123 species and 17,020 in-
graphical distances has been taken into account. Significance levels are
dividuals of pteridophytes, herbs, and palms were recorded in the
based on 999 permutations ( ∗∗∗ P < 0.001, ∗∗ P < 0.01, ∗ P < 0.05).
Cuyabeno 1-ha plot. Pteridophytes were represented by 29 species
(4668 individuals), herbs by 70 species (6607 individuals) and
palms by 24 species (5745 individuals); nine of the palm species Pteridophytes Herbs
can grow into the canopy. The complete species list is presented in
the Appendix. Sørensen Index
The following results are based on 0.01-ha (10 × 10 m) sub- Palms 0.31∗∗∗ (0.28∗∗∗ ) 0.14∗∗ (0.10∗ )
plots of the 1-ha plot. Palm density was highest in the highest Pteridophytes 0.25∗∗∗ (0.18∗∗ )
part of the plot, and showed a significant correlation with elevation Steinhaus Index
(Pearson’s r = 0.46, P < 0.001). In contrast, the densities of pterido- Palms 0.24∗∗∗ (0.19∗∗∗ ) 0.29∗∗∗ (0.24∗∗ )
phytes and herbs were not correlated with elevation. The densities Pteridophytes 0.26∗∗∗ (0.12∗∗ )
of herbs and palms were lowest on the steep slopes (Pearson’s r =
−0.20, P = 0.045 between herb density and slope; r = −0.32,
P = 0.001 between palm density and slope), whereas the density
of pteridophytes was not correlated with slope. Species richness per dance data (Steinhaus index) was used. Difference in herb species
subplot was highest in the lowest part of the plot for pteridophytes composition was also correlated with difference in slope (Mantel’s
and herbs (Pearson’s r = −0.22 between species richness and ele- r = 0.08, P = 0.027 with Sørensen index; Mantel’s r = 0.19, P =
vation, P = 0.027), whereas the opposite was true for palm species 0.001 with Steinhaus index). These results are in agreement with
richness (Pearson’s r = 0.30, P = 0.002). Pteridophytes showed the results of the ordination analyses (Fig. 1).
significantly higher species richness on the steeper slopes (Pearson’s Mantel tests involving soil data had to be limited to the
r = 0.31, P = 0.002), whereas the species richness of herbs and 20 subplots for which soil samples had been analyzed. The Mantel
palms was not correlated with slope. correlations between floristic distances and distances based on the
The first axes of all floristic ordinations show patterns that various soil characteristics varied between 0 and 0.75 (Table 4).
could be related to topography (Fig. 1). When all three plant groups There were clear differences in these results both among the plant
were combined for the ordination analysis, the main floristic gradi- groups and among the soil variables.
ent was clearly that from the top of the hill toward the lower parts For pteridophytes and palms, significant Mantel correlations
of the plot, with the central ridge having intermediate values. An were found with several soil factors. For herbs, on the contrary, only
ordination based on pteridophytes only showed a clear dichotomy soil Al content yielded a significant Mantel correlation, and even
between the hilltop and the lower parts, with the lowest ordination that only when herb abundance data were used (Steinhaus index).
scores in the highest parts of the plot. The ordinations on herbs In the case of pteridophytes, the Sørensen index gave generally
only and on palms only also reflected the elevation gradient within higher Mantel correlations with the soil characteristics than did the
the plot. In the herb ordination (and to a lesser degree in the palm Steinhaus index. With palms, the opposite was true.
ordination), there also seemed to be a floristic gradient related to the The soil properties that most often yielded significant Mantel
steepness of slopes, with the lowest ordination scores in the flattest correlations with floristic distances were the contents of exchange-
parts of the plot. able cations, Ca, Na, Al, and sand. Five of the soil variables (K, P
and silt content, pH, and LOI) did not produce significant Mantel
CORRELATIONS BETWEEN FLORISTIC AND EDAPHIC RESEMBLANCE correlations with any plant group.
MATRICES.—The floristic patterns of the three plant groups, as mea- As with the full set of subplots, difference in elevation was
sured by floristic distance matrices comparing the 100 subplots, were highly correlated with difference in floristic composition of all
significantly correlated (Mantel’s r = 0.14–0.31; Table 3). When plant groups. The Mantel correlation coefficients involving ele-
the effect of geographical distances between subplots was taken into vation difference were usually clearly higher than any of the Mantel
account by running a partial Mantel test, the correlations between correlations involving difference in soil physical or chemical prop-
plant groups decreased slightly (Mantel’s r = 0.10–0.28), but re- erties. The exception to this tendency was the series of Mantel tests
mained statistically significant. that involved floristic distances computed on palm abundance data
When floristic patterns were compared with topographic pat- (Table 4; Fig. 2). Differences in floristic composition were most
terns among the 100 subplots, difference in elevation was found correlated to differences in elevation in the case of pteridophytes,
to be positively correlated with difference in floristic composition and least correlated in the case of herbs.
with all three plant groups (Mantel’s r = 0.32–0.45, P < 0.001), Geographical distance was highly correlated with floristic dis-
regardless of whether presence–absence (Sørensen index) or abun- tance of pteridophytes and herbs whether presence–absence or
 Edaphic and Floristic Variation 473

TABLE 4. Mantel correlations between distance matrices based on floristic, soil and terrain characteristics using twenty 10 × 10 m subunits within a 1-ha plot at Cuyabeno,
Amazonian Ecuador. The floristic distance matrices were calculated separately for each plant group using two different indices (Sørensen index for presence–absence
data, Steinhaus index for abundance data). All distance matrices for soil and terrain characteristics were based on Euclidean distance. Significance levels are based
on 999 permutations ( ∗∗∗ P < 0.001, ∗∗ P < 0.01, ∗ P < 0.05). The proportion of statistically significant correlations expected at random at the P < 0.05 level
is one out of 20 tests, clearly lower than found here.

Pteridophytes Herbs Palms

Sørensen Steinhaus Sørensen Steinhaus Sørensen Steinhaus

Cation sum 0.42∗ 0.27∗ 0.10 0.16 0.35∗ 0.66∗∗

Ca 0.47∗ 0.35∗∗ 0.11 0.17 0.39∗ 0.75∗∗∗
K −0.04 0.02 −0.15 −0.12 −0.05 0.08
Mg 0.20 0.12 0.05 0.07 0.22 0.44∗∗
Na 0.35∗ 0.17 0.00 −0.04 0.25∗ 0.36∗
Al 0.30∗ 0.22∗ 0.20 0.25∗ 0.28∗ 0.51∗∗
pH 0.05 0.05 −0.02 −0.04 0.06 0.17
LOI −0.10 0.00 −0.04 −0.01 0.06 0.01
P −0.04 −0.03 −0.20 −0.09 0.02 0.02
Clay 0.21 0.20∗ 0.00 0.11 0.07 0.24∗
Silt −0.01 −0.01 −0.12 −0.02 −0.16 −0.06
Sand 0.40∗ 0.21∗ 0.06 0.05 0.37∗ 0.59∗∗
Elevation 0.73∗∗∗ 0.63∗∗∗ 0.43∗∗∗ 0.49∗∗∗ 0.55∗∗∗ 0.51∗∗∗
Slope 0.00 0.00 0.15 0.19∗ −0.15 −0.10
Geo distance 0.20∗ 0.43∗∗∗ 0.39∗∗∗ 0.56∗∗∗ 0.21∗ 0.19
Partial Mantel correlations between floristic distances and those shown at left after taking the effect of geographical distances into account
Cation Sum 0.40∗ 0.22∗ 0.04 0.08 0.33∗ 0.65∗∗
Al 0.27 0.13 0.12 0.14 0.24 0.48∗∗
Sand 0.41∗ 0.24∗ 0.07 0.07 0.38∗ 0.60∗∗
Elevation 0.71∗∗∗ 0.57∗∗∗ 0.34∗∗∗ 0.38∗∗∗ 0.52∗∗∗ 0.48∗∗∗
Partial Mantel correlations between floristic and geographical distances after taking the effect of the distances shown at left into account
Cation Sum 0.14 0.41∗∗∗ 0.38∗∗∗ 0.55∗∗∗ 0.17 0.11
Al 0.14 0.40∗∗∗ 0.36∗∗∗ 0.53∗∗∗ 0.16 0.08
Sand 0.22∗ 0.44∗∗∗ 0.39∗∗∗ 0.56∗∗∗ 0.23∗ 0.24∗
Elevation −0.08 0.29∗∗ 0.29∗∗ 0.48∗∗∗ 0.03 0.02

abundance data were used (Table 4; Fig. 2). In the case of palms, relation between floristic and geographical distances. Only ele-
a significant Mantel correlation was only obtained with presence– vation was found to have a clear effect on all the correlations,
absence data. The geographical distances yielded, in most cases, although also cation content and Al content caused the correla-
lower Mantel correlation coefficients than differences in elevation. tions involving pteridophyte or palm abundance data to become
For palm and pteridophyte presence–absence data, geographical dis- nonsignificant.
tances also yielded lower correlations than some of the soil variables. All multiple regression analyses retained elevation in the final
Partial Mantel tests were run to remove the effect of geographi- model. Three of them also retained geographical distance, and two
cal distance in order to determine whether the correlations between retained one of the soil cations (Table 5).
floristic differences and differences in the most important soil and
topographic variables could be explained by the effects of spatial
proximity (Table 4). The Mantel correlations involving the con- DISCUSSION
tent of exchangeable cations or sand content were affected only
marginally, and the correlations involving elevation mainly in the SOIL VARIATION.—The soil characteristics of the Cuyabeno 1-ha
case of herbs. Most correlations involving Al content, however, be- plot can be compared to soil characteristics documented in three
came nonsignificant when the effect of geographical distance was other studies in western Amazonia, one in Yasuni national park (60–
partialled out. 90 km to the south of Cuyabeno; Tuomisto et al. 2003a), one in the
Partial Mantel tests were also run to remove the effect of district of Loreto in northern Peru (500–600 km to the southeast;
the most important soil and topographic variables from the cor- Ruokolainen & Tuomisto 1998), and one in the district of Madre
474 Poulsen, Tuomisto, and Balslev

1
Multiple regressions on distance matrices using twenty 10 × 10 m
Pteridophytes (Sørensen)

TABLE 5.
subunits within a 1-ha plot at Cuyabeno, Amazonian Ecuador. The
distance matrices are as in Table 4. Initially, the model for each inde-
pendent matrix included all variables that show a significant simple
Mantel test result in Table 4, and those retained in backward elimi-
nation are included in the final model and shown here. Significance
levels are based on 999 permutations.
0
1
Independent matrices in
Pteridophytes (Steinhaus)

Dependent matrix final model R2 (%) P

Pteridophytes (Sørensen) Elevation, Na 57 <0.001

Pteridophytes (Steinhaus) Elevation, Geo distance 45 <0.001
Herbs (Sørensen) Elevation, Geo distance 25 <0.001
Herbs (Steinhaus) Elevation, Slope, Geo distance 44 <0.001
Palms (Sørensen) Elevation 30 <0.001
0
1 Palms (Steinhaus) Elevation, Ca 61 0.002
Herbs (Sørensen)

de Dios in southern Peru (1500 km to the south–southeast; Phillips

et al. 2003).
The soil cation content in Cuyabeno was intermediate between
0 those observed in Loreto and those observed in Yasuni and Madre
1 de Dios. The mean soil cation content in the Cuyabeno plot was
0.78 cmol(+)/kg. In Madre de Dios, soil cation content in 88 plots
Herbs (Steinhaus)

averaged 3.65 cmol(+)/kg (Phillips et al. 2003). In the Yasuni area,

soil cation content of 27 transects averaged 9.7 cmol(+)/kg, and no
transect had a cation content of less than 1 cmol(+)/kg (Tuomisto
et al. 2003a). In contrast, out of 32 Loreto sites, two-thirds had a
cation content below that of Cuyabeno (Ruokolainen & Tuomisto
0 1998). The observed mean cation content in Cuyabeno conformed
1 to predictions based solely on the pteridophyte species composition
in the plot (Tuomisto & Poulsen 1996).
Palms (Sørensen)

The values of soil aluminum content found in the Cuyabeno

plot were very high as compared to those found in the Loreto study;
only three of the 32 study sites in Loreto had higher mean Al
content than the Cuyabeno plot (Ruokolainen & Tuomisto 1998).
This contrasts with the situation in Yasuni, where one-third of
0 the study plots had higher mean soil Al content than Cuyabeno
1 (Tuomisto et al. 2003a). The Al contents reported for Madre de
Dios are surprisingly low, averaging only 1 percent of the values
Palms (Steinhaus)

reported for the two other regions and Cuyabeno (Phillips et al.
2003).
Soil cation content within the Cuyabeno plot ranged between
0.47 and 2.06 cmol(+)/kg, i.e., by a factor of 4.4. In earlier studies,
variation of this magnitude has been found among sites whose
0 floristic differences suggest that they represent different forest types
0 4 8 12 16 10 1 10 1.4 10 1.8 10 2.2 (Ruokolainen & Tuomisto 1998, Tuomisto et al. 2003a). Some of
Elevation difference (m) Geographical distance (m)
this variation seems to be related to catena formation, as many of
the soil variables were correlated with elevation. This underlines the
FIGURE 2. The floristic similarity between the 20 subplots (in which soils importance of collecting soil samples from different topographical
were sampled) within the 1-ha plot at Cuyabeno measured by two indices positions when the purpose is to make comparisons among sites;
(Sørensen and Steinhaus) for three plant groups (pteridophytes, herbs and palms) soil samples from different topographical positions within the same
plotted against elevational differences and geographical distances. hectare may give quite different results.
 Edaphic and Floristic Variation 475

SPECIES RICHNESS AND LOCAL ENVIRONMENTAL GRADIENTS.—A pat- studies. All of the soil variables studied in the present paper, except
tern similar to the one found in Cuyabeno, with highest species sand and silt, yielded significant Mantel correlations with trees in
richness of pteridophytes and herbs toward the lower part of a the Madre de Dios study mentioned above (Phillips et al. 2003).
slope, has been found both in Amazonian transects (Tuomisto et al. It is interesting to note that, although conducted at different
1995, Tuomisto & Poulsen 2000) and in a 1-ha plot in a Bornean spatial scales, the studies involving pteridophytes yielded relatively
rain forest (Poulsen 1996). In both 1-ha plots, the lower part of similar results as to which soil properties were related to floristic
the slope had about four times as high soil cation concentrations patterns. While the present study was conducted within a plot of
as the higher part, but Poulsen (1996) speculated that water reten- 100 × 100 m, the other studies were conducted at the landscape
tion of the soil might be a more important factor explaining the scale with the largest interplot distances exceeding 25 km. Thus, the
species richness patterns. However, since many of the measured soil same edaphic factors that affect plant distributions at the landscape
properties are correlated with elevation, which in turn is correlated scale may also affect plant distribution at the local scale. Given
with soil moisture and drainage, the relative roles of these factors that each of these factors has well-known direct or indirect effects
as determinants of floristic or species richness patterns cannot be on plant growth, this result is not entirely unexpected, although
separated even in the statistical sense, and proving that there is a our study does not, of course, prove that similar results would be
causal relationship would require experimental studies and more obtained in other hectares of the forest.
study sites. Other factors not included in our study, such as forest Spatial autocorrelation can be expected to be important within
structure and gap formation (both of which influence light inten- a small, square sample plot, which would decrease the unique con-
sity), may also affect the observed species richness patterns. There tribution of soil variation to explaining floristic variation (Legendre
is some evidence that even these factors may be related to topogra- & Legendre 1998). Indeed, geographical distances yielded signifi-
phy (Denslow 1987, 1995; Gale 1997; Trichon 1997), which may cant Mantel correlations with all of the plant groups (except palms,
enhance its prominence in our results. when the Steinhaus index was used), and many of these correla-
Palms in Cuyabeno differed from pteridophytes and herbs in tions remained significant after the influence of edaphic distances
having the highest species richness on the higher end of the slope. was partialled out. However, most of the Mantel correlations be-
Several other studies (Kahn & De Castro 1985; Kahn 1987; Sven- tween floristic and edaphic distances that had been significant in
ning 1999; Vormisto et al. 2000, 2004a) have shown that topo- the simple Mantel test also remained significant after the effect of
graphic and edaphic gradients influence local density and species geographical distances was partialled out. This suggests that these
richness of palms. correlations were not just due to spatial proximity.
An interesting question is whether variation in elevation and Interestingly, in the partial Mantel tests involving both geo-
edaphic properties in the Cuyabeno plot also affect local tree species graphical distance and difference in elevation, the latter showed
richness, as the number of tree species in the plot is among the highly significant correlations with all plant groups, whereas geo-
highest ever recorded (Valencia et al. 1994). The results of Johnston graphical distance remained significant only with herbs and with
(1992) indicate that trees may react to soil variation at this spatial pteridophytes when the Steinhaus index was used. In other words,
scale, and our present results indicate that the Cuyabeno plot is not floristic differences within the 1-ha plot were more related to differ-
perceived as homogeneous by some other plant groups. However, ences in elevation (which correlates with drainage and catena-related
1 ha is a very small plot for analyzing distribution patterns of soil properties) than to geographical distance. This interpretation
canopy trees, for which these questions could better be answered is in agreement with the multiple regression results: all models re-
using larger but equally intensively studied plots, such as the 25-ha tained elevation, but only half of them also retained geographical
plot in Yasuni (Valencia et al. 2004). distance. These results suggest that even at the local scale, floristic
patterns may be more readily explained by environmental patterns
FLORISTIC VARIATION AND EDAPHIC GRADIENTS.—Floristic differ- than by spatially contagious random processes such as dispersal.
ences between subunits in the Cuyabeno 1-ha plot were found It is noteworthy that those soil properties that yielded signif-
to be most consistently correlated with differences in elevation, icant Mantel correlations with floristic patterns, also yielded high
which produced highly significant Mantel correlations with all plant correlations with elevation difference (compare Tables 2 and 5).
groups. Out of the actual soil variables, soil aluminum content, This underlines the difficulty in quantifying the relative effects of
soil cation content (especially Ca), and soil texture (especially sand alternative explanatory factors. However, two of the multiple re-
content) yielded significant Mantel correlations. The latter three gression models retained a soil cation distance matrix in addition to
were also important in both the Yasuni study and the Loreto study elevation difference, which indicates that in spite of their intercor-
mentioned above (Ruokolainen & Tuomisto 1998, Tuomisto et al. relation, both had a unique contribution to explaining the variation
2003a), although the significance of Al varied between plant groups in the floristic distance matrix.
in the Yasuni study; and Na was not a significant factor in either Whereas floristic patterns in all three plant groups were corre-
of the earlier studies whereas K was. The correlations of differences lated with differences in elevation, the herbs differed from the other
in LOI and soil phosphorus concentration with floristic differences two groups by not being correlated with differences in any of the
were significant in one of the two landscape-scale studies but not the soil characteristics. On the other hand, herbs were the only group
other; in the present study, both were nonsignificant. Differences in whose floristic patterns correlated with slope differences (when us-
soil pH produced only nonsignificant Mantel test results in all three ing the Steinhaus index). In Malaysia, certain ground herbs have
476 Poulsen, Tuomisto, and Balslev

been found to be restricted to slopes (Kiew 1978, 1988), and in a statistical approach for assessing similarity of species composition with
lowland rain forest in Singapore, small-seeded species of understory incidence and abundance data. Ecol. Let. 8: 148–159.
CLARK, D. A., D. B. CLARK, AND J. M. READ. 1998. Edaphic variation and the
woody plants favored steep microslopes (Metcalfe et al. 1998). Even
mesoscale distribution of tree species in a neotropical rainforest. J. Ecol.
though we did not measure seed size of the herbs in the Cuyabeno 86: 101–112.
plot, this factor may very well affect herb density and species com- ——–, ———, R. M. SANDOVAL, AND M. C. CASTRO VINICIO. 1995. Edaphic
position on slopes. and human effects on landscape-scale distributions of tropical rain forest
Positive correlations between floristic patterns of different plant palms. Ecology 76: 2581–2594.
groups have been obtained in several tropical studies (Tuomisto et al. CLARK, D. B., M. W. PALMER, AND D. A. CLARK. 1999. Edaphic factors and
the landscape-scale distributions of tropical rain forest trees. Ecology 80:
1995, 2003a, 2003b, 2003c; Ruokolainen et al. 1997; Ruokolainen 2662–2675.
& Tuomisto 1998; Vormisto et al. 2000; Poulsen et al. 2005; Duque COSTA, F. R. C., W. E. MAGNUSSON, AND R. C. LUIZA. 2005. Mesoscale distribu-
et al. 2005). All of these studies were done at the landscape or tion patterns of Amazonian understorey herbs in relation to topography,
regional scale, and therefore encountered much more environmental soil and watersheds. J. Ecol. 92: 863–878.
DEBSKI, I., D. F. R. P. BURSLEM, P. A. PALMIOTTO, J. V. LAFRANKIE, H. S.
variation than the present study. It is therefore not surprising that
LEE, AND N. MANOKARAN. 2002. Habitat preferences of Aporosa in two
the correlation coefficients between plant groups were higher than Malaysian forests: Implications for abundance and coexistence. Ecology
in the present study; however, it is highly interesting that similar 83: 2005–2018.
relationships between different plant groups can also be documented DENSLOW, J. S. 1987. Tropical rainforest gaps and tree species diversity. Ann.
at the local scale. Rev. Ecol. Syst. 18: 431–451.
———. 1995. Disturbance and diversity in tropical rain forests: The density
Plots of 1-ha have commonly been used to measure plant alpha
effect. Ecol. Appl. 5: 962–968.
diversity in tropical rain forests, especially that of trees; however, all DUIVENVOORDEN, J. F. 1995. Tree species composition and rain forest-
plant groups included in our study presented a significant degree environment relationships in the middle Caquetá area, Colombia, NW
of beta diversity within the 1-ha plot. Furthermore, much of the Amazonia. Vegetatio 120: 91–113.
variation in beta diversity could be explained by differences in el- DUQUE, A. J., J. F. DUIVENVOORDEN, J. CAVELIER, M. SÁNCHEZ, C. POLANÍA,
AND A. LEÓN. 2005. Ferns and Melastomataceae as indicators of vascular
evation and soil properties within the plot, which indicates it was
plant composition in rain forests of Colombian Amazonia. Plant Ecol.
not caused by random variation. Palms, which include large and 178: 1–13.
tree-like species, yielded the highest Mantel correlations with soil ———, M. SÁNCHEZ, J. CAVELIER, AND J. F. DUIVENVOORDEN. 2002. Different
patterns in our study. Whether other trees would show similar pat- floristic patterns of woody understorey and canopy plants in Colombian
terns cannot be estimated with the data available at present, but Amazonia. J. Trop. Ecol. 18: 499–525.
the results of Johnston (1992) suggest that they might. Therefore, GALE N. 1997. Modes of tree death in four tropical forests. Ph.D. Thesis, Aarhus
University, DK.
when estimates of alpha diversity are based on sample plots of 1 ha, HARMS, K. E., R. CONDIT, S. P. HUBBELL, AND R. B. FOSTER. 2001. Habitat
the possibility that some degree of beta diversity may be involved associations of trees and shrubs in a 50-ha Neotropical forest plot. J.
should be kept in mind when interpreting the results; different Ecol. 89: 947–959.
hectares most probably differ in their heterogeneity. HOLMGREN, P. K., N. H. HOLMGREN, AND L. C. BARNETT. 1990. Index herbari-
orum. Part I: The herbaria of the world. 8th edition. New York Botanical
Garden. pp. 693.
HUBBELL, S. P. 2001. A unified neutral theory of biodiversity and biogeography.
ACKNOWLEDGMENTS Princeton University Press, Princeton, New Jersey.
———, AND R. B. FOSTER. 1986. Commonness and rarity in a neotropical
We thank Gustavo Cañas for collecting the soil samples in the forest: Implications for tropical tree conservation. In M. E. Soulé (Ed.).
Cuyabeno plot, and Inger Juste at AAU and Jim Solomon at MO for Conservation Biology, Sinauer. pp. 205–231.
providing updates on the species identifications from their respective JOHNSTON, M. H. 1992. Soil-vegetation relationships in a Tabonuco forest
community in the Luquillo Mountains of Puerto Rico. J. Trop. Ecol. 8:
herbarium databases. We are grateful for the financial support from
253–263.
the Danish Natural Science Research Council (#21-01-0617 and KAHN, F. 1987. The distribution of palms as a function of local topography in
21-02-0552) and the Academy of Finland. Amazonian terra-firme forests. Experientia 43: 251–259.
———, AND A. DE CASTRO. 1985. The palm community in a forest of Central
Amazonia, Brazil. Biotropica 17: 210–216.
LITERATURE CITED KIEW, R. 1978. Floristic components of the ground flora of a tropical lowland
rainforest at Gunung Mulu National Park, Sarawak. Pertanika 1: 112–
BAILLIE, I. C., P. S. ASHTON, M. N. COURT, J. A. R. ANDERSON, E. A. FITZ- 119.
PATRICK, AND J. TINSLEY. 1987. Site characteristics and the distribution ———. 1988. Herbaceous flowering plants. In Earl of Cranbrook (Ed.).
of tree species in mixed dipterocarp forest on Tertiary sediments in cen- Malaysia, Pergamon Press, Oxford, pp. 56–76.
tral Sarawak, Malaysia. J. Trop. Ecol. 3: 201–220. LEGENDRE P., AND L. LEGENDRE. 1998. Numerical ecology. 2nd English edition.
BALSLEV, H., R. VALENCIA, C. G. PAZ Y MIÑO, H CHRISTENSEN, AND I. NIELSEN. Elsevier Science BV, Amsterdam. p. xv and p. 853.
1998. Species count of vascular plants in one hectare of humid low- LIEBERMAN, M., D. LIEBERMAN, G. S. HARTSHORN, AND R. PERALTA. 1985.
land forest in Amazonian Ecuador. In F. Dallmeier and J. A. Comiskey Small-scale altitudinal variation in lowland wet tropical forest vegetation.
(Eds.). Forest biodiversity in north, central and south America and the J. Ecol. 73: 505–516.
Caribbean: Research and monitoring. Man and the Biosphere Series. 21, METCALFE, D. J., P. J. GRUBB, I. M. TURNER. 1998. The ecology of very
pp. 585–594. Unesco and Parthenon, Paris. small-seeded shade-tolerant trees and shrubs in lowland rain forest in
CHAO, A., R. L. CHAZDON, R. K. COLWELL, AND T. -J. SHEN. 2005. A new Singapore. Plant Ecol. 134: 131–149.
 Edaphic and Floristic Variation 477

PHILLIPS, O. L., P. NÚÑEZ VARGAS, A. LORENZO MONTEAGUDO, A. PEÑA ———, AND ———. 2000. Pteridophyte diversity and species compostion in
CRUZ, M. E. CHUSPE ZANS, W. GALIANO SÁNCHEZ, M. YLI-HALLA, four Amazonian rain forests. J. Veg. Sci. 11: 383–396.
AND S. ROSE. 2003. Habitat association among Amazonian tree species: ———, ———, K. RUOKOLAINEN, R. MORAN, C. QUINTANA, J. CELI, AND
A landscape-scale approach. J Ecol. 91: 757–775. G. CAÑAS. 2003a. Linking floristic patterns with soil heterogeneity
PITMAN, N. C. A., J. TERBORGH, M. R. SILMAN, AND P. NUÑEZ V. 1999. Tree and satellite imagery in Ecuadorian Amazonia. Ecol. Appl. 13: 352–
species distributions in an upper Amazonian forest. Ecology 80: 2651– 371.
2661. ———, ———, AND R. C. MORAN. 1998. Edaphic distribution of some
POULSEN, A. D. 1996. Species richness and density of ground herbs within a plot species of the fern genus Adiantum in Western Amazonia. Biotropica
of lowland rainforest in north-west Borneo. J. Trop. Ecol. 12: 177–190. 30: 392–399.
———, AND H. BALSLEV. 1991. Abundance and cover of ground herbs in an ———, AND K. RUOKOLAINEN. 1994. Distribution of Pteridophyta and Melas-
Amazonian rain forest. J. Veg. Sci. 2: 315–322. tomataceae along an edaphic gradient in an Amazonian rain forest. J.
———, AND I. H. NIELSEN. 1995. How many ferns are there in one hectare of Veg. Sci. 5: 25–34.
tropical rain forest? Amr. Fern J. 85: 29–35. ———, ———, A. D. POULSEN, R. C. MORAN, C. QUINTANA, G. CAÑAS, AND
———, D. HAFASHIMANA, G. EILU, I. B. LIENGOLA, C. E. N. EWANGO, AND J. CELI. 2002. Distribution and diversity of pteridophytes and Melas-
T. B. HART. 2005. Composition and species richness of forest plants tomataceae along edaphic gradients in Yasuni National Park, Ecuadorian
along the Albertine Rift, Africa. Biologiske Skrifter 55: 129–143. Amazonia. Biotropica 34: 516–533.
PYKE, C. R., R. CONDIT, S. AGUILAR, AND S. LAO. 2001. Floristic composition ———, ———, M. AGUILAR, AND A. SARMIENTO. 2003b. Floristic patterns
across a climatic gradient in a neotropical lowland forest. J. Veg. Sci. 12: along a 43-km long transect in an Amazonian rain forest. J. Ecol. 91:
553–566. 743–756.
RUOKOLAINEN, K., A. LINNA, AND H. TUOMISTO. 1997. Use of Melastomataceae ———, ———, AND M. YLI-HALLA. 2003c. Dispersal, environment, and floris-
and pteridophytes for revealing phytogeographic patterns in Amazonian tic variation of western Amazonian forests. Science 299: 241–244.
rain forests. J. Trop. Ecol. 13: 243–256. ———, ———, R. KALLIOLA, A. LINNA, W. DANJOY, AND Z. RODRIGUEZ.
———, AND H. TUOMISTO. 1998. Vegetación natural de la zona de Iquitos. In R. 1995. Dissecting Amazonian biodiversity. Science 269: 63–66.
Kalliola and S. Flores Paitán (Eds.). Geoecologı́a y desarrollo amazónico: VALENCIA, R., H. BALSLEV, AND G. PAZ Y MIÑO. 1994. High tree alpha-diversity
Estudio integrado en la zona de Iquitos, Perú., pp. 253–365. Annales in Amazonian Ecuador. Biodiv. Cons. 3: 21–28.
Universitatis Turkuensis Ser A II 114. ———, R. B. FOSTER,G. VILLA, R. CONDIT, J. -C. SVENNING, C. HERNÁNDEZ,
———, ———, J. VORMISTO, AND N. PITMAN. 2002. Two biases in estimating K. ROMOLEROUX, E. LOSOS, E. MAGÅRD, AND H. BALSLEV. 2004. Tree
range sizes of Amazonian plant species. J. Trop. Ecol. 18: 935–942. species distributions and local habitat variation in the Amazon: Large
SABATIER, D., M. GRIMALDI, M. -F. PRÉVOST, J. GUILLAUME, M. GODRON, M. forest plot in eastern Ecuador. J. Ecol. 92: 214–229.
DOSSO, AND P. CURMI. 1997. The influence of soil cover organization VAN REEUWIJK, L. P. 1993. Procedures for soil analysis. International Soil Ref-
on the floristic and structural heterogeneity of a Guianan rain forest. erence and Information Centre Technical Paper 9, fourth edition. In-
Plant Ecol. 131: 81–108. ternational Soil Reference and Information Centre, Wageningen, The
SVENNING, J. -C. 1999. Microhabitat specialization in a species-rich palm com- Netherlands.
munity in Amazonian Ecuador. J. Ecol. 87: 55–65. VORMISTO, J., H. TUOMISTO, AND J. OKSANEN. 2004a. Palm distribution pat-
———. 2001. On the role of microenvironmental heterogeneity in the ecology terns in Amazonian rainforest: What is the role of topographic variation?
and diversification on neotropical rain-forest palms (Arecaceae). Bot. J. Veg. Sci. 15: 485–494.
Rev. 67: 1–53. ———, O. PHILLIPS, K. RUOKOLAINEN, H. TUOMISTO, AND R. VÁSQUEZ. 2000.
TER STEEGE H., V. G. JETTEN, A. M. POLAK, AND M. J. A. WERGER. 1993. A comparison of fine-scale distribution patterns of four plant groups in
Tropical rain forest types and soil factors in a watershed area in Guyana. an Amazonian rainforest. Ecography 23: 349–359.
J. Veg. Sci. 4: 705–716. ———, J. -C. SVENNING, P. HALL, AND H. BALSLEV. 2004b. Diversity and
TRICHON, V. 1997. Hétérogénéité spatiale d’une forêt tropicale humide de dominance in palm (Arecaceae) communities in terra firme forests in the
Sumatra: Effet de la topographie sur la structure floristique. Ann. Sci. western Amazon basin. J. Ecol. 92: 577–588.
For. 54: 431–446. WEBB, C. O, AND D. R. PEART. 2000. Habitat associations of trees and seedlings
TUOMISTO, H., AND A. D. POULSEN. 1996. Influence of edaphic specialization in a Bornean rain forest. J. Ecol. 88: 464–478.
on pteridophyte distribution in neotropical rain forests. J. Biogeo. 23: WHITMORE, T. C., R. PERALTA, AND K. BROWN. 1985. Total species count in a
283–293. Costa Rican rain forest. J. Trop. Ecol. 1: 375–378.
478 Poulsen, Tuomisto, and Balslev

APPENDIX. Species list of the three plant groups inventoried in a 1-ha plot at Cuyabeno, APPENDIX. Continued.
Amazonian Ecuador. Number of individuals is shown for each species. For
Costaceae
palms, it is indicated in parentheses which stratum of the forest (understory,
Costus lasius Loesener 92
canopy) each species reaches in maturity.
Costus scaber Ruı́z & Pav. 11
Terrestrial pteridophytes Costus sp. nov. 131
Adiantum humile Kunze 914 Cyclanthaceae
Adiantum obliquum Willd. 1 Evodianthus funifer (Poit.) Lindm. 9
Adiantum terminatum Kunze 1004 Cyperaceae
Asplenium cirrhatum Willd 8 Calyptrocarya bicolor (Pfeiffer) Koyama 14
Cyathea lasiosora (Kuhn) Domin 317 Calyptrocarya glomerulata (Brogn.) Urban 188
Cyclodium meniscioides (Kunze) Mett. 16 Hypolytrum scraederianum Nees 32
Danaea lepriuerii Kunze 22 Scleria secans (L.) Urban 22
Danaea sp. aff. nodosa (L.) Sm. 5 Gesneriaceae
Elaphoglossum styriacum Mickel 9 Nautilocalyx ecuadoranus Wiehler 192
Lindsaea cf. divaricata Klotzsch 41 Pearcea sp. 40
Lindsaea lancea (L.) Bedd. var. lancea 1 Heliconiaceae
Lindsaea sp. 1 Heliconia schummaniana Loes. 3
Lomagramma guianensis (Aubl.) Ching 321 Heliconia spathocircinata Aristeg. 2
Lomariopsis nigropaleata Holttum 177 Heliconia velutina L. Anderss. 208
Microgramma megalophylla (Desv.) Sota 2 Marantaceae
Nephrolepis rivularis (Vahl) Krug 22 Calathea altissima (Poepp. & Endl.) Koern. 11
Polybotrya caudata Kunze 2 Calathea lagoagriana Kennedy 64
Polybotrya osmundacea Willd. 269 Calathea micans (Mathieu) Koern. 11
Polybotrya pubens Mart 690 Calathea microcephala (Poepp. & Endl.) Koern. 88
Saccoloma inaequale (Kunze) Mett. 170 Ischnosiphon hirsutus O. G. Peters 78
Salpichlaena hookeriana (O. Kuntze) Alston 27 Ischnosiphon killipii Macbr. 99
Salpichlaena volubilis (Kaulf.) J. Smith 84 Ischnosiphon obliquus (Rudge) Koern. 1
Selaginella exaltata (Kunze) Spring 72 Ischnosiphon puberulus Loes. 98
Selaginella parkeri (Hook & Grev.) Spring 8 Monotagma laxum (Poepp. & Endl.) K. Schum. 195
Poaceae
Selaginella revoluta Baker 1
Cryptochloa unispiculata Soderstrom 95
Thelypteris biformata (Rosenstock) R.M. Tryon 4
Ichnanthus pallens (Sw.) Munro ex Benth. 3
Trichomanes pinnatum Hedw. 317
Olyra caudata Trin. 4
Triplophyllum dicksonioides (Fée) Holttum 2
Olyra latifolia L. 12
Triplophyllum funestum (Kunze) Holttum 161
Angiosperm ground herbs Pariana radiciflora Sagot ex Doell in Mart. 1069
Araceae Piresia goeldii Swallen 486
Anthurium acrobates Sodiro 46 Piresia leptophylla Soderstrom 550
Anthurium atropurpureum R.E. Schult. & Maguire 37 Piresia sympodica (Doell) Swallen 108
Anthurium brevispadix Croat 3 Rubiaceae
Anthurium croatii Madison 573 Geophila cordifolia Miq. 120
Anthurium ernestii Engl. var. ernestii 3 Geophila gracilis (Ruiz & Pav.) DC. 587
Anthurium loretense Croat 5 Geophila repens I. M. Johnston 5
Anthurium oxycarpum Poepp. 2 Psycotria ulviformis Steyerm. 6
Anthurium sp. 1 2 Zingiberaceae
Anthurium sp. 2 2 Renealmia asplundii Maas 1
Chlorospatha longipoda (K. Krause) Madison 22 Renealmia monosperma Miq. 28
Dieffenbachia parvifolia Engl. vel aff. 7 Renealmia thyrsoidea (Ruı́z & Pav.) Poepp. & Endl. ssp. thyrsoidea 9
Monstera obliqua Miq. 28 Palms
Monstera sp. aff. obliqua Miq. 27 Aiphanes ulei (Dammer) Burret (understory) 6
Monstera sp. 1 8 Astrocaryum chambira Burret (canopy) 23
Philodendron alatum Poepp. 96 Attalea maripa (Aubl.) Mart. (canopy) 65
Philodendron asplundii Croat & M.L.C. Soares 367 Bactris maraja Mart. (understory) 24
Philodendron columbianum R.E. Schult. 20 Bactris simplicifrons Mart. (understory) 14
Philodendron ernestii Engl. 18 Bactris sp. 1 (understory) 39
Philodendron fragrantissimum (Hook.) Kunth 18 Bactris sp. 2 (understory) 49
Philodendron hylaeae G.S. Bunting 5 Desmoncus orthacanthos Mart. (canopy liana) 18
Philodendron linnaei Kunth 7 Euterpe precatoria Mart. (canopy) 94
Philodendron panduriforme (Kunth) Kunth 1 Geonoma deversa (Poit.) Kunth (understory) 98
Philodendron pedatum (Hook.) Kunth 5 Geonoma longepedunculata Burret (understory) 514
Philodendron wittianum Engl. 547 Geonoma macrostachys Mart. (understory) 168
Philodendron sp. 1 18 Geonoma maxima (Poit.) Kunth (understory) 43
Philodendron sp. 2 3 Geonoma polyandra Skov (understory) 222
Stenospermation amomifolium (Poepp.) Schott 4 Geonoma stricta var. piscicauda (Dammer) A. J. Hend. (understory) 59
Syngonium podophyllum Schott 52 Geonoma stricta (Poit.) Kunth var. stricta (understory) 1162
Bromeliaceae Geonoma triglochin Burret (understory) 9
Bromelia plumieri (E. Morren) L.B. Smith 1 Hyospathe elegans Mart. (understory) 480
Genus indet 1 1 Iriartea deltoidea Ruiz & Pav. (canopy) 121
Genus indet 2 2 Mauritia flexuosa L.f. (canopy) 1
Commelinaceae Oenocarpus bataua Mart. (canopy) 2514
Tradescantia zanonia (L.) Sw. 4 Oenocarpus mapora H. Karst. (canopy) 3
Geogenanthus sp. 1 Socratea exorrhiza (Mart.) H. Wendl. (canopy) 1
Wettinia maynensis Spruce (canopy) 18