doi:10.1017/S0266467406003282 Printed in the United Kingdom
Species–habitat associations in a Sri Lankan dipterocarp forest
C. V. S. Gunatilleke
, I. A. U. N. Gunatilleke
, S. Esufali
, K. E. Harms
D. F. R. P. Burslem
§ and P. S. Ashton#
Department of Botany, University of Peradeniya, Peradeniya 20400, Sri Lanka
Republic of Panama
‡ School of Forestry and Environmental Studies, Yale University, New Haven, CT 06511-2189, USA
§ School of Biological Sciences, University of Aberdeen, Cruickshank Building, St Machar Drive, Aberdeen AB24 3UU, Scotland, UK
# Arnold Arboretum, Harvard University, 22 Divinity Avenue, Cambridge, MA 02138, USA
Accepted November 2005)
habitats for the 205 species with stems
≥ 1 cm dbh inhabiting a 25-ha plot in the Sinharaja rain forest, Sri Lanka. The
habitats were steep spurs, less-steep spurs, steep gullies and less-steep gullies, all at either lower or upper elevations.
Mean stem density was signiﬁcantly greater on the upper spurs than in the lower, less-steep gullies. Stem density
was also higher on spurs than in gullies within each elevation category and in each upper-elevation habitat than in
its corresponding lower-elevation habitat. Basal area varied less among habitats, but followed similar trends to stem
density. Species richness and Fisher’s alpha were lower in the upper-elevation habitats than in the lower-elevation
habitats. These differences appeared to be related to the abundances of the dominant species. Of the 125 species
subjected to torus-translation tests, 99 species (abundant and less abundant and those in different strata) showed at
least one positive or negative association to one or more of the habitats. Species associations were relatively more
frequent with the lower-elevation gullies. These and the previous ﬁndings on seedling ecophysiology, morphology
and anatomy of some of the habitat specialists suggest that edaphic and hydrological variation related to topography,
accompanied by canopy disturbances of varying intensity, type and extent along the catenal landscape, plays a major
role in habitat partitioning in this forest.
Key Words: Environmental heterogeneity, habitat specialization, rain forest, Sinharaja Forest Dynamics Plot, species-
habitat associations, Sri Lanka, torus translations
Both niche partitioning and dispersal-assembly processes
have been invoked to explain species co-existence and
controls on plant distribution in species-rich tropical tree
communities (Hubbell 2001, Potts et al. 2004, Whitﬁeld
2002, Wright 2002). A role for niche partitioning is
suggested by associations between plant distributions and
environmental conditions at a variety of spatial scales in
both the New and Old World Tropics (Baillie et al. 1987,
Debski et al. 2002, Fine et al. 2005, Gartlan et al. 1986,
Gimaret-Carpentier et al. 1998, 2003; Harms et al.
2001, Itoh et al. 2003, Phillips et al. 2003, Potts et al.
2002). The dispersal-assembly perspective proposes that
Corresponding author: I. A. U. N. Gunatilleke, Department of Botany,
Sri Lanka. Email: firstname.lastname@example.org, email@example.com
communities are non-equilibrium assemblages of species
brought together by accidents of dispersal, and that
localized niche partitioning plays a limited role in species
coexistence. The importance of seed-dispersal limitation
for determining the distribution of species at small scales
has been demonstrated in recent research in tropical
forests (Dalling et al. 2002, Hubbell et al. 1999, Webb &
Peart 2001). However, the relative importance of the two
sets of mechanisms in controlling structure of tropical
rain-forest communities that are rich in closely related
species is poorly understood. This results in part because,
in most cases, the potentially subtle differences in life-
history characteristics among species with contrasting
habitat associations have not been examined.
Central to understanding the distribution patterns of
plant species is the identiﬁcation of habitats at scales
that are relevant to plant populations. The limitations
of small plots in differentiating local habitats have led to
C. V. S. GUNATILLEKE ET AL.
the establishment of large plots (16–52 ha) in tropical
forests where all individuals
≥ 1 cm diameter at breast
height (dbh) have been mapped, measured and identiﬁed
to species (Condit 1995, Condit et al. 1996, Harms et al.
2001, Losos & Leigh 2004, Manokaran et al. 1992,
Sukumar et al. 1992, Valencia et al. 2004). Such data
sets now provide opportunities to test species-habitat
relationships as one step towards understanding the
factors that determine species-distribution patterns.
The 25-ha Forest Dynamics Plot (FDP) at Sinharaja,
south-west Sri Lanka, is among the most topographically
heterogeneous FDPs co-ordinated within the network
of the Center for Tropical Forest Science (CTFS), and
has the highest elevational range (151 m) of the CTFS
plots (comparative data for the various CTFS plots are
available on the CTFS website: http://www.ctfs.si.edu).
While supporting a very large number of stems per
unit area relative to the other large plots, the Sinharaja
FDP also has several series of closely related congeneric,
sympatric species (Ashton et al. 2004, Gunatilleke et al.
2004). The majority of these are endemic to Sri Lanka.
Understanding the presence or absence of habitat pre-
ferences, especially among these congeneric species could
shed some light on the means by which they coexist.
Species-habitat associations have now been described
for FDPs in a semi-deciduous forest on Barro Colorado
Island in Panama (Harms et al. 2001), lowland evergreen
forests at Yasuni in Ecuador (Valencia et al. 2004)
and Lambir in Sarawak (Davies et al. 2005). At
Sinharaja, unlike the other three sites, our interpretation
of differences in species distribution was facilitated by
a substantial body of experimental research that has
investigated the mechanistic basis of species-habitat
associations among closely-related and sympatric species
within the important tree genera Shorea, Mesua and
2003). The combination of habitat associations plus
species traits and performance characteristics provides
a powerful opportunity to address the challenge of
determining the extent to which differences in species
responses to resource availability contribute to their co-
existence in species-rich tropical forests (Hubbell 2001).
The Sinharaja FDP was divided into eight habitats based
on elevation, convexity and slope to address the following
questions: (1) Do stem density, basal area, species richness
and representation by different growth forms vary among
habitats? (2) What proportion of species is signiﬁcantly as-
sociated with one or more of these habitats? (3) Are more
species associated with some habitats than with others?
(4) Are more-abundant species differentially associated
with habitats compared with less-abundant species?
(5) Do species of different growth forms, i.e. structural
guilds, differentially associate with these habitats? (6) Are
signiﬁcant associations, especially differences among
congeneric species, consistent with the available exper-
imental evidence for their ecophysiological differences?
The area studied is the Sinharaja Forest Dynamics Plot
(FDP), a 500
× 500-m (25-ha) permanent study plot
(Figure 1). The Sinharaja FDP is located in the lowland
rain forest of the Sinharaja UNESCO World Heritage
Site at the centre of the ever-wet south-western region
of Sri Lanka (6
21–26 N, 80
21–34 E). The forest has
been classiﬁed as a Mesua–Doona community (de Rosayro
1942), and on a regional scale it represents a mixed
dipterocarp forest (Ashton 1964, Whitmore 1984).
elevational range of 424 m to 575 m asl. The Sinharaja
FDP includes a valley lying between two slopes, a steeper
higher slope facing south-west and a less-steep slope
facing north-east (Figure 1). Seepage ways, spurs, small
hillocks, at least two perennial streams and several
seasonal streamlets cut across these slopes. The ﬂoristics
and forest structure within the plot as a whole have
been documented in Gunatilleke et al. (2004). The
Sinharaja FDP is representative of the ‘ridge-steep slope-
valley’ landscape of the lowland through mid-elevational
rain forests of south-western Sri Lanka. This landform
is a result of differential weathering and erosion of
lithologically less-resistant Precambrian metamorphic
bedrock along structurally controlled parallel strike ridges
and valleys (Cooray 1984, Erb 1984).
To establish the Sinharaja FDP, we followed the
methodology established by Hubbell & Foster (1983) and
Manokaran et al. (1992), to maintain census uniformity
with similar plots within the CTFS network. The Sinharaja
FDP was established in 1993, when it was demarcated
on the horizontal plane into 625 quadrats of 20
× 20 m
) each. The trees in the plot were censused over
≥ 1 cm dbh were measured. Each stem
was mapped and identiﬁed to species, using the National
Herbarium of Sri Lanka, and Dassanayake & Fosberg
Habitats of the Sinharaja FDP were identiﬁed by three
physical parameters, viz. elevation, slope and convexity,
in each of the 20
× 20-m quadrats. The mean of the
elevations at the four corners of each quadrat gave the
quadrat’s elevation. Each quadrat was divided into four
triangular planes, each formed by joining three corners
of the quadrat. The average angular deviation of these
planes from horizontal provided the slope (Harms et al.
2001). Convexity was calculated as in Yamakura et al.
(1995), i.e. as a quadrat’s mean elevation relative to
the mean elevations of its eight immediate neighbouring
quadrats (the focal quadrat mean elevation minus the
mean elevation of the neighbouring quadrats). For each of
the perimeter quadrats of the plot, for which the number of
neighbouring quadrats was
as the elevation of the centre point of the focal quadrat
minus the mean elevation of its four corners. Positive
values indicate convex surfaces, whereas negative values
indicate concave surfaces.
Bivariate scatterplots for each pair of topographic
variables conﬁrmed that they were independent of each
other, with r
values ranging between 0.0356 and 0.141.
topographic properties, so we used all three to deﬁne
eight topographic habitats. Each 20
× 20-m quadrat was
assigned to one of two categories of elevation (upper
vs. lower, divided by the median elevation value for the
FDP), slope (steep vs. less-steep divided by the median
slope value), and convexity (Table 1, Figure 2a).The
abbreviations of the habitat categories used in the entire
paper are explained in Table 1.
× 20-m quadrat of the Sinharaja Forest Dynamics Plot.
Number (and %) of quadrats
Total area on plot (ha)
Upper-elevation steep spurs (USS)
Upper-elevation steep gullies (USG)
Upper-elevation less-steep spurs (ULS)
Upper-elevation less-steep gullies (ULG)
Low-elevation steep spurs (LSS)
Low-elevation steep gullies (LSG)
Low-elevation less-steep spurs (LLS)
Low-elevation less-steep gullies (LLG)
elevation, slope and convexity, each at two levels. Distribution patterns of (b) Mesua nagassarium (blue) found predominantly on upper-elevation
steep spurs and Mesua ferrea (red) found predominantly on upper steep and less-steep gullies. Distribution patterns of (c) Shorea worthingtonii
(black) found predominantly on upper steep spurs, Shorea trapezifolia (blue) found predominantly on the low-elevation less-steep spurs, and Shorea
To assess the structural characteristics of the vegetation
in the different habitats, the means of density and
basal area per quadrat in each habitat were compared.
Similarly, species richness and Fisher’s alpha diversity per
quadrat were calculated and compared among habitats.
Signiﬁcant differences in species richness and Fisher’s
alpha diversity among habitats were determined using
torus-translation tests, described below.
per quadrat among habitats in the Sinharaja 25-ha Forest Dynamics Plot. Total number of free-standing species identiﬁed in the plot was 205.
Signiﬁcance among the respective values column-wise was tested using two-tailed torus-translation tests, P
< 0.025 for either tail. An asterisk (
Mean no. of individuals
Mean basal area (m
Mean Fisher’s alpha
Signiﬁcant associations of species with habitats
Positive and negative associations of species with habitats
were determined by torus-translation tests (Harms et al.
2001). The tests assess the similarity between the spatial
structure of each focal species population and each
habitat. For each species, the observed relative densities of
stems in each of the habitats were compared with expected
relative densities. To obtain the expected values, the true
habitat map was shifted about a two-dimensional torus
by 20-m increments to exhaustively produce all possible
20-m translations of the true habitat map in the four
cardinal directions. Each of the 625 maps provided an
estimate of the expected relative density.
A species was signiﬁcantly positively associated with a
particular habitat if its relative density in the true habitat
> 97.5% of the values obtained from translated
relative density in the true map was
from translated maps. In the Sinharaja FDP, 205 tree
species with stems
≥ 1 cm dbh and 10 species of liana
have been identiﬁed. For the torus-translation tests, we
used the 125 tree species with a density
≥ 100 individuals
in the 25-ha plot.
We also used torus-translations to test whether species
richness, Fisher’s alpha diversity, stem density and basal
area differed among habitats. In each case, the observed
value for a given habitat was compared with a frequency
distribution of expected values generated by an exhaust-
ive set of 20-m incremental torus-translations (analogous
to the procedure used to assess species associations).
Spatial distribution of habitats
The most extensive habitat was the LLG (6.2 ha), whereas
the least extensive and most fragmented was the ULG
(1.3 ha; Table 1, Figure 2a). The remaining habitats
ranged from 1.9 to 4.3 ha in extent. USS and ULS were
greater in extent (4.2–4.3 ha) than the USG and ULG (1.3–
2.7 ha). The extent of the LLG was similar to that of the
three remaining low-elevation habitats combined.
The LLG had the lowest density of individuals
≥ 1 cm
dbh, whereas the USS and ULS had the highest densities
> 400 individuals per quadrat, Table 2); in these cases
the densities depart signiﬁcantly from expectations. The
densities of the remaining habitats had values between
these extremes. Spurs at both elevations, irrespective of
whether they were steep or less steep, had signiﬁcantly
higher densities compared with gullies at the same
Mean basal area among habitats ranged from 1.22 m
in the LLG to 2.36 m
in the USS, although no mean
value differed signiﬁcantly from expectations (Table 2).
The basal area of the tree community on spurs was higher
than that in gullies at each of the two elevations, as with
stem density. The value for each upper-elevation habitat
was greater than that of the corresponding habitat at
Species richness per quadrat showed little variation
among habitats and ranged from 46.7 in the USS to
a high of 56.9 in the LSS (Table 2). Species diversity
per quadrat (measured using Fisher’s alpha) among the
habitats ranged from 14.6 to 21.1 (Table 2). In the upper-
elevation habitats, where the diversity was at the lower
end of the range, spurs showed lower values than gullies.
The diversity values of the low-elevation habitats were
more or less similar, but among them the LLG had the
highest diversity. Diversity was signiﬁcantly higher in LLG
than in USS. The differences among all other values were
not statistically signiﬁcant.
Based on torus-translation tests, a total of 175 signiﬁcant
associations (94 positive and 81 negative) were observed
Forest Dynamics Plot, based on two-tailed torus-translation tests, P
Total no. of signiﬁcant associations
No. of positive associations
in each habitat
No. of negative associations
Totals in upper-elevation habitats
Totals in low-elevation habitats
(Table 3). LLG produced the highest number of signiﬁcant
associations. There were 66 signiﬁcant associations
among the four upper-elevation habitats, of which 18
were positive and 48 were negative. The corresponding
values in the four low-elevation habitats totalled 109,
with 76 positive and 33 negative (Table 3). In the
upper-elevation habitats, spurs had more signiﬁcant asso-
ciations (mostly negative) than gullies, but in the lower-
elevation habitats the pattern was reversed and the gullies
had more associations (mostly positive) than spurs.