全 文 ：Received 8 Dec. 2003 Accepted 8 May 2004
Supported by the Grants from National Science Fund of China for Distinguished Young Scholars (39825106).
* Author for correspondence. Tel: +86 (0)10 82594676; E-mail: .
http://www.chineseplantscience.com
Acta Botanica Sinica
植 物 学 报 2004, 46 (9): 1015-1024
Spatial Heterogeneity of Population Structure of the Mangrove Bruguiera
gymnorrhiza at Yingluo Bay, South-China Coast
LIANG Shi-Chu, DONG Ming*
(Laboratory of Quantitative Vegetation Ecology, Institute of Botany, The Chinese Academy
of Sciences, Beijing 100093, China)
Abstract: The spatial heterogeneity, including distribution pattern, tree perimeter and height
differentiation, and canopy structure heterogeneity, of Bruguiera gymnorrhiza (L.) Lamk populations at
Yingluo Bay, South-China Coast was investigated using the positioning index (CE), differentiation index (TC
and TH), Shannon-Wiener diversity index (D), and Ripley’s K -functions. Most populations showed random
distribution and low differentiation in perimeters and heights of individuals, while a few showed clumped
distribution and clear differentiation. Canopy and gap patches were analyzed at multiple horizontal and
vertical scales using geographic information system (GIS). The mosaic patterns of canopy and gap patches
are different among populations, and could be quantitatively described with the Shannon-Wiener diversity
index based on crown projection. The spatial heterogeneity of the canopy structure changed with spatial
scales, but this kind of change would remain relatively stable over a range of scales. This scale range
could be regarded as the referenced scale for a regeneration or ecological management unit for the
forest.
Key words: mangrove; population structure; spatial heterogeneity; canopy projection; scale
The investigation of the spatial and temporal patterns
of individuals, species, populations, communities, or land-
scapes is now a current topic of research, because spatial
variability is recognized to be a universal ecological char-
acteristic of ecosystems. Identifying and characterizing
the population spatial structure using various quantita-
tive methods are often necessary in ecological studies
(Mladenoff et al., 1993; Batista and Maguire, 1998). The
measurement of structural heterogeneity, however, not as
simple as might be expected. Population spatial structure
has been described, in the most general terms, as the spa-
tial configuration patterns of individuals, such as point
patterns and their variation in size classes, and arrange-
ment of individuals into different canopy layers across
multiple spatial scales (Zenner and Hibbs, 2000; Pélissier
and Goreaud, 2001). To quantitatively describe these spa-
tial structures, numerous approaches have been proposed,
of which nearest neighbor methods and K-functions, spa-
tial autocorrelation and covariance modeling, geostatistical
and spatial econometric modeling, multivariate techniques,
and spatial interaction models are the most common ones
(Haase, 1995; Dale et al., 2002; Fortin et al., 2002).
The stand structure can be split up into three structural
characteristics: positioning, mixture and differentiation, and
they can be quantified with the spatial indexes such as the
positioning index of Clark and Evan (CE), the segregation
index of Pielou (S), the mixture index of Von Gadow (DM)
and the differentiation index of Von Gadow (T), which are
ecologically useful both in describing the stand structure
more accurately and in reflecting the operative processes
causing stand development (Clark and Evans, 1954; Kint
et al., 2000; Lexer et al., 2000; Kint et al., 2003). Because of
problems of canopy accessibility, and the lack of efficient
sampling and modeling methods, most previous spatial
studies of tree populations have focused on the descrip-
tion of tree stems, ignoring the horizontal and vertical struc-
ture of the canopies (Chen et al., 1997; Song et al., 1997).
To address these problems, some efficient methods, such
as TSTRAT, a vertical tree stratification program devel-
oped to quantify the vertical aspect of stand structure
(Latham et al., 1998), and those using geographic informa-
tion system (GIS) and Lidar technology (Song et al., 1997;
Lefsky et al., 1999), have been proposed.
Like other ecosystems, mangrove ecosystems also ex-
hibit various hierarchical levels of spatial and temporal
heterogeneity (Farnsworth, 1998). However, only a few
case studies have been reported so far (Farnsworth and
Ellison, 1996; Osunkoya and Creese, 1997; Ellison, 2002).
Detailed studies in a wide range of locations should there-
fore be conducted to examine spatial s tructural
Acta Botanica Sinica 植物学报 Vol.46 No.9 20041016
characteristics of mangrove populations and their relation-
ships with environmental factors to provide more informa-
tion on mangrove development. On the South-China coast,
Bruguiera gymnorrhiza is a common and important spe-
cies found in the landward part of intertidal mangroves
(Lin and Hu, 1983; Wang et al., 2003). It often forms a
distinct zone mixed with only a few individuals of other
mangrove species. Located at the area nearest to the land,
the forests of this species now suffer more and more im-
pacts from increases in boat traffic, mangrove timber
harvesting, and tourism. In anticipation of changes of the
forests arising from such developments, a baseline eco-
logical study on population structure should be carried
out. Based on these considerations, this study was de-
signed to assess the spatial heterogeneity of the popula-
tion structure within a stand to increase understanding of
the effects of heterogeneity on pattern-process relation-
ships in mangrove forests. To capture stand heterogene-
ity in terms of point patterns of tree distributions, two
commonly cited methods, the positioning index CE and
Ripley’s K-function, were applied, and to quantify the dif-
ference in perimeter and height between neighboring trees,
the differentiation index T was employed. In addition, stem-
mapped crowns were generated and analyzed as canopy
patches within a GIS to quantify patterns within the strata
of canopy. Based on these quantitative methods, we were
concerned with the following four main questions: (1) how
are B. gymnorrhiza trees distributed within a community
and at what scales; (2) are their distribution patterns con-
sistent with the species’ biological and ecological habits,
and intra- or inter-species interactions; (3) how can we
quantify the spatial distributions of height classes (i.e.
sub-populations) in terms of vertical and horizontal canopy
structure; and (4) can the above-selected spatial indices
be effectively applied to comparison of structural hetero-
geneity among different B. gymnorrhiza populations.
1 Materials and Methods
1.1 Study area and investigated forest
The study was undertaken at Yingluo Bay (21°28 N,
109°43 E), where is located geographically to the eastern
and northern coast of Beibu Gulf, South China Sea and
belongs administratively to Guangxi Province in South
China. The climate there belongs to subtropical marine mon-
soon zone, with mean annual air temperature of 22.9 ºC.
The absolute maximum air temperature is 38.2 ºC, while the
absolute minimum is 1.5 ºC. Local tides have a semi-diurnal
regime with an average of 2.53 m and a maximal range of
6.25 m. Mean annual temperature of sea water is 23.5 ºC,
salt content 20‰-23‰ , and pH 7.6-7.8. Intertidal man-
groves cover an area of approximately 1 500 hm2 domi-
nated mainly by Bruguiera gymnorrhiza (L.) Lamk,
Rhizophora stylosa , Kandelia candel, Aegiceras
corniculatum, and Avicennia marina.
Study plots were established in B. gymnorrhiza forest,
which is 3.5-7.0 m in height, 10-30 m in width, and zonally
distributed in the landward part of the intertidal mangroves,
approximately parallel to the shorelines (Liang, 1996). The
constructive species B. gymnorrhiza extends throughout
the forest, forming a virtual monospecific forest. A few
trees of other mangrove species are mostly sporadic along
the seaward edge of the forest.
1.2 Plot design and measurements
Four typical study plots, numbered as Q1, Q2, Q3 and
Q4, were selected in terms both of substrate conditions,
which practically represent different stages in a succes-
sional sequence of tidal flats (Liang, 1996), and of popula-
tion density and structure. At each plot, a 80 m× 10 m
rectangular quadrat was established in B. gymnorrhiza
forest, which was subdivided into10 m× 10 m grids to
ensure an accurate and efficient sampling of stem
coordinates. Tree height, diameter at breast height (DBH),
crown diameter, and the x and y coordinate of all trees ≥
2.5 cm DBH were measured. Soil samples were collected
from the top 20 cm layer of the soil, and analyzed for par-
ticle-size distribution, and so on.
1.3 Data analysis methods
1.3.1 The positioning index The positioning index (CE),
was used to express the extent to which a forest stand
deviates from the “Poisson stand”, and defined as (Clark
and Evans, 1954; Kint et al., 2003):
CE = = (1)
where ri is the distance between tree i and its nearest
neighbor (m), N is the total number of trees in the plot, A is
the area of the plot (m2), and P is the perimeter of the plot
(m). A Poisson stand has a CE value of 1. In order to test
the calculated CE values against a significant deviation of 1,
the null hypothesis (H0: CE = 1 and H1: CE≠ 1) was used:
c = with σrE = = (2)
whereσrE is the standard deviation of rE in a Poisson-
forest of density r.
1.3.2 The differentiation index The differentiation
index (T), includes the perimeter differentiation index (TC)
rA
rE
1
N
0.5 + 0.051 4 + 0.041AN
P
N
P
N3/2
Σ ri
N
i=1
rA- rE
σrE
0.261 36
N× r
0.261 36
N2/A
LIANG Shi-Chu et al.: Spatial Heterogeneity of Population Structure of the Mangrove Bruguiera gymnorrhiza at Yingluo
Bay, South-China Coast 1017
and the height differentiation index (TH) (Kint et al.,
2000; Kint et al., 2003). TC of a single tree i with perim-
eter Ci in relation to its three nearest neighbors is de-
fined as:
TCi = [1- ] (3)
where Cj is the perimeter of the jth nearest neighbor of tree
i. Thus,
TC = TCi (4)
TH of a single tree i with height Hi in relation to its three
nearest neighbors is defined as:
THi = [1- ] (5)
where Hj is the height of the jth nearest neighbor of tree i.
Thus,
TH = THi (6)
1.3.3 Spatial tree pattern analysis Mapped tree pat-
terns of all trees ≥ 2.5 cm DBH were analyzed using
Ripley’s K-function, which determines the consis-
tency of the empirical distribution of distances among
individuals with the Poisson expectations, and is
given by the following equation (Pélissier and
Goreaud, 2001):
K(r) = kij (7)
where kij=1 if the distance between trees i and j is less
than r, and 0 otherwise. The intensity:
l = (8)
is related to N, the number of trees in the plot area A. To
simplify the interpretation, a linearized function L(r) pro-
posed by Besag (1977) was used:
L(r) = K(r)/p - r (9)
A 99% confidence interval of L(r) was computed to test
the null hypothesis of spatial randomness using the Monte
Carlo method with 10 000 simulated random patterns.
All calculations and simulations were performed using
ADS in ADE-4 package (http://pbil.univ-lyon1.fr/ADE-4/
ADE-4.html; Thioulouse et al., 1997).
1.3.4 Canopy structure and heterogeneity
To explore the relationship between DBH and crown
diameter (Y), the following non-linear crown width model
was used:
Y = b0-b1eb
2
×DBH (10)
Two-dimensional canopy projections for individual
canopy strata were generated and analyzed to quantify
patterns within the strata of the canopy. Based on most
crowns of B. gymnorrhiza trees at Yingluo Bay being more
regular and ellipse-shaped (Liang and Wang, 2002a), each
crown projection was simulated by generating an elliptical
polygon using the following equation:
+ =1 (11)
where (xi, yi) are the coordinates for tree i, and ai and bi are
half the maximum and minimum axis width of the crown of
tree i, respectively.
Trees in the canopy layer were divided into three sub-
populations based on tree height classes: lower layer (2-
4 m), intermediate layer (4-6 m), and upper layer (<6 m). A
GIS coverage of two-dimensional crow projections of all
trees for each sub-population was generated. At Yingluo
Bay, the canopy of B. gymnorrhiza forests is discontinued,
and there are various sizes of canopy gaps. Thus based
on the size range of both canopy patches and canopy gap
(here defined as areas not occupied by crowns) patches,
all of the above GIS layers were then divided into grids of
1.00, 1.67, 2.00, 2.50, 3.30, 5.00, and 10 m to explore the
hierarchical structures of canopy patches at multiple scales.
To quantify the structural heterogeneity of canopies across
scales, the Shannon-Wiener diversity index (D) was
calculated:
D = -Σ Pi ln Pi (12)
where Pi is the proportion of non-gap area occupied by
each sub-population i.
2 Results
2.1 Size structure
Table 1 shows that only three mangrove species were
found in B. gymnorrhiza forest at Yingluo Bay, and B.
gymnorrhiza absolutely dominates and plays a crucial role
in the structural configuration of the forest. The DBH and
height classes of B. gymnorrhiza populations varied in
different plots (Fig.1). Plots Q1 and Q2 had most trees with
8.5-14.5 cm DBH, accounting for 77.6% and 67.1%,
respectively. In plot Q3, 86.5% of trees were 11.5-20.5 cm
DBH; in plot Q4, 68.9% of trees were 2.5-5.5 cm DBH. The
absence of trees was found in the DBH classes of 2.5-
8.5 cm and 20.5-26.5 cm in plot Q1, 2.5-5.5 cm in plot Q2
and 2.5-8.5 cm in plot Q3. At larger scale, the height struc-
ture of B. gymnorrhiza forest could be divided into one or
two layers. In plots Q1, Q2 and Q3, only one height layer,
1
N
1
NΣi=1
N
MIN (Hi, Hj)
MAX (Hi, Hj)
1
3Σj=1
3
1
NΣi=1
N
1
l Σi=1
N
Σ
j≠i
N
A
(xi+ai cos q)2
a2i
(yi+bi sin q)2
b2i
MIN (Ci, Cj)
MAX (Ci, Cj)
1
3Σj=1
3
Acta Botanica Sinica 植物学报 Vol.46 No.9 20041018
i.e. tree layer, was found; trees <2.5 m height were absent.
In contrast, in plot Q4, where trees <2.5 m height were
abundant and accounted for 75.0%, the forest had two
height layers, i.e. shrub and tree layer.
2.2 Indexes of CE, TH and TC
The spatial index values for B. gymnorrhiza popula-
tions at Yingluo Bay are summarized in Table 2. In which,
values of CE clearly indicated that B. gymnorrhiza trees
present in plots Q1, Q2 and Q3 were randomly positioned,
while in plot Q4 clumped. No substantial differences were
found between results for TC and TH. TH values of plots
Q1, Q2 and Q3 were less than 0.2, indicating that neighbor-
ing trees show low differentiation in heights and that only
limited height differences between neighboring trees can
be observed, but those of plot Q4 larger than 0.4, indicat-
ing that neighboring trees show clear differentiation in
heights. TC values of plots Q1 and Q3 were less than 0.2,
indicating that neighboring trees show low differentiation
in perimeters, but those of plots Q2 and Q4 ranged be-
tween 0.2 and 0.4, indicating that neighboring trees show
moderate differentiation in perimeters.
2.3 Spatial pattern
The spatial patterns detected using Ripley’s univariate
Table 1 Soil texture, species compositions and their basic statistic characteristics of Bruguiera gymnorrhiza forest stands sampled at
Yingluo Bay
Plot number Q1 Q2 Q3 Q4
Soil texture Light clay, Medium clay, Sandy loam, Sandy loam,
slightly hardened semi-hardened muddy muddy
Number of trees (stems/hm2)
Bruguiera gymnorrhiza 838 (838) 988 (988) 650 (650) 2 250 (563)
Rhizophora stylosa 0 (0) 63 (63) 25 (25) 50 (50)
Kandelia candel 25 (0) 0 (0) 13 (0) 63 (0)
Mean crown diameter of all trees (m) 2.2 ± 0.7b 2.0 ± 0.5b 3.0 ± 0.7c 1.5 ± 0.9a
Mean crown diameter of canopy trees (m) 2.2 ± 0.7a 2.0 ± 0.5a 3.0 ± 0.7b 2.8 ± 0.6b
Mean height of all trees (m) 4.1 ± 0.9c 5.3 ± 1.2d 3.5 ± 0.6b 1.8 ± 1.4a
Mean DBH of all trees (cm) 42.6 ± 13.05c 13.0 ± 4.4c 15.8 ± 3.9b 6.3 ± 4.0a
Values in the brackets are tree density of the canopy layer. Values (X ± SE) that are significantly different (P < 0.05) from each other are
indicated by different superscripted letters. Crown diameter is the arithmetic average of diameters of the maximum and minimum axis of
the crown. DBH, the diameter at breast height of trees; Q1, Q2, Q3 and Q4 refer to plots established at different sites.
Fig.1. Frequency of trees by height class (top panel) and by
DBH class (bottom panel) for Bruguiera gymnorrhiza popula-
tions in the four plots (Q1, Q2, Q3 and Q4) at Yingluo Bay. In
terms of tree growth and development characteristics, the mea-
sured tree heights were grouped in height classes of 1m range
except trees <1.5 m height, and the diameters were grouped in
DBH classes of 3 cm range. Abbreviations are the same as in
Table 1.
Table 2 Indices of CE, TH and TC for Bruguiera gymnorrhiza
populations in the four plots established at Yingluo Bay
Plot number Q1 Q2 Q3 Q4
CE 1.012 1.047 0.979 0.923*
c 0.195 0.802 -0.290 -1.991
TH 0.187 0.174 0.131 0.415
TC 0.194 0.241 0.192 0.398
The symbol * indicates a significant difference from 1 at a level of
5%. CE = 1 indicates a tendency towards random spacing, while CE
<1 towards clumping. T values of 0 to 0.2 indicate low differentiation,
0.2 to 0.4 moderate di fferent iat ion , and 0 .4 to 0.6 clear
differentiation. c, the standard variate of the normal curve; CE,
the positioning index of Clark and Evan; TC, the perimeter differ-
entiation index of Von Gadow; TH, the height differentiation index
of Von Gadow; Q1, Q2, Q3 and Q4 refer to plots established at
different sites.
LIANG Shi-Chu et al.: Spatial Heterogeneity of Population Structure of the Mangrove Bruguiera gymnorrhiza at Yingluo
Bay, South-China Coast 1019
Fig.2. Diagram of the L(r) function for Bruguiera gymnorrhiza populations at Yingluo Bay. Solid line, values of the L(r) function;
dashed line, 0.01 confidence intervals for complete spatial randomness (CSR). For poisson patterns, L(r) = 0; for clumped patterns,
L(r)0; for regular patterns, L(r) < 0. A value of L(r) outside the confidence interval is interpreted as a significant departure from CSR
towards clumping or regularity. Q4U, the upper layer of plot Q4 with trees of >7 cm DBH; Q4L, the lower layer of plot Q4 with trees
of ≤7 cm DBH. Abbreviations are the same as in Table 1.
L(r) function showed that B. gymnorrhiza trees in plots
Q1, Q2 and Q3 were significantly randomly distributed at
almost all scales, only with an exception of a tendency
toward regularity at the scales of 1 m in plot Q3 (Fig.2). The
trees in plot Q4 showed clumped distribution for distance
scales r > 0.8 m, and random distribution at smaller
distances. Further analyses indicated that the clumped distri-
bution of the B. gymnorrhiza population in plot Q4 resulted
from the sapling trees with DBH ≤7cm, which accounted for
75.0% of the total trees and grew mostly around their parent
trees (Fig.2, Q4L); but the trees with larger DBH were signifi-
cantly randomly distributed at all scales (Fig.2, Q4U).
2.4 Canopy heterogeneity
Figure 3 shows that the B. gymnorrhiza populations
had an open canopy. Of the canopy trees in the forest, B.
gymnorrhiza accounted for 91.8%-100.0%, while R. stylosa
below 9%. Crown diameters of B. gymnorrhiza trees in-
creased monotonically with their DBHs (Fig.4); however,
most crowns were less than 5 m in diameter, with a few
exception. The tree crowns in plots Q3 and Q4 tended to
be more spread out and had larger diameters than those in
plots Q1 and Q2. The mean crown diameters for B.
gymnorrhiza trees in plots Q1, Q2, Q3 and Q4 were (2.2 ±
0.7), (1.9 ± 0.5), (3.0 ± 0.7) and (2.7 ± 0.6) m, respectively.
Figure 3 shows that the horizontal patterns of canopy
structures, as represented by aggregates of crown
projections, were diverse and different between
populations. Moreover, the areas of canopy and gap
patches change evidently from the lower to the upper
canopy, except those for plot Q1(Fig.5). Heterogeneity of
the canopy structures measured by the Shannon-Wiener
index (D) varied with spatial scales (Table 3).
3 Discussion
3.1 Spatial heterogeneity and scale
Heterogeneity occurs when some quantitative or quali-
tative descriptors of an ecosystem vary significantly from
one location to another (Pélissier and Goreaud, 2001).
Acta Botanica Sinica 植物学报 Vol.46 No.9 20041020
Interpretations of such spatial variations may depend on
the scale of observation (Levin 1992; Dale et al., 2002). It
has been found that the spatial distribution of trees is
usually heterogeneous, and can exhibit one pattern at
smaller scales but a different one at other scales. For
example, Moeur (1993) reported that trees in hemlock
Fig.3. Horizontal spatial patterns of Bruguiera gymnorrhiza populations measured at Yingluo Bay. The dots denote the relative
positions of trees, and the elliptical polygons are the two-dimensional projections of crowns. Abbreviations are the same as in Fig.2.
Fig.4. Estimated crown diameter (Y) with tree size (DBH) for Bruguiera gymnorrhiza populations at Yinglio Bay. Solid lines are
produced based on a non-linear regression Equation (10): Q1, Y = 5.331 5-7.669 4e0.070 3DBH, R = 0.865***, n = 67; Q2, Y = 7.441 9-
7.243 1e0.021 5DBH, R = 0.865***, n = 79; Q3, Y = 5.129 1-5.724 7e0.065 5DBH, R = 0.672***, n = 43; Q4, Y = 4.993 2-5.334 0e0.073 2DBH,
R = 954***, n = 180. ***, the significance at a level of 0.001. Abbreviations are the same as in Table 1.
LIANG Shi-Chu et al.: Spatial Heterogeneity of Population Structure of the Mangrove Bruguiera gymnorrhiza at Yingluo
Bay, South-China Coast 1021
forests tend toward regular patterns at smaller scales but
clustered patterns at larger scales. Additionally, there has
been a hypothesis: the wider the range of scales examined,
the more levels of patchiness are generally discovered,
and the more likely the researcher will point out the com-
plex nature of heterogeneity (Azovsky et al., 2000). The
results in Fig.2 of our study support these points of views.
Table 3 shows that the Shannon-Wiener index of both
canopy and gap patches varied with spatial scales. Finer
scales may include only one crown type or canopy gap.
For example, at the scale of 1 to1.7 m, the canopy patch
was in fact composed of crown of only one tree or of some
small sapling trees (Liang and Wang, 2002b). As the mea-
sured scale increases, the number of crown types increase
until all types are included.
3.2 Spatial heterogeneity and species characteristics
and substrate conditions
Heterogeneity can result either from exogenous or en-
dogenous factors. Many biological processes, such as
reproduction, intra- or inter-species interactions, death,
can induce a spatially heterogeneous pattern of popula-
tions (Pélissier and Goreaud, 2001; Antos and Parish, 2002).
For example, some authors (Hubbell, 1997; Pitman et al.,
1999) emphasized that dispersal pattern of propagules con-
trols the spatial distribution of plant species and their
coexistence. The tendency to clumped distribution in the
B. gymnorrhiza trees might result mainly from sapling or
small trees aggregating beneath large trees, because it has
been found during the course of field investigation that
most seedlings and saplings of this species often aggre-
gate around parent trees, only a few growing under canopy
gaps and forming distinct, separated patches. This pat-
tern may be attributed to the distribution and redistribu-
tion of propagules and to the occurrence of safe sites for
germination and establishment. B. gymnorrhiza is a vi-
viparous mangrove species. The length of its propagules
is on average 20 cm, with the average diameter of about
1.5 cm. Through their gravitational forces, most propagules
Table 3 Shannon-Wiener index (D) of the canopy composition for the Bruguiera gymnorrhiza populations in the four plots (Q1, Q2,
Q3 and Q4) at Yingluo Bay, calculated at different measured scales
Scale (m)
Q1 Q2 Q3 Q4
Canopy Gap Canopy Gap Canopy Gap Canopy Gap
1.0 0.610 0.319 0.701 0.321 0.557 0.310 0.542 0.421
1.7 0.714 0.475 0.875 0.476 0.627 0.385 0.547 0.451
2.0 0.704 0.470 0.884 0.518 0.646 0.443 0.606 0.632
2.5 0.733 0.544 0.943 0.604 0.667 0.493 0.643 0.751
3.3 0.716 0.677 1.013 0.822 0.655 0.559 0.676 0.841
5.0 0.617 0.730 0.959 0.915 0.573 0.584 0.663 0.933
10.0 0 0 0.294 0.368 0.294 0.368 0.604 0.997
Fig.5. Canopy and gap areas at different tree heights for
Bruguiera gymnorrhiza populations at Yingluo Bay. Abbrevia-
tions are the same as in Table 2.
Fig.6. The TC and TH distribution for Bruguiera gymnorrhiza
populations at Yingluo Bay. Abbreviations are the same as in
Table 2.
Acta Botanica Sinica 植物学报 Vol.46 No.9 20041022
can self plant beneath the crowns of parent trees. Thus,
although gaps in the canopy allow seedlings and saplings
of trees to escape competition or predation (Smith, 1987;
Osborne and smith, 1990), the seedlings and saplings of B.
gymnorrhiza often show a clumped distribution around
their parent trees, and moreover the scale size of the
clumped patches is determined mostly by the crown size
of parent trees (Liang and Wang, 2002b). In plot Q4, mean
crown diameter of canopy trees was (2.7 ± 0.6) m, while
sapling trees showed a clumped distribution at the scales
of 0.8 to 2.5 m. Therefore, the conclusion from these analy-
ses is that the fine-scale spatial pattern of B. gymnorrhiza
is generally determined by its self-planting characteristics.
Intra- or inter-species competition for nutrient and spa-
tial resources may be the main constraint to trees and lead
to changes in the spatial patterns within and between life
stages on both a small and large scale. Generally, spatial
pattern of trees experiencing competitive thinning may
become simply less clumped (He and Li, 1999), more regu-
lar (Moeur, 1993), or random with time (Dovciak et al., 2001).
Figure 2 shows that sapling trees (Q4L) are clumped at
scales r >0.8 m, while mature trees (Q1, Q2, Q3 and Q4U)
were randomly distributed at almost all measured scales.
These results suggest that spatial distributions of B.
gymnorrhiza populations may change with both time and
scale, and intra-species competition and self-thinning be
the two major processes causing pattern dynamics of B.
gymnorrhiza populations. Moreover, according to Figs.3,
5, the B. gymnorrhiza forests were more open. Thus, we
can hypothesize that among mature trees, the density-de-
pendent mortality has greatly decreased, even ceased, and/
or was overruled by density-independent factors, such as
typhoon and human disturbance.
In the four measured plots, B. gymnorrhiza forests were
composed of almost pure stands, but the density of canopy
trees were relatively lower, with 613-1 051 stems/hm2 for
trees ≥2.3 m tall. Figure 3 shows that the canopies were
more open and discontinued and the coverage density
ranged from about 43% to 54%. The horizontal patterns of
canopy structures, as represented by aggregates of crown
projections, were diverse and different between
populations. Canopy gaps, as the inverse of canopy
patches, were characterized by an irregular morphology,
being highly connected. However, the nest ing
phenomenon, i.e. sapling trees grow under canopy gaps,
was not common. Take plot Q4 as an example, the canopy
gaps reached a total area of about 421.6 m2, but most trees
≤ 7cm DBH aggregate around their parent trees (Fig.3,
Q4).
The substrate conditions of mangrove forests influ-
ence plant growth, population density and consequently
affect forest dynamics and spatial structure. At Yingluo
Bay, it has been found that the ecological successional
trend of substrates in B. gymnorrhiza forests is: muddy→
slightly hardened→ semi-hardened flats. Table 1 shows
that in plots Q1 and Q2, the forest soils have been more or
less hardened, which are not benefit to the dispersal and
development of viviparous seedlings and thus directly in-
fluence the quantity and spatial distribution of individuals
in the regenerative layer; in plot Q3, there are plentiful rock
outcrops that reduce the spatial positions for individuals
to occupy; in plot Q4, the natural regeneration is relative
better, with a more obvious undergrowth layer. The trees
with DBH of ≤7 cm are absent in Q1, Q2, and Q3, but they
are abundant in plot Q4 and account for 75.0% of the total.
Therefore, it is apparent that frequency distribution pat-
terns of TC and TH values are closely related with both the
substrate conditions of forest stands and the regenerative
processes of the species.
3.3 Quantitative description of canopy structure
Mangrove forests are structurally complex within the
canopy, where gaps generated by natural and anthropo-
genic disturbance create a discontinuous landscape
(Minchinton, 2001). The ability of GIS to quantify canopy
structure and heterogeneity, both in terms of species com-
position and spatial configuration, across scales has been
demonstrated. Influenced by the density of trees, tree
location, the species composition, and the shape and size
of the crown, canopy heterogeneity varies significantly
among size and height classes (Song et al., 1997; Chen
and Bradshaw, 1999). Table 3 shows that the spatial con-
figuration patterns of canopy patches of B. gymnorrhiza
trees were different among different populations and scale
dependent, although the canopies were relative more open
and the heterogeneity degree of the canopy structure of
each population relatively gently changed with spatial
scales. This suggests that the methods used, including
the generating of a GIS coverage of two-dimensional crown
projections and the Shannon-Wiener diversity index (D),
are capable of detecting even fine structure changes, such
as the height differentiation of canopy structure, and sug-
gests that the spatial characteristics of the population struc-
ture are a function of spatial scale. Effectively quantifying
the complexity and dynamics of the canopy structure of
each population within the forest will help to understand
the overall structure and function of the forest.
3.4 Applicability of spatial indices
The positioning index, CE, is based on the distance
LIANG Shi-Chu et al.: Spatial Heterogeneity of Population Structure of the Mangrove Bruguiera gymnorrhiza at Yingluo
Bay, South-China Coast 1023
from an individual to its nearest neighbor, irrespective of
direction (Clark and Evans, 1954). Although the determi-
nation of a series of such distances measured in a given
population can seem quite laborious, CE is easy to
calculate, can be corrected for edge effects, and offers a
statistical test (Kint et al., 2000). Ripley’s K-function uses
all plant-to-plant distances to characterize point patterns.
For the determination of tree spatial patterns, the results
of both the CE values in Table 2 and the Ripley’s K-func-
tions in Fig.2 for the B. gymnorrhiza populations were the
same. For differentiation, the only nearest-neighbor index
known to us is T, which can be also calculated for every
single tree and then categorized and counted for a fre-
quency distribution. This approach is especially useful
for interpreting index-change over time (Kint et al., 2003).
Figure 6 shows that the TC and TH values of 52%-60% of
B. gymnorrhiza trees were lower than the average value
each. This means that only limited perimeter and height
differences between neighboring trees can be observed.
All these results suggested that both indexes of CE and T
and K-function offer an effective description of the spatial
structure of B. gymnorrhiza populations. They are capable
of detecting even fine structure changes, such as the com-
petition-based shift of perimeter and height differentiation.
Acknowledgements: We thank Dr. MI Xiang-Cheng for
his valuable help in the calculation of Equation (10), and
Dr. Vincent Kint, Dr. Elizabeth Farnsworth, and Dr. Hans
Zuuring for offering us their papers.
References:
Antos J A, Parish R. 2002. Dynamics of an old-growth, fire-
initiated, subalpine forest in southern interior British
Columbia: tree size, age, and spatial structure. Can J For
Res, 32: 1935-1946.
Azovsky A I, Chertoprood M V, Kucheruk N V, Rybnikov P V,
Sapozhnikov F V. 2000. Fractal properties of spatial distri-
bution of intertidal benthic communities. Mar Biol, 136: 581-
590.
Batista J L F, Maguire D A.1998. Modeling the spatial structure
of tropical forests. For Ecol Manage, 110: 293-314.
Besag J. 1977. Contribution to the discussion of Dr Ripley’s
paper. J Roy Stat Soc. B39:193-195.
Chen J, Desanker P V, Reed D D, Bradshaw G A, Franklin J F.
1997. Modeling canopy structure and heterogeneity across
scales: from crowns to canopy. For Ecol Manage, 96: 217-
229.
Chen J, Bradshaw G A. 1999. Forest structure in space: a case
study of an old growth spruce-fir forest in Changbaishan
Natural Reserve, PR China. For Ecol Manage, 120: 219-
233.
Clark P J, Evans F C. 1954. Distance to nearest neighbour as a
measure of spatial relationships in populations. Ecology,
35: 445-453.
Dale M R T, Dixon P, Fortin M J, Legendre P, Myers D E,
Rosenberg M S. 2002. Conceptual and mathematical rela-
tionships among methods for spatial analysis. Ecography,
25: 558-577.
Dovciak M, Frelich L E, Reich P B. 2001. Discordance in spatial
patterns of white pine (Pinus strobus) size-classes in a patchy
near-boreal forest. J Ecol, 89: 280-291.
Ellison A M. 2002. Macroecology of mangroves: large-scale pat-
terns and processes in tropical coastal forest. Trees, 16: 181-
194.
Farnsworth E J. 1998. Issues of spatial, taxonomic and temporal
scale in delineating links between mangrove diversity and
ecosystem function. Global Ecol Biogeogr Lett, 7: 14-25.
Farnsworth E J, Ellison A M. 1996. Scale-dependent spatial and
temporal variability in biogeography of mangrove-root
epibiont communities. Ecol Monogr, 66: 45-66.
Fortin M J, Dale M R T, Hoef J V. 2002. Spatial analysis in
ecology. El-Shaarawi A, Piegorsch W W. The Encyclopedia
of Environmetrics. Vol. 4. Chichester: John Wiley and Sons,
Ltd. 2051-2058.
Haase P. 1995. Spatial pattern analysis in ecology based on
Ripley’s K-function: Introduction and methods of edge
correction. J Veg Sci, 6: 575-582.
He Y-H, Li C-L. 1999. The ecological geographic distribution,
spatial pattern and collecting history of Cycas panzhihuaensis
populations. Acta Phytoecol Sin , 23: 23-30. (in Chinese
with English abstract)
Hubbell S P. 1997. A unified theory of biogeography and relative
species abundance and its application to tropical rain forests
and coreal reefs. Coral Reefs, 16: 9-21.
Kint V, Lust N, Ferris R, Olsthoorn A F M. 2000. Quantifica-
tion of forest stand structure applied to scots pine (Pinus
sylvestris L.) forests. Invest Agr Sist Recur For Fuera de
Serie, 1: 147-163.
Kint V, van Meirvenne M, Nachtergale L, Geudens G, Lust N.
2003. Spatial methods for quantifying forest stand structure
evolution: a comparison between nearest-neighbor indices
and geostatistics. For Sci, 49: 36-49.
Latham P A, Zuuring H R, Coble D W. 1998. A method for
quantifying vertical forest structure. For Ecol Manage, 104:
157-170.
Lefsky M A, Cohen W B, Acker S A, Parker G G, Spies T A,
Harding D. 1999. Lidar remote sensing of the canopy struc-
ture and biophysical properties of Douglas-fir western hem-
lock forests. Remote Sens Environ, 70: 339-361.
Acta Botanica Sinica 植物学报 Vol.46 No.9 20041024
Levin S A. 1992. The problem of pattern and scale in ecology.
Ecology, 73: 1943-1967.
Lexer M J, Lexer W, Hasenauer H. 2000. The use of forest
models for biodiversity assessments at the stand level. In-
vest Agr Sist Recur For Fuera de Serie, 1: 297-316.
Liang S-C. 1996. Study on the mangrove communities in Yingluo
Bay of Guangxi. Acta Phytoecol Geobot Sin , 20: 310-321.
(in Chinese with English abstract)
Liang S C, Wang B S. 2002a. Fractal characteristics of popula-
tion canopy structure of the mangrove, Bruguiera
gymnorrhiza (L.) Lamk. Mar Sci Bull, 5: 54-61 .
Liang S-C, Wang B-S . 2002b. Fractal analysis of sapling patch
shape of the mangrove species, Bruguiera gymnorrhiza.
Guihaia, 22: 481-484. (in Chinese with English abstract)
Lin P, Hu J-T.1983. The mangrove communities in Guangxi.
Acta Bot Sin,25:95-97. (in Chinese)
Minchinton T E. 2001. Canopy and substratum heterogeneity
influence recruitment of the mangrove Avicennia marina. J
Ecol, 89: 888-902.
Mladenoff D J, White M A, Pastor J, Crow T R. 1993. Compar-
ing spatial pattern in unaltered old-growth and disturbed for-
est landscapes. Ecol Appl, 3: 294-306.
Moeur M. 1993. Characterizing spatial patterns of trees using
stem-mapped data. For Sci, 39: 756-775.
Osunkoya O O, Creese R G. 1997. Population structure, spatial
pattern and seedling establishment of the grey mangrove,
Avicennia marina var. australasica, in New Zealand. Aust J
Bot, 45: 707-725.
Osborne K, Smith T J. 1990. Differential predation on man-
grove propagules in open and closed canopy forest habitats.
Vegetatio, 89: 1-6.
Pélissier R, Goreaud F. 2001. A practical approach to studying
the spatial structure in simple cases of heterogeneous veg-
etation stands. J Veg Sci, 12: 99-108.
Pitman N C A, Terborgh J, Silman M R, Nuñez P V. 1999. Tree
species distribution in an upper Amazonian forest. Ecology,
80: 2651-2661.
Smith T J. 1987. Effects of light and intertidal position on seed-
ling survival and growth in tropical tidal forests. J Exp Mar
Biol Ecol, 110: 133-146.
Song B, Chen J, Desanker P V, Reed D D, Bradshaw G A,
Franklin J F. 1997. Modeling canopy structure and hetero-
geneity across scales: from crowns to canopy. For Ecol
Manage, 96: 217-229.
Wang B-S , Liang S-C, Zhang W-Y , Zan Q-J . 2003. Mangrove
flora of the world. Acta Bot Sin , 45: 644-653.
Thioulouse J, Chessel D, Dolédec S, Olivier J M. 1997. ADE-4:
a multivariate analysis and graphical display software. Stat
Comp, 7: 75-83.
Zenner E K, Hibbs D E. 2000. A new method for modeling the
heterogeneity of forest structure. For Ecol Manage, 129:
75-87.
(Managing editor: HAN Ya-Qin)