Plant Ecology 161: 123–135, 2002.                                          123
   © 2002 Kluwer Academic Publishers. Printed in the Netherlands.



An ordination study to view vegetation structure dynamics in disturbed
and undisturbed mangrove forests in Kenya and Sri Lanka

F. Dahdouh-Guebas 1,*, J.G. Kairo 2, L.P. Jayatissa 3, S. Cannicci 4 and N. Koedam 1
1
Laboratory of General Botany and Nature Management, Mangrove Management Group, Vrije Universiteit
Brussel, Pleinlaan 2, B-1050 Brussels, Belgium; 2Kenyan Marine and Fisheries Research Institute, P.O. Box
81651, Mombasa, Kenya; 3Department of Botany, University of Ruhuna, Matara, Sri Lanka; 4Dipartimento di
Biologia Animale e Genetica ’Leo Pardi’, Università degli Studi di Firenze, Via Romana 17, I-50125 Firenze,
Italy; *Author for correspondence
Received 27 October 2000; accepted in revised form 20 September 2001


Key words: CCA, DCA, Disturbance, Forestry, Propagule predation, Rehabilitation


Abstract

The mangrove vegetation of a disturbed and undisturbed site in both Kenya and Sri Lanka was investigated in
the field for three vegetation layers: adult trees, young trees, and juvenile trees. A minimum of 25 sample points,
in which the vegetation was described and environmental factors (salinity, light intensity, land/water ratio, abun-
dance of herbivorous crabs and snail abundance) were measured or estimated, were taken on each site. Detrended
correspondence analysis (DCA) and canonical correspondence analysis (CCA) were used to summarize the data
bulk, to investigate the vegetation dynamics (e.g., comparability of species distribution in the three vegetation
layers), and/or to link the vegetation data to the environmental factors. Results showed that species clusters were
relatively easy to delineate, whether mangrove zonation was present or not. Among the environmental factors,
the abundance of propagule predators (mostly sesarmid crabs) contributed significantly to the variation in veg-
etation and could be an explanatory parameter for the observed vegetation data in a majority of sites. In the site
where it was not, the most important factor in the ordination was the land/water ratio, which is important at the
ecological level as well (link between water level and vegetation dynamics). However, none of the environmen-
tal factors could successfully explain the total variability in the vegetation data suggesting that other, more de-
termining factors exist. Our results further provide information on the dynamic or non-dynamic nature of a forest
and on its ability to rejuvenate, and may contribute to appropriate forestry management guidelines in the future.


Introduction                              ate results with direct application to forest manage-
                                    ment planning (e.g., Holmgren et al. (1997);
Vegetation dynamics, defined as changes in stand            Holmgren and Thuresson (1998)). Studying appropri-
structure and composition over time, are a major as-          ate forestry variables and parameters and applying
pect of vegetation ecology (e.g., Putz and Chan            them to specific vegetation layers in a certain se-
(1986); Smith and Huston (1989); Heil and Van             quence can generate invaluable data on past, present,
Deursen (1996); Murali et al. (1998); Dahdouh-Gue-           and even future vegetation structure dynamics.
bas et al. (2000c)), yet such research has received            Among the tools that are available to analyse veg-
little attention. Very often, the study of vegetation         etation data, ordination algorithms are an appropriate
dynamics relies on remote sensing, combined with            choice to explore the relationship between the vege-
geographic information systems (GIS), and usually           tation structure and the environment (Kent and Coker
constitutes the only retrospective basis of comparison         1992). However, in mangrove ecology, particularly in
to actual vegetation data (Dahdouh-Guebas et al.            studies on vegetation dynamics, ordination algorithms
2000c). In addition, this type of research can gener-         have seldom been utilized (e.g., Ukpong (1995)). In
124

the past, vegetation dynamics have been deduced
from ordinations focusing on one time stamp (e.g.,
from principal component analysis by Hobbs and
Grace (1981)), but these analyses create problems that
can be overcome by repetitive analysis over time.
Studies of the mangrove vegetation using sequential
aerial photography has shown that autogenous
changes can occur in mangrove ecosystems (Dah-
douh-Guebas et al. 2000c). Since mangroves become
increasingly threatened by various human impacts
(e.g., ITTO (1993); Pernetta (1993); IUCN (1996);
Farnsworth and Ellison (1997); Kjerfve et al.
(1997)Dahdouh-Guebas et al. (2000a, 2000b, 2000c
(in press), 2001)), there is a need to investigate the
mangrove vegetation with the purpose of predicting
changes in the future. Adverse alterations in vegeta-
tion structure could then be countered in time by var-
ious forms of human interference, such as artificial
rehabilitation.
  In this study we investigate and compare the veg-
etation structure and composition from different veg-
etation layers of undisturbed and disturbed mangrove   Figure 1. Map of the western part of the Indian Ocean showing
sites in Kenya and Sri Lanka using two ordination     the locations of our study sites in Kenya and Sri Lanka. For de-
techniques. In order to interpret presence/absence and  tailed cartographic material on these study sites we refer to Dah-
abundance data from different vegetation layers, we    douh-Guebas et al. (2000a, 2000b, in press); Gallin et al. (1989);
                             Kairo (2001).
employed the use of detrended correspondence anal-
ysis and canonical correspondence analysis, to ex-    both Mida Creek and Gazi Bay constitute a fringing
plore the comparability of species distribution in the  forest type (cf. Lugo and Snedaker (1974)). The man-
different vegetation layers and to test the hypothesis  groves in Gazi Bay have long been over-exploited for
that the environmental factors measured contribute    wood and, likewise, is disturbed. Although woodcut-
significantly to the observed variability in the vegeta-  ting is also occurring in Mida Creek (Dahdouh-Gue-
tion, respectively. In addition, we demonstrate how    bas et al. 2000a), the mangroves are much less af-
such an ordination study may reveal the vegetation    fected in Mida Creek than in Gazi Bay (Kairo 2001).
structure dynamics, test the underlying explanatory      Of the two mangrove forests investigated along the
hypotheses, and allow for prediction of future vege-   southwestern side of Sri Lanka, the first one was lo-
tation changes.                      cated in the Pambala area of Chilaw Lagoon, in Sri
                             Lanka’s intermediate climate zone (Mueller-Dombois
                             1968). The mangroves there are of the fringe type
Description of the study sites              (c.f., Lugo and Snedaker (1974)), and have a rather
                             irregular distribution along a complex of creeks (i.e.,
Single undisturbed and disturbed study sites were     Marambettiya Ela, Bate Ela, Pol Ela, and Dutch
chosen in both Kenya and Sri Lanka (Figure 1). The    Channel). Most freshwater influx originates from the
Kenyan study sites were located in Mida Creek (3°20    Karambalan Oya catchment, whereas outflow to the
S, 40°00 E) and Gazi Bay (4°26 S, 39°30 E). Mida     sea is possible at Chilaw (07°35 48 N, 079°47 25
Creek has a narrow opening towards the ocean lo-     E) and Toduwawa (07°29 30 N, 079°48 16 E). The
cated about 100 km north of Mombasa. The creek has    Pambala site is well preserved and has the greatest
no overland freshwater input, but benefits from a high   species diversity in Southwestern Sri Lanka (Jayatissa
groundwater outflow (Tack and Polk 1999). Gazi Bay,    et al. 2002). Nevertheless, the mangroves in Chilaw
a mangrove bay widely open to the ocean, is located    Lagoon have recently been subjected to strong an-
about 40 km south of Mombasa. Unlike Mida Creek,     thropogenic influences as a result of shrimp farming
Gazi Bay is fed by two seasonal rivers. Mangroves in
                                                       125

(Foell et al. 1999; Dahdouh-Guebas et al. 2000c, in    upper or lower intertidal areas. In contrast, Pambala
press).                          contains a plateau with pools and Galle possesses a
  The second Sri Lankan mangrove forest was lo-     multitude of small islands and pools, primarily result-
cated between Galle and Unawatuna (06°01 N –       ing from the burrowing activities of the mangrove
80°14 E) in the wet climate zone of Sri Lanka (Muel-   mud lobster Thalassina anomala Herbst.
ler-Dombois 1968). This basin and riverine mangrove
type (cf. Lugo and Snedaker (1974)) covers an area
of 1.5 km 2 and is located at about 600 m from the    Methods
Indian Ocean shore. Two rivers run through the man-
grove forest–the Thalpe Ela, which discharges into    In each of the mangrove sites, sample points (Table 1)
the ocean, and the Galu Ganga, which is a tributary    along parallel as well as orthogonal transects were
of the former. The mangrove forest in Galle has been   chosen at 10 m intervals. In each sample point, 4
subjected to anthropogenic influence over the last 50   quadrants were established as in the Point-Centred
years (Dahdouh-Guebas et al. 2000c).           Quarter Method of Cottam and Curtis (1956). The
  All ten East-African mangrove species occur along   height and D 130, a new term coined by Brokaw and
the Kenyan coast - Avicennia marina (Forsk.) Vierh.,   Thompson (2000) to deal with the unstandardised use
Bruguiera gymnorrhiza (L.) Lam., Ceriops tagal      of diameter at breast height (DBH), of the adult and
(Perr.) C.B. Robinson, Heritiera littoralis Dryand.,   young trees (Ht < 1.3 m or D 130 < 2.5 cm) closest to
Lumnitzera racemosa Willd., Pemphis acidula Forst.,    the sample point were recorded in each quadrant (=
Rhizophora mucronata Lam., Sonneratia alba Sm.,      quarter). Total vegetation coverage in the 5 × 5 m 2
Xylocarpus granatum Koen, and X. moluccensis       quadrat (= square) nearest to the sampling point was
(Lamk.) Roem. The common species present in Pam-     estimated in each quadrant (as in the Braun-Blanquet
bala (Sri Lanka) are Aegiceras corniculatum (L.)     relevé method). Measurements were taken in July
Blanco, Avicennia offıcinalis L., Bruguiera gymnor-    1996 for Mida Creek, in January 1997 for Galle, in
rhiza, B. sexangula (Lour.) Poir., Excoecaria agal-    February 1997 for Pambala, and in July 1997 for Gazi
locha L., Lumnitzera racemosa, Rhizophora apicu-     Bay.
lata Bl., R. mucronata, and Xylocarpus granatum. In      In the same 5 × 5 m quadrats, mangrove juveniles;
Galle the common species are B. gymnorrhiza, B. sex-   defined as propagules, seeds, or young plants with no
angula, E. agallocha, H. littoralis, and R. apiculata.  more than 6 leaves; were identified and counted. At
A considerable area in the forest was covered with the  each sample point the following environmental fac-
non-mangrove herbaceous species Fimbristylis sal-     tors were measured: salinity (using an Atago refrac-
bundia (Nees) Kunth subsp. pentaptera (Nees) T.      tometer, S/Mill-E, Japan), light intensity (using a
Koyama. For a more detailed description of the man-    Lutron luxmeter, LX105, France), and, in the 10 × 10
grove species and mangrove associates in Pambala     m 2 area around the sample point, centre land/water
and Galle, we refer to Jayatissa et al. (2002). The    ratio (visually estimated), crab burrow density, and
taxonomy and nomenclature of mangrove species are     Terebralia palustris L. snail density. All factors were
according to Tomlinson (1986); Duke and Jackes      investigated in March-April 1998 and November
(1987); Duke (1991).                   1999 for both Galle and Pambala, in April 1999 for
  Apart from the more species diverse nature of     Gazi Bay, and in May 1999 for Mida Creek. The dry
mangroves in Sri Lanka, there are two other main     season is in January and February, whereas the peaks
differences between the Kenyan and Sri Lankan sites.   of the wet season are around May and November in
The first is the spring tidal amplitude, which is more   both Kenya and Sri Lanka.
than 3.5 m in Kenya, but less than 1 m in Sri Lanka      Vegetation data were inserted into two types of
(Spalding et al. 1997); locally the tidal amplitude    matrices – one based on presence/absence data (0 or
changes as little as 15 cm per week throughout the    1), and the other based on abundance data, ranging
year (Dahdouh-Guebas et al. 2000c). The second dif-    from 0 – 4 per sample point (i.e., 4 quadrants togeth-
ference concerns mangrove species zonation. In      er). For each of these two types of matrices, three
Kenya zonation is very pronounced in all mangrove     sub-types of matrices were created, which included
areas, whereas in Sri Lanka species zonation is sel-   adult and young trees alone, or adult and young trees
dom present (i.e., very localized partial or semi-zona-  combined with juveniles based on either a separate
tion). In both Sri Lankan sites, there was no slope in  input for each of the fieldwork expeditions or a com-
                                                                                             126




Table 1. Summary of the structural vegetation data for the 4 sites located in Kenya (KE) and Sri Lanka (LK).

Site          Disturbance      Number of sample    Mangrove species     Tree density (stems  Basal area      Mean tree height (m) Complexity Index
                                      1             -1       2
                       points         present (in order of   0.1 ha )       (m /0.1ha)                 (C.I.) 3
                                          2
                                   importance )

Mida, KE        undisturbed      25           C. tag, R. muc, A.    107.6         0.697        8.9          3.3
                                   mar, B. gym, X. gra
Gazi, KE        disturbed       99           R. muc, S. alb, A.    89.2         2.144        5.6          2.5
                                   mar, C. tag, B. gym
Pambala, LK      undisturbed      127          R. muc, L. rac, R.    191.5         1.012        6.2          12.0
                                   api, B. sex, A. off, E.
                                   aga, B. gym, X. gra,
                                   H. lit, A. cor
Galle, LK       disturbed       116          E. aga, R. api, B.    131.1         0.764        4.7          2.5
                                   gym, H. lit
1
 A. cor = Aegiceras corniculatum, A. mar = Avicennia marina, A. off = Avicennia offıcinalis, B. gym = Bruguiera gymnorrhiza, B. sex = Bruguiera sexangula; C. tag = Ceriops tagal, E.
aga = Excoecaria agallocha, H. lit = Heritiera littoralis, L. rac = Lumnitzera racemosa, R. api = Rhizophora apiculata, R. muc = Rhizophora mucronata, S. alb = Sonneratia alba, X. gra
= Xylocarpus granatum
2
 Ranked according to the importance value of Curtis (1959)
3
 C.I. = the product of number of species, basal area (m 2 0.1 ha −1), mean tree height (m) and tree density 0.1 ha −1 × 10 −3 expressed in a 0.1 ha plot (Holdridge et al. 1971)
                                                        127

bination of all fieldwork expeditions. In our view the    Results
latter was most representative for the study of the
mangrove vegetation structure dynamics, since the      From the structural attributes of the vegetation, tree
place (= quadrat) that a juvenile reaches is more im-    densities and forest heights are higher in the less dis-
portant than the time (= fieldwork expedition) in      turbed forests (Table 1). This is true for basal area in
which it reaches that particular quadrat. However, the   the Sri Lankan site, but not in the Kenyan sites. Gazi,
spread of fieldwork expeditions indicate the annual     although disturbed, is dominated by Rhizophora mu-
variability of juvenile distribution. The separate ma-   cronata with a presence of very large Avicennia ma-
trices of adult and young trees tell us more about the   rina and Sonneratia alba trees, the latter two species
relation with environmental variables, since the      of which greatly influence the total basal area of the
‘noise’ that new juveniles with a compromised future    forest. Nevertheless, the undisturbed sites (Mida and
might introduce in the matrices is removed.         Pambala) are structurally more complex than the dis-
  Using PC ORD for Windows (McCune and Mef-        turbed sites (Gazi and Galle), according to the com-
ford 1997), DCA (Detrended Correspondence Analy-      plexity indices (Table 1).
sis: Hill and Gauch (1980)) or CCA (Canonical Cor-       The variation in species data and the comparabil-
respondence Analysis: Ter Braak (1986, 1994)) were     ity of the distribution of species in the different veg-
applied. The percentage of variance in the matrix that   etation layers in each of the sites were analysed in the
is explained by each axis was calculated. For the      DCA (Figure 2, Table 2), whereas the hypothesis-
DCA this calculation was done a posteriori by calcu-    testing combination of species and environmental
lating the coefficient of determination (r 2) between    data is shown in the CCA (Figure 3, Table 3). For a
Euclidean distances in the ordination space and those    majority of the ordinations (87%), those based on ma-
in the original space. The environmental variables and   trices with abundance data (0–4) had better explana-
parameters used in the CCA were selected according     tory power than those based on presence-absence val-
to their relevance as suggested in the scientific litera-  ues only. The exclusion of juveniles from the input
ture. Environmental variables included salinity (aver-   matrices, explored in the research phase, did not alter
aged if not significantly different between seasons),    the explanatory power of the ordination or the ease
light intensity, land cover, and abundance of herbivo-   of interpretation substantially and are therefore not
rous crabs and snails. Monte Carlo tests were per-     shown. Except for Pambala, the ordinations incorpo-
formed to determine the significance of the eigenval-    rating the juvenile trees can be interpreted without
ues and of the species-environment correlations up to    severe complications and the presence of the juve-
a significance level of 0.001 (= 999 runs).         niles is important from a future vegetation structure
  It is essential to recall and emphasize at this point  stand point. For Pambala, the combination of many
that DCA was used only to explore and view the data     species and their juveniles render interpretation diffi-
set. Independent from all causal factors, the distribu-   cult and therefore the ordination was repeated omit-
tion of tree species in adult, young, and juvenile veg-   ting the juvenile trees (Figure 4, Table 4). The omis-
etation layers in the DCA provides insight into the     sion of the outliers formed by young Heritiera littora-
present and potential future vegetation dynamics. Past   lis trees in this ordination, did likewise not affect the
dynamics were reflected by frequency distributions in    ordination results.
part based on D 130 classes at 2.5 cm intervals. The      Although the species-environment correlations in
CCA, on the other hand, was used as a hypothesis-      the CCA were highly significant along the first axis,
testing technique to understand some of the determin-    except for Mida Creek, and along the second axis,
ing environmental factors important in the establish-    except for Mida Creek and Galle, the environmental
ment of a particular mangrove vegetation structure.     variables measured explained but a small part of the
After obtaining the CCA results, based on both veg-     variation in species data. Generally, the variance ex-
etation and environmental data, the results from the    plained was low in all sites (Table 2). Nevertheless
DCA, based exclusively on vegetation data, was used     some remarkable patterns can be recognized. For in-
once more to evaluate whether the environmental fac-    stance, some species-environment associations can be
tors chosen contributed substantially to the total vari-  recognized in the CCA and information on mangrove
ability in vegetation.                   vegetation structural dynamics can be inferred based
                              on both types of ordinations. However, such an inter-
128




Figure 2. Plots of the detrended correspondence analysis in Mida (a), Gazi (b), Pambala (c) and Galle (d). For Mida Xylocarpus granatum
formed an outlier to the ordination and was therefore omitted from the input matrix. Points of one species are encircled and where necessary
the points have been shaded to ease interpretation. See Table 2 for numerical results. See Fig. 4a for a simplified graph of Pambala. (AT =
adult trees, YT = young trees, JT = juvenile trees, propagules and seeds, Acor = Aegiceras corniculatum, Amar = Avicennia marina, Aoff =
Avicennia offıcinalis, Bgym = Bruguiera gymnorrhiza, Bsex = Bruguiera sexangula; Ctag = Ceriops tagal, Eaga = Excoecaria agallocha,
Hlit = Heritiera littoralis, Lrac = Lumnitzera racemosa, Rapi = Rhizophora apiculata, Rmuc = Rhizophora mucronata, Rspp = Rhizophora
juveniles that could not be identified unambiguously to species level, Salb = Sonneratia alba, Xgra = Xylocarpus granatum).

pretation can be performed independently from the            bivorous crabs and snails are present (Figure 3). The
indirect ordination (DCA) only (see discussion).             same can be observed more clearly in Figure 4b for
  The species-environment correlations that can be           Pambala, where the species associated with low crab
recognized from the biplots in Figure 3, bearing in           abundances are all propagule producing species,
mind that the account of other non-measured biotic            namely: Aegiceras corniculatum, Avicennia offıcina-
and abiotic factors remain to be elucidated but are           lis, Bruguiera gymnorrhiza, Rhizophora apiculata,
different by site. One of the common results is the fact         and R. mucronata. In Galle, crab abundance was so
that species clusters usually appear where fewer her-          poorly associated with the second axis as compared
                                                                     129




Figure 3. Biplots of the canonical correspondence analysis in Mida (a), Gazi (b), Pambala (c) and Galle (d). For Mida Xylocarpus granatum
formed an outlier to the ordination and was therefore omitted from the input matrix. Points of one species are encircled and where necessary
the points have been shaded to ease interpretation. See Table 3 for numerical results. See Fig. 4b for a simplified graph of Pambala. (% Land
= % land cover, the rest being water; Crabs = # herbivorous crabs m −2; Snails = # snails m −2; Sal = salinity and Lux = light intensity, both
of them with mention of the year of sampling. Species legend as in Figure 2).

to other environmental factors, that it did not even           ing on the margin of an island with roots in the wa-
show on the biplot (Figure 3d). In Galle, where tidal           ter. In the zoned mangroves (Gazi and Mida) the spe-
influence is minimal, a slope is absent and a mosaic            cies assemblage clusters are more separated and rec-
of islands and pools is present. The land-water ratio           ognisable than in the semi-zoned (Pambala) or non-
is responsible for most of the variation in vegetation          zoned (Galle) mangroves. Between the less disturbed
data (intraset correlation = −0.870). In the same site          sites (Mida and Pambala) and the disturbed sites
the association of Excoecaria agallocha and Heriti-            (Gazi and Galle), no clear and relevant differences
era littoralis with high land cover values, and that of          were found in the ordinations.
Rhizophora apiculata with intermediate land cover               Frequency distributions based on D 130 classes at
values (Figure 3d), reflects the field observation that           2.5 cm intervals were combined with frequency dis-
the former two species are usually located on top of           tributions of juveniles during the fieldwork missions
the small islands, whereas R. apiculata is often grow-
130

                                 Table 2. Eigenvalues and an assessment of how the variance ex-
                                 plained is distributed among the primary axes of the detrended
                                 correspondence analysis in Mida (a), Gazi (b), Pambala (c) and
                                 Galle (d). The evaluation is performed by calculating the coeffi-
                                 cient of determination (r 2) between distances in the ordination
                                 space and distances in the original space (see text). See figure 2 for
                                 graphical results.

                                 (a) Mida              Axis 1    Axis 2    Axis 3

                                 Eigenvalue             0.370    0.154    0.066
                                 r2  variance explained (%)    57.3     8.30     7.70
                                 (b) Gazi              Axis 1    Axis 2    Axis 3

                                 Eigenvalue             0.404    0.195    0.092
                                 r2  variance explained (%)    54.1     10.0     3.5
                                 (c) Pambala             Axis 1    Axis 2    Axis 3

                                 Eigenvalue             0.787    0.434    0.251
                                 r2  variance explained (%)    20.5     9.4     −3.7
                                 (d) Galle              Axis 1    Axis 2    Axis 3

                                 Eigenvalue             0.436    0.217    0.153
                                 r2  variance explained (%)     23.1     15.6     4.8


                                 emphasize and compare the potential of various re-
                                 mote sensing technologies, case-studies, and future
                                 applications are being reported (e.g., Rehder and
                                 Patterson (1986); Tassan (1987); Aschbacher et al.
                                 (1995); Blasco et al. (1998); Holmgren and Thures-
                                 son (1998); Hyyppä et al. (2000)). In contrast to this
                                 growing literature base, our study proposes a slightly
                                 different approach and creates a similar but non-
                                 mapped output using ordination techniques to provide
                                 new insight into the past, present, and future vegeta-
                                 tion structure of mangrove forests. Of particular im-
                                 portance are predictive capabilities for mangrove deg-
                                 radation. Mangrove areas may well have a luxurious
                                 vegetation cover and thus be mapped successfully, in
                                 the field, and in particular in the youngest layers of
Figure 4. Graphical output of the DCA (a) and the CCA (b) for
                                 the understory, the situation may be different com-
adult and young mangrove trees in Pambala. (Legends as in Fig-
ures 2 and 3).                          pared to the past (Kairo 2001). In fact, the understory
                                 of an extensive forest may indicate degradation.
and are shown in Figure 5. It can be seen that for          As shown in the results, species clusters are clearly
some species certain D 130 classes are missing.         distinguishable and to a certain extent the zonation or
                                 semi-zonation present in the sites can be detected
                                 along a line that is closely associated to the first axis
Discussion                            (Figure 2). Some species, such as Ceriops tagal,
                                 Rhizophora apiculata and R. mucronata, have re-
Vegetation investigations have developed a greater        stricted dispersal (species cluster with small coverage
dependence on remote sensing technology in recent        on the first axis), whereas others, such as Avicennia
years (e.g., Gang and Agatsiva (1992); Cohen et al.       marina and Excoecaria agallocha, are distributed
(1996); Ramachandran et al. (1998); Dahdouh-Gue-         over considerably larger areas (species clusters with
bas et al. (1999)), and ‘reviews’ and ‘advances’ that      a wide coverage on the first axis). For Avicennia ma-
                                                                    131

Table 3. Eigenvalues, variance in species data, and species-environment (Spp-Env) correlations for the primary axes from the canonical
correspondence analysis in Mida (a), Gazi (b), Pambala (c) and Galle (d). The p-values from the Monte Carlo tests, which tested the sig-
nificance of the eigenvalues and the Spp-Env correlations, are indicated between brackets and significant results with an asterisk. See Fig-
ure 3 for graphical results.

(a) Mida                    Axis 1                Axis 2                Axis 3

Eigenvalue                   0.097 (p = 0.317)          0.035 (p = 0.551)          0.011 (p = 0.766)
Variance explained (%)             10.3                 3.7                 1.2
Pearson Spp-Env correlation          0.536 (p = 0.702)          0.481 (p = 0.543)          0.408 (p = 0.402)
(b) Gazi                    Axis 1                Axis 2                Axis 3

Eigenvalue                   0.054 (p = 0.022)*          0.032 (p = 0.004)*          0.012 (p = 0.162)
% variance explained              3.7                 2.2                 0.9
Pearson Spp-Env correlation          0.505 (p = 0.028)*          0.423 (p = 0.025)*          0.327 (p = 0.095)*
(c) Pambala                  Axis 1                Axis 2                Axis 3

Eigenvalue                   0.348 (p = 0.001)*          0.130 (p = 0.002)*          0.067 (p = 0.002)*
% variance explained              7.4                 2.8                 1.4
Pearson Spp-Env correlation          0.693 (p = 0.031)*          0.666 (p = 0.001)*          0.418 (p = 0.258)
(d) Galle                   Axis 1                Axis 2                Axis 3

Eigenvalue                   0.127 (p = 0.001)*          0.039 (p = 0.055)          0.023 (p = 0.065)
% variance explained              6.6                 2.0                 1.2
Pearson Spp-Env correlation          0.636 (p = 0.001)*          0.364 (p = 0.204)          0.360 (p = 0.054)


Table 4. Eigenvalues and an assessment of how the variance explained is distributed among the primary axes of the DCA (a) and the CCA
(b) in Pambala. The evaluation of the DCA is performed simply by calculating the coefficient of determination (r 2) between distances in the
ordination space and distances in the original space (see text). The p-values from the Monte Carlo tests, which tested the significance of the
eigenvalues and the Spp-Env correlations in the CCA, are indicated between brackets and significant results with an asterisk. See Figure 4
for graphical results.

(a) DCA                     Axis 1                Axis 2                Axis 3

Eigenvalue                   0.788                0.465                 0.301
r2  variance explained (%)          32.8                 13.6                 1.5
(b) CCA                     Axis 1                Axis 2                Axis 3

Eigenvalue                   0.294 (p = 0.007)*          0.117 (p = 0.006)*          0.051 (p = 0.210)
% variance explained              5.7                 2.3                  1.0
Pearson Spp-Env correlation           0.634 (p = 0.025)*          0.476 (p = 0.054)           0.372 (p = 0.274)


rina this is clearly linked to its pioneer nature (cf.          crease of the terrestrial nature of the vegetation. E.
Osborne and Berjak (1997)). A. marina spreads seeds            agallocha is a disturbance resistant species (Tomlin-
over a wider area than where they can actually estab-           son 1986) that can also easily colonise new areas over
lish. Therefore they display a distribution pattern for          a small temporal scale (Dahdouh-Guebas et al.
juvenile trees that is less associated to that for the          2000c). It is always located at the highest topographi-
adult trees than that for young trees. This same life           cal levels in the mangrove and therefore often asso-
history characteristic is true for Avicennia offıcinalis.         ciated to terrestrial non-mangrove species such as the
This cannot be confirmed, however, for Heritiera lit-           herbaceous Fimbristylis salbundia subsp. pentaptera
toralis, where young trees were observed in the sec-           (pers. obs.). In total, less than half of the species clus-
tion of Pambala where conspecific adult trees are ab-           ters display the associated adult-young tree distribu-
sent. Adult Heritiera trees were observed in Pambala           tion as compared to the adult-juvenile tree distribu-
but outside the study area (pers. obs.). Contrary to the         tion.
pioneering role played by these species, the same pat-            The results of the frequency distributions show
tern observed for E. agallocha is explained by an in-           that Xylocarpus granatum is not rejuvenating in
132

Mida, whereas it is a relatively new species in Pam-     the environmental factors measured in this study. Vi-
bala. Also, Heritiera littoralis can be considered to    sual observations in Kenya (Mida and Gazi) have, in
show a relatively recent establishment in Galle. To-     contrast, revealed that the vegetation zone dominated
gether with Avicennia and Rhizophora species, the      by Rhizophora mucronata and that dominated by Ce-
above two species are known to grow to very tall trees    riops tagal, located at higher intertidal areas, corre-
in Kenya and Sri Lanka. This is not true for other      sponds with the upper topographical limit of the area
species, which may explain the smaller D 130 classes     flooded during neap tide periods and consequently
represented. The frequency distribution for Bruguiera    with the time the vegetation is submerged. According
gymnorrhiza in Mida (Fig. 5a) lacks two successive      to Brakel (1982) this elevation corresponds to at least
D 130 classes in between classes that are actually rep-   1 hour of immersion per tidal cycle, but this can vary
resented. Although no such anomalous patterns in       significantly between sites (Dahdouh-Guebas et al., in
D 130 classes are observed for Rhizophora mucronata     press). Under the lower tidal influences in Sri Lanka,
or Ceriops tagal in the same site, their patterns were    the conditions for Rhizophora are different and tem-
also more variable than expected. Comparing all sites    porally irregular, but nevertheless also vary from per-
confirms the degree of disturbance, since the less dis-    manent flooding by freshwater in the rainy season to
turbed Mida and Pambala clearly have more and taller     very little flooding in the dry season. However, in
trees than Gazi and Galle. The observed D 130 fre-      Galle, many Rhizophora trees are in contact with
quency distributions further revealed some differ-      pools. In Pambala, association with pools is only the
ences between disturbed and undisturbed sites in       case for the most landward and the most creekward
terms of number and size of trees. In Mida, although     Rhizophora individuals. Contact at both ends with
a less disturbed site, the anomalous frequency distri-    water probably leads to a below-ground supply for the
butions observed for Bruguiera gymnorrhiza, Ceriops     Rhizophora trees located farther from the surface wa-
tagal, and Rhizophora mucronata could be the result     ter. Whereas the influence of groundwater supply on
of a higher harvesting preference for these species, as   mangrove distribution has clearly been demonstrated
was shown by Dahdouh-Guebas et al. (2000a); Kairo      by Tack and Polk (1999) for the Kenyan coast, in Sri
(2001). However, the species harvested are obviously     Lanka no conclusive results have been found. Another
preferred by man, but the species harvested are not     variable that probably plays a substantial role is the
necessarily the species that regenerate.           soil texture. It has recently been found that soil tex-
  The appearance of species clusters in the CCA      ture is a major explanatory factor in the distribution
where fewer herbivorous crabs and snails are present,    of sesarmid crabs (Ballerini et al. 2000; Cannicci et
except for the species that are predated less or not at   al. 2000), and a recent study on rooting of mangrove
all, like Excoecaria agallocha, Lumnitzera racemosa,     species in Sri Lanka suggests that soil texture also in-
or Xylocarpus granatum, agrees with the observed       fluences the distribution of mangrove species (Dah-
predation intensity by these animals in Kenya (Dah-     douh-Guebas et al. 2001). Other edaphic characteris-
douh-Guebas et al. 1997, 2000b) and Sri Lanka (Dah-     tics, such as nutrient availability, may be important
douh-Guebas 2001). The land/water ratio in Galle,      explanatory factors but remain untested within ordi-
which was the most important explanatory variable      nation investigations.
for vegetation data (Figure 3d), is probably due to its     This study shows that ordinations can provide in-
high ecological importance to vegetation dynamics. It    formation with respect to the dynamics of a vegeta-
was suggested recently that the water level in man-     tion assemblage by using different vegetation layers
grove areas that lack a clear slope (and therefore      per species as an input, rather than single species. The
hardly display zonation) is a major factor in the es-    dynamic or non-dynamic nature of a particular forest
tablishment of a particular vegetation structure and     is captured by ordination approaches, and annual
that, interacting with propagule predation, it is a driv-  monitoring of the juveniles can reveal whether or not
ing force in the dynamics of a mangrove forest (Dah-     the forest is rejuvenating. The combination of a for-
douh-Guebas 2001). Sites with an irregular topogra-     est’s dynamism and its rejuvenation provides a pos-
phy like Galle will create microhabitats for juvenile    sible indication on the destiny of a forest and may
and adult mangroves and associated fauna that will      raise the awareness to protect or to rehabilitate a man-
change with different hydrological factors.         grove area. Combined with remote-sensing and GIS
  Rhizophora species are among the species that dis-    and in-depth knowledge of silvicultural restoration
play distribution patterns that are difficult to link with  techniques, the methodololgy used in the present
                                                                     133




Figure 5. Frequency distributions of D 130 classes (at 2.5 cm intervals) combined with frequency of juveniles during the fieldwork missions.
The right part of the frequency distributions provide information about the history of the forest, whereas the left part may be used for infer-
ence about the future. The young Heritiera littoralis tree in Pambala was omitted because it does not display on the logarithmic scale.

study opens doors to the development of an early-             velopment of an emergency plan for declining man-
warning system. Such a system may allow for the de-
134

grove forests long before degradation symptoms are           Cohen W.B., Kushla J.D., Ripple W.J. and Garman S.L. 1996. An
visible in the field.                           introduction to digital methods in remote sensing of forested
                                     ecosystems: focus on the Pacific Northwest, USA. Environ-
                                     mental Management 20: 421–435.
                                    Cottam G. and Curtis J.T. 1956. The use of distance measures in
                                     phytosociological sampling. Ecology 37: 451–460.
Acknowledgements                            Curtis J.T. 1959. The Vegetation of Wisconsin. An Ordination of
                                     Plant Communities. University of Wisconsin Press, Madison,
                                     Wisconsin, USA.
Many thanks are due for the fieldwork assistance of           Dahdouh-Guebas F. 2001. Propagule predation in Sri Lankan man-
Fatuma M. Saidi, Daglas W.N. Thisera, Anouk Ver-             groves and its effect on vegetation structure dynamics. In: Man-
heyden, Ilse Van Pottelbergh, Abdulhakim Abubakr             grove vegetation structure dynamics and regeneration. F. Dah-
Ali Jilo, and Saman Nishante. We are very grateful to           douh-Guebas. PhD Dissertation, Vrije Universiteit Brussel,
                                     Brussels, Belgium.
Fabienne van Rossum (Vrije Universiteit Brussel,
                                    Dahdouh-Guebas F., Coppejans E. and Van Speybroeck D. 1999.
General Botany and Nature Management) for assis-             Remote sensing and zonation of seagrasses and algae along the
tance during data analysis and to Ken Krauss (U.S.            Kenyan coast. Hydrobiologia 400: 63–73.
Geological Survey, National Wetlands Research Cen-           Dahdouh-Guebas F., Mathenge C., Kairo J.G. and Koedam N.
ter) for providing critical comments and some lan-            2000a. Utilization of mangrove wood products around Mida
                                     Creek (Kenya) amongst subsistence and commercial users.
guage style corrections. The first author is a postdoc-
                                     Economic Botany 54: 513–527.
toral researcher from the Fund for Scientific Research         Dahdouh-Guebas F., Thisera D.W.N., Wickramasinghe A. and
(FWO-Flanders). Research financed by the European             Koedam N. 2001. Regeneration status of mangroves under nat-
Commission (Contract EBR IC18-CT98-0295), and               ural and nursery conditions in Galle and Pambala, Sri Lanka.
with a specialisation fellowship of the Institute for the         In: Mangrove vegetation structure dynamics and regeneration.
                                     F. Dahdouh-Guebas. PhD Dissertation, Vrije Universiteit Brus-
Promotion of Innovation by Science and Technology
                                     sel, Brussels, Belgium.
in Flanders (IWT).                           Dahdouh-Guebas F., Verneirt M., Tack J.F. and Koedam N. 1997.
                                     Food preferences in Neosarmatium meinerti de Man (Decapo-
                                     da: Sesarminae) and its possible effect on the regeneration of
References                                mangroves. Hydrobiologia 347: 83–89.
                                    Dahdouh-Guebas F., Verheyden A., De Genst W., Hettiarachchi S.
                                     and Koedam N. 2000b. Four decade vegetation dynamics in Sri
Aschbacher J., Ofren R., Delsol J.P., Suselo T.B., Vibulsresth S.
                                     Lankan mangroves as detected from sequential aerial photog-
  and Charrupat T. 1995. An integrated comparative approach to
                                     raphy: a case study in Galle. Bulletin of Marine Science 67:
  mangrove vegetation mapping using advanced remote sensing
                                     741–759.
  and GIS technologies: preliminary results. Hydrobiologia 295:
                                    Dahdouh-Guebas F., Verneirt M., Tack J.F., Van Speybroeck D. and
  285–294.
                                     Koedam N. 1998. Propagule predators in Kenyan mangroves
Ballerini T., Macia A. and Cannicci S. 2000. Zonation of Mangrove
                                     and their possible effect on regeneration. Marine and Freshwa-
  Crabs at Inhaca Island (Mozambique): a new hypothesis. Ab-
                                     ter Research 49: 345–350.
  stract Volume of the Meeting on Mangrove Macrobenthos, 7-11
                                    Dahdouh-Guebas F., Zetterström T., Rönnbäck P., Troell M. and
  September, Mombasa, Kenya.
                                     Koedam N. 2000c. Conservation of mangroves vs. develop-
Blasco F., Gauquelin T., Rasolofoharinoro M., Denis J., Aizpuru
                                     ment of shrimp farming in Pambala - Chilaw lagoon, Sri Lanka.
  M. and Caldairou V. 1998. Recent advances in mangrove stud-
                                     In: Flos R. and Creswell L. (eds), Responsible Aquaculture in
  ies using remote sensing data. Marine and Freshwater Research
                                     the New Millenium. European Aquaculture Society, Oostende,
  49: 287–296.
                                     Belgium, p. 167.
Brakel W.H. 1982. Tidal patterns on the East African coast and
                                    Dahdouh-Guebas F., Zetterström T., Rönnbäck P., Troell M., Wick-
  their implications for the littoral biota. In: Proceedings of the
                                     ramasinghe A. and Koedam N. 2002. Recent changes in land-
  Symposium on the Coastal and Marine Environment of the Red
                                     use in the Pambala-Chilaw Lagoon complex (Sri Lanka) inves-
  Sea, Gulf of Aden and Tropical Western Indian Ocean 2., pp.
                                     tigated using remote sensing and GIS: conservation of man-
  403–418.
                                     groves vs. development of shrimp farming. Environment, De-
Brokaw N. and Thompson J. 2000. The H for DBH. Forest Ecol-
                                     velopment and Sustainability (in press).
  ogy and Management 129: 89–91.
                                    Dahdouh-Guebas F., Verneirt M., Cannicci S., Kairo J.G., Tack J.F.
Cannicci S., Skov M.W. and Vannini M. 2000. The influence of
                                     and Koedam N. An explorative study on grapsid crab zonation
  abiotic and biotic factors on the distribution of mangrove crabs
                                     in mangrove forests in Kenya. Wetlands Ecology and Manage-
  at Mida Creek (Kenya): a multivariate approach. Abstract Vol-
                                     ment (in press).
  ume of the Meeting on Mangrove Macrobenthos, 7-11 Septem-
                                    Duke N.C. 1991. A systematic revision of the mangrove genus
  ber, Mombasa, Kenya.
                                     Avicennia (Avicenniaceae) in Australasia. Australian System-
                                     atic Botany 4: 299–324.
                                                                    135

Duke N.C. and Jackes B.R. 1987. A systematic revision of the      Kjerfve B., de Lacerda L.D. and Diop E.H.S. 1997. Mangrove
  mangrove genus Sonneratia (Sonneratiaceae) in Australasia.       Ecosystem Studies in Latin America and Africa. UNESCO,
  Blumea 32: 277–302.                           Paris, France.
Farnsworth E.J. and Ellison A.M. 1997. The global conservation     Lugo A.E. and Snedaker S.C. 1974. The ecology of mangroves.
  status of mangroves. Ambio 26: 328–334.                 Annual Review of Ecology and Systematics 5: 39–64.
Foell J., Harrison E. and Stirrat R.L. 1999. Participatory ap-     McCune B. and Mefford M.J. 1997. Multivariate Analysis of Eco-
  proaches to natural resource management – the case of coastal      logical Data. Version 3.09. MjM Software, Gleneden Beach,
  zone management in the Puttalam District. Summary findings        Oregon, USA.
  of DFID-funded research ‘participatory mechanisms for sus-     Mueller-Dombois D. 1968. Ecogeographic analysis of a climate
  tainable development of coastal ecosystems’ (Project R6977),      map of Ceylon with particular reference to vegetation. The
  School of African and Asian studies. University of Sussex,       Ceylon Forester 8: 3–20.
  Falmer, Brighton, UK.                        Murali K.S., Siddappa Setty R., Ganeshaiah K.N. and Uma
Gallin E., Coppejans E. and Beeckman H. 1989. The mangrove         Shaanker R. 1998. Does forest type classification reflect spatial
  vegetation of Gazi bay (Kenya). Bulletin de la Société Royal      dynamics of vegetation? An analysis using GIS techniques.
  Botanique de Belgique 122: 197–207.                   Current Science 75 : 220–227.
Gang P.O. and Agatsiva J.L. 1992. The current status of mangroves    Osborne D.J. and Berjak P. 1997. The making of mangroves: the
  along the Kenyan coast: a case study of Mida Creek mangroves      remarkable pioneering role played by seeds of Avicennia ma-
  based on remote sensing. Hydrobiologia 247: 29–36.           rina. Endeavour 21: 143–147.
Heil G.W. and Van Deursen W.P.A. 1996. Searching for patterns      Pernetta J.C. 1993. Marine protected areas needs in the South
  and processes: modelling of vegetation dynamics with geo-        Asian Seas Region. Sri Lanka. A Marine Conservation and De-
  graphical information systems and remote sensing. Acta Botan-      velopment Report. IUCN, Gland, Switzerland, vii + 67 pp.
  ica Neerlandica 45: 543–556.                    Putz F.E. and Chan H.T. 1986. Tree growth, dynamics, and pro-
Hill M.O. and Gauch H.G. 1980. Detrended correspondence analy-       ductivity in a mature mangrove forest in Malaysia. Forest Ecol-
  sis: an improved ordination technique. Vegetatio 42: 47–58.       ogy and Management 17: 211–230.
Hobbs R.J. and Grace J. 1981. A study of pattern and process in     Ramachandran S., Sundaramoorthy S., Krishnamoorthy R.,
  coastal vegetation using principal component analysis. Vegeta-     Devasenapathy J. and Thanikachalam M. 1998. Application of
  tio 44: 137–153.                            remote sensing and GIS to coastal wetland ecology of Tamil
Holdridge L., Grenke W.C., Hatheway W.H., Liang T. and Tosi J.A.      Nadu and Andaman and Nicobar group of islands with special
  1971. Forest Environment in Tropical Life Zones. Pergamon        reference to mangroves. Current Science 75: 236–244.
  Press, NY, pp. 747.                         Rehder J.B. and Patterson S.G. 1986. Mangrove mapping and mon-
Holmgren P. and Thuresson T. 1998. Satellite remote sensing for       itoring. In: Damen M.C.J., Smit G.S. and Verstappen H.Th.
  forestry planning – a review. Scandinavian Journal of Forest      (eds), Symposium on Remote Sensing for Resources Develop-
  Research 13: 90–110.                          ment and Environmental Management. ITC, Enschede, The
Holmgren P., Thuresson T. and Holm S. 1997. Estimating forest        Netherlands, pp. 495–497.
  characteristics in scanned aerial photographs with respect to    Smith T. and Huston M. 1989. A theory of the spatial and temporal
  requirements for economic forest management planning. Scan-       dynamics of plant communities. Vegetatio 83: 49–69.
  dinavian Journal of Forest Research 12: 189–199.          Spalding M., Blasco F. and Field C. 1997. World Mangrove Atlas.
Hyyppä J., Hyyppä H., Inkinen M., Engdahl M., Linko S. and Zhu       The International Society for Mangrove Ecosystems, Okinawa,
  Y.H. 2000. Accuracy comparison of various remote sensing        Japan, 178 pp.
  data sources in the retrieval of forest stand attributes. Forest  Tack J.F. and Polk P. 1999. The influence of tropical catchments
  Ecology and Management 128: 109–120.                  upon the coastal zone: modelling the links between groundwa-
ITTO 1993. The economic and environmental value of mangrove         ter and mangrove losses in Kenya, India and Florida. In: Harper
  forests and their present state of conservation. Prepared for the    D. and Brown T. (eds), Sustainable management in tropical
  International Tropical Timber Organisation. Japan International     catchments. John Wiley & Sons Ltd, London, UK.
  Association for Mangroves and International Society for Man-    Tassan S. 1987. Evaluation of the potential of the Thematic Map-
  grove Ecosystems, Okinawa, Japan.                    per for marine application. International Journal of Remote
IUCN 1996. Conservation management plan for ten selected man-        Sensing 8: 1455–1478.
  grove sites in southwestern Sri Lanka. IUCN, Gland, Switzer-    Ter Braak C.J.F. 1986. Canonical correspondence analysis: a new
  land, November 1996.                          eigenvector technique for multivariate direct gradient analysis.
Jayatissa L.P., Dahdouh-Guebas F. and Koedam N. 2002. A review       Ecology 67: 1167–1179.
  of the floral composition and distribution of mangroves in Sri    Ter Braak C.J.F. 1994. Canonical community ordination. Part I:
  Lanka. Botanical Journal of the Linnean Society 138: 29–43.       Basic theory and linear methods. Ecoscience 1: 127–140.
Kairo J.G. 2001. Ecology and Restoration of Mangrove Systems in     Tomlinson P.B. 1986. The Botany of Mangroves. Cambridge Uni-
  Kenya. PhD Dissertation, Vrije Universiteit Brussel, Belgium.      versity Press, Cambridge, UK, 419 pp.
Kairo J.G., Gwada P.O., Dahdouh-Guebas F., Koedam N., Van        Ukpong I.E. 1995. An ordination study of mangrove swamp com-
  Speybroeck D. and Ochieng C. Natural regeneration status of       munities in West-Africa. Vegetatio 116: 147–159.
  Mangrove Forests in Mida Creek. (in preparation) (unpub-      Verheyden A., Dahdouh-Guebas F., Thomaes K., De Genst W.,
  lished).                                Hettiarachchi S. and Koedam N. High resolution vegetation
Kent M. and Coker P. 1992. Vegetation Description and Analysis.       data for mangrove research as obtained from aerial photogra-
  A practical Approach. John Wiley & Sons, Chichester, UK.        phy. (submitted).