MARINE ECOLOGY PROGRESS SERIES
    Vol. 272: 77–92, 2004                                          Published May 19
                           Mar Ecol Prog Ser




   Human-impacted mangroves in Gazi (Kenya):
  predicting future vegetation based on retrospective
  remote sensing, social surveys, and tree distribution
F. Dahdouh-Guebas1, 2,*, I. Van Pottelbergh1, J. G. Kairo3,1, S. Cannicci4, N. Koedam1
  1
   Biocomplexity Research Team, Laboratory of General Botany and Nature Management, Mangrove Management Group,
                 Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium
      2
       Uitgegeven met steun van de Universitaire Stichting van België, Egmontstraat 11, 1000 Brussels, Belgium
            3
             Kenya Marine and Fisheries Research Institute, PO Box 81651, Mombasa, Kenya
4
 Dipartimento di Biologia Animale e Genetica ‘Leo Pardi’, Università degli Studi di Firenze, Via Romana 17, 50125 Firenze, Italy




     ABSTRACT: Gazi Bay, Kenya, covers an area of 18 km2, and its mangroves are degraded. We present
     a quantitative and qualitative evaluation of the degradation of vegetation structure and dynamics of
     mangrove communities over a period of 25 yr, using aerial photography in a geographical informa-
     tion system (GIS), combined with ground-truth data for different vegetation layers, and with faunal
     and environmental factors. Retrospective analysis and understanding of current practices were aided
     by interviews with Gazi village elders and by field observations of mangrove tree stumps. GIS-based
     vegetation maps were combined with data obtained using the point-centred quarter method
     (PCQM), an accuracy analysis was performed, and forestry parameters were derived from the resul-
     tant PCQM data. In addition to general human-induced degradation of vegetation structure and
     floristic composition of the seaward mangrove zone, a particular sandy beach is expanding at the
     expense of mangrove, whereas the back mangrove zone has undergone minor changes. Aerial
     photographs of 1992 and current field data show an apparent zonation of 6 different monospecific or
     mixed mangrove communities, with a high importance of Rhizophora mucronata in each community
     and each vegetation layer. Retrospective vegetation structure was combined with correspondence
     analyses on the PCQM data derived for adult, young and juvenile trees in order to make predictions.
     Present dynamics initiated by anthropogenic degradation of mangroves continues, even though
     human impact has diminished. We predict that under a ‘no impact scenario’, the sandy ridge will
     continue to expand, that this will speed up under a ‘mangrove cutting scenario’, and that a scenario
     altering the complex topography will lead to a major re-organisation of the mangrove and terrestrial
     vegetation structure.

     KEY WORDS: Mangroves · Dynamics · Aerial photography · Macrobenthos · Multivariate analysis ·
     Forecasting · Interviews · Kenya
                Resale or republication not permitted without written consent of the publisher




           INTRODUCTION                    frame (Dahdouh-Guebas et al. 2000b). The driving
                                     forces of these dynamics are not known. Therefore,
  Understanding vegetation dynamics is important            and because mangroves are probably fundamentally
for conservation, restoration and sustainable exploita-         different in various ecological settings, an insight into
tion purposes. Mangrove formations, which are often           the processes involved in the dynamic shifts in man-
undervalued and destroyed as a result of various forms         grove vegetation structure is required. There is a lack
of human activity (e.g. Farnsworth & Ellison 1997),           of data on the undisturbed state of mangroves (apart
display major structural changes within a short time-          from observations supplied by local inhabitants). An


*Email: fdahdouh@vub.ac.be                       © Inter-Research 2004 · www.int-res.com
78                      Mar Ecol Prog Ser 272: 77–92, 2004




investigation of the anthropogenic or endogenic/auto-         Retrospective analysis, understanding of current
ecological factors in mangrove dynamics is necessary.       practices and forecasting were also aided by inter-
In Kenya, where the adverse effect of predation on         views with Gazi village elders and by visual observa-
the initial regeneration of mangroves has been docu-        tions of mangrove tree stumps in the field.
mented (Dahdouh-Guebas et al. 1997, 1998), the exper-
imental design of reforestation projects might benefit
from this type of fundamental research. Furthermore,               MATERIALS AND METHODS
it serves as a basis for deciding whether or not human
interference in the form of management or restoration         Study area. In Gazi Bay (4° 26’ S, 39° 30’ E), located
is appropriate. Providing data on the potential of a        about 45 km south of Mombasa, more than 6.15 km2 of
mangrove stand to successfully renew and rejuvenate        mangrove forest has developed (Fig. 1), much of which
can assist predictions (Dahdouh-Guebas 2002).           is subject to severe human impact (e.g. Beeckman et
  The first objective of this study was to make a quan-      al. 1989, Gallin et al. 1989, Vanhove et al. 1992, Kairo
titative and qualitative evaluation of the past and        1995a, Schrijvers et al. 1995, Fondo & Martens 1998,
present vegetation structure dynamics in a mangrove        Aloo 2000, Dahdouh-Guebas et al. 2000a, Hoorweg et
forest in Gazi Bay, Kenya, emphasizing and confirm-        al. 2000, Abuodha & Kairo 2001, Kairo et al. 2001). This
ing the anthropogenic degradation that has taken          study concentrates on a mangrove stretch adjacent to
place in the past. The study was aimed at interpreting       the village of Gazi (see Fig. 3), one of the most affected
past and present mangrove vegetation structure using        areas. Kidogoweni and Mkurumuji, 2 seasonal rivers,
aerial photography analysis and field studies of the        provide most of the overland freshwater input into
adult vegetation. A second objective was to extra-         the bay, but tidal influences (spring tidal amplitude
polate the interpretation to future dynamics through        is about 3.5 to 4 m) are more important. All 10 East-
fieldwork on young and juvenile vegetation layers.         African mangrove species, i.e. Avicennia marina (Forsk.)
The study combines remote sensing data and ordina-         Vierh., Bruguiera gymnorrhiza (L.) Lam., Ceriops tagal
tion analysis to make predictions based on vegetation       (Perr.) C.B. Robinson, Heritiera littoralis Dryand., Lum-
history.                              nitzera racemosa Willd., Pemphis acidula Forst., Rhizo-




Fig. 1. (a) Kenya. (b) Gazi Bay, showing areas of mangroves, seagrasses and coral reef. (c) Location of study site in Gazi Bay and
approximate extent of aerial photograph of 1972 (dashed rectangle) and that of 1992 (dotted rectangle) (see Fig. 3). (For detailed
                     map of coast see Dahdouh-Guebas et al. 2002b)
                             Dahdouh-Guebas et al.: Human-impacted mangroves                       79




phora mucronata Lam., Sonneratia alba Sm., Xylocar-                       1:12 500 respectively, from the Kenyan Survey Depart-
pus granatum Koen. and X. moluccensis (Lamk.) Roem.                       ment. These were photographically enlarged (Kodak
(nomenclature according to Tomlinson 1986, Duke &                        5052 TMX) using a 50 mm macro lens, then scanned at
Jackes 1987, Duke 1991), occur along the Kenyan                         300 dots per inch (dpi), and digitised in a geographic
coast, and their presence has been reported in Gazi                       information system (GIS), using MapInfo software
Bay. X. moluccensis was, however, not observed                         on a Macintosh computer platform to outline polygon
during the course of this study.                                vectors corresponding to mangrove zones. In this study
 Climatic data are available for Mombasa (Fig. 2), and                     the term ‘zonation’ is defined as ‘banding of vegeta-
the data list an average annual rainfall of 1136.47 mm                     tion types with a certain floristic composition, usually
(period 1890–2001) and an average annual tempera-                        strongly dominated by a single species’. The identifi-
ture of 26.15°C (period 1931–2001). The rainfall trend                     cation of vegetation assemblages was based on the
during the study period (1997–1999) was not different                      image attributes of tonality, texture and structure
from the average, except for the period October 1997                      (Dahdouh-Guebas et al. 2001) as successfully applied
to February 1998, which included an El Niño event and                      to the identification of mangroves in the framework of
had considerably more rainfall than the usual pattern                      mangrove vegetation dynamics research in the past
(October 826 mm, November 317 mm, December 285                         (Dahdouh-Guebas et al. 2000b, Verheyden et al. 2002).
mm, January 213 mm, February 50 mm). Apart from an                       The aerial photographs were not co-registered and
early study (Jeathold & Smidt 1976), no detailed cli-                      overlayed for a number of reasons. First, there was a
matic data were available for the specific study area.                     lack of reliable ground control-points common to both
However, the rainfall pattern in Gazi Bay is the same                      aerial photographs to enable effective co-registering of
as in Mombasa (Teathold & Smidt 1976) with an annual                      both photographs. Second, an overlay of the 2 photo-
average that may be slightly higher in Gazi due to the                     graphs, with their many vegetation assemblages in a
direction of the trade-winds (south-east and north-                       small area, resulted in an excessive amount of vegeta-
east), the differences in relief, and hence a less than                     tion classes that obscured interpretation (Dahdouh-
perfect east-west gradient in the rainfall (Hoorweg et                     Guebas et al. 2000b); it was thus far more appropriate
al. 2000).                                           to be able to compare and interpret the original vege-
 Photogrammetry. We obtained 2 aerial photographs,                       tation maps from the 2 photographs separately. Third,
one for 1972, the other for 1992, at scales of 1:7500 and                    the overlay of the ground-truth transects on the most
                                                recent photograph was more accurate using visual
                 Mombasa (Kenya)             400          cues such as trees than using the Global Positioning
                (04° 00’ S, 039° 36’ E)                     System (GPS) (Dahdouh-Guebas et al. 2000b, Dah-
                                     300
                                                douh-Guebas 2002). However, in order to extract some
                                     200          surface data with respect to the vegetation assem-
          50                           100          blages, 5 GPS points with an average accuracy of 10.5 m
                                                were relied upon to georeference the photographs.
                                     90
                                                The 1992 photograph was ground-truthed in 1997 as a
                                        Rainfall (mm)




          40                           80          result of the delayed availability of the photographs,
Temperature (°C)




                                     70          and for practical and logistic considerations. Incorpora-
                                     60
                                                tion of this time-gap of 5 yr was based on the absence
          30
                                                of major visual differences in areal extent or floristic
                                     50
                                                composition of vegetation assemblages in the field
          20                           40          (pers. obs. by F.D.G. & J.G.K. on a semestrial basis
                                     30          since 1992), and the absence of substantial differences
                                                from the zonation in 1993 (cf. Dahdouh-Guebas et al.
          10                           20
                                                2002b, in press). Nevertheless, the results of this study
                                     10          were interpreted with careful consideration of this time
                                                gap. Ground-truthing consisted of visits to all the man-
            J  A  S  O  N  D  J  F  M  A  M  J             grove assemblages outlined in the digitisation of the
                        Month                      photograph as well as land-creek transects through the
Fig. 2. Climate diagram for Mombasa. Monthly rainfall data                   mangrove vegetation.
(upper continuous line) are from Lieth et al. (1999) for period                  Field survey. The field survey served to determine
1890–1985 complemented with data from Meteorological                      the vegetation structure in different layers and to
Department in Mombasa for period 1986–2001. Monthly tem-                    investigate a selection of environmental factors in the
perature data (lower continuous line) are from Lieth et al.
(1999) for period 1931–1990 complemented with data from
                                                study area (see below). During the a first field survey
Meteorological Department in Mombasa for period 1991–2001                    in July to August 1997 (wet season), 6 transects (5 par-
80                    Mar Ecol Prog Ser 272: 77–92, 2004




allel transects and 1 orthogonal) were surveyed in the      tis 1956) as described by Cintrón & Schaeffer Novelli
most anthropogenically degraded part of our study site      (1984). The individual tree closest to the sample point
(Fig. 3). The orthogonal transect was covered in order      was identified and recorded in each of 4 quadrants,
to corroborate a zonation pattern observed on the        and its height and diameter, d130 (130 cm height sensu
image and in the other transects, but it did not provide     Brokaw & Thompson [2000], formerly referred to as
additional information and is therefore not shown in       DBH, the diameter at breast height) were measured.
the relevant figures. Its vegetation data were however      The stem diameter of mangrove species with aerial
used in the multivariate analyses (see later subsection).    roots at 130 cm height was measured 30 cm above the
Sampling at 10 m intervals was undertaken using the       highest roots. Other anomalies to the application of
point-centred quarter method, PCQM (Cottam & Cur-        the PCQM (e.g. forking stems) were dealt with as




                                                new fishermen’s place




                                                Fig. 3. Aerial photographs
                                                of study site in (a) 1972
                                                and (b) 1992. The latter
                                                shows important local fea-
                                                tures and approximate ex-
                                                tent of 1972 photograph
                                                (dashed rectangle). In (b),
                                                the 2 arrows pointing left
                                                (centre of photo) and that
                                                pointing right (upper right)
                                                indicate approximate lo-
                                                cations of photographs in
   old fishermen’s place        Gazi Mosque    historic palace/slave house      Fig. 5a,b,c respectively
                  Dahdouh-Guebas et al.: Human-impacted mangroves                 81




described by Cintrón & Schaeffer Novelli (1984). The     duration of the fieldwork, in order to complement the
total vegetation cover was estimated in percentages in    retrospective data. Apart from historic aerial photo-
the 5 m × 5 m quadrat nearest to the sampling point (the   graphs and mangrove tree stumps, village elders are
sample point thus formed the common corner point of     the only source of information on the past extent and
the 4 quadrats located in the 4 quadrants). Each tran-    composition of mangroves. The complementation of
sect was sampled for vegetation description within 3 d    basic scientific data with such sociological information
to locate and describe mangrove juveniles (propagules    is an added value for research (e.g. Cormier-Salem
or seeds up to the stage of sapling), young mangrove     1999). Free and open-ended interviews on the history
trees (trees smaller than 1.3 m or with a d130 < 2.5 cm,   of the mangrove vegetation in Gazi (emphasising
but which had reached the sapling stage, a plant with    changes in area and species composition) were carried
more than 6 leaves), and adult mangrove trees. These     out in Kiswahili at 11 different families in the village.
3 stages represent different phases critical to the sur-   At the time of the survey there were a total of 119
vival of mangroves in this area: (1) mangrove juveniles   houses in Gazi (Vandeput 1999), and on average 1
that are subject to potential stressful conditions in the  family occupies one house (pers. obs. by F.D.G. &
pre- and post-dispersal development of the plant, par-    I.V.P.). The age of the people interviewed was > 40 yr,
ticularly propagule predation, which is known to be a    a period extending retrospectively well beyond the
significant constraint in Kenya (Dahdouh-Guebas et      date of the first aerial photograph (1972).
al. 1997, 1998); (2) young mangrove trees, which have     Data analysis. Vegetation maps of 1972 and 1992
survived predation disturbance but are still vulnerable   were drawn up and the field transects were traced onto
to other disturbance and stress factors such as drought,   the most recent aerial photograph (1992) using visual
water movement, erosion, sedimentation, etc.; (3) adult   cues. The data originating from the PCQM, usually
mangrove trees, plants that are well established and     used to calculate forest structural parameters only,
are likely to survive, provided no devastating natural    were here transformed into visual data that allow their
or human impacts occur. Only the adult vegetation      use in a GIS. PCQM transect data for adult, young and
could be used as a link to the aerial photographs as it   juvenile mangrove species were visualised in separate
represented the remotely sensed canopy. The diameter     GIS layers as squares representing the 4 quadrants
of young mangrove trees and juveniles was not mea-      superimposed on the vegetation map of 1992. The ratio
sured as they were too small to be relevant in this study.  of PCQM sample-points occupied by the adult man-
  A second field study was organised in March 1999     grove species dominant for the remotely sensed vege-
(dry season) in order to detect possible seasonal      tation class in which the sample points are located, and
changes in the distribution and abundance of man-      the sample points occupied by a non-dominant tree,
grove juveniles. We were unable to sample a represen-    was calculated as a map-accuracy measure. A similar
tative dry season in 1998 due to the El-Niño event      measure for the map of 1972 was unfortunately not
at the transition of 1997 to 1998 (mentioned earlier).    possible as no field data were available.
Along 2 additional transects (Transects A and B), cross-    From the PCQM data relative density (DEr = [number
ing the complete vegetation belt (including terrestrial   of individuals of species total number of individuals] ×
vegetation) from the village to the creek, mangrove     100), relative dominance (DOr = [dominance of a
species were recorded to confirm the land-sea zona-     species dominance for all species] × 100) and relative
tion and to detect any zone not visible from the aerial   frequency (FRr = [frequency of a species sum of fre-
photograph (particularly species encroaching on the     quencies of all species] × 100) were computed for the
terrestrial vegetation). Environmental factors such as    adult trees using the methods described by Cintrón &
salinity (determined with an Atago refractometer with    Schaeffer Novelli (1984) in order to calculate the im-
a precision of 1 ‰, over a scale from 0 to 100 ‰), light   portance value (IV) of Curtis (1959) for each mangrove
intensity (determined with a Lutron lux-meter with a     tree species. According to this method, relative domi-
precision of 4% of the resolution scale, which ranged    nance is based on tree diameter. Relative density and
from 1 to 100 lux), herbivorous crab abundance and      relative frequency were also calculated for young and
snail abundance (by burrow and individual counts in     juvenile mangrove trees. The IV for these vegetation
1 m2 quadrats in each of the 4 PCQM quadrants) were     layers was calculated as for the adult tree layers, but
also recorded. These 4 environmental factors were      omitting relative dominance.
selected in particular because they have been reported     Statistical analyses. A G-test (Sokal & Rohlf 1981)
by univariate studies to play a significant role in the   was performed on the absolute numbers of sample
establishment of vegetation structure (Smith 1987, 1992,   points located in a certain vegetation class or zone con-
Dahdouh-Guebas et al. 1998, Matthijs et al. 1999).      taining a certain mangrove species in order to detect
  Visual observations and interviews with local people   differences in abundance between the adult and
were carried out whenever possible over the whole      young vegetation layers at a particular location; this
82                     Mar Ecol Prog Ser 272: 77–92, 2004




may reflect a static or spatially dynamic nature of      multivariate context, and to determine the extent of
regeneration. It was also used together with a χ2-test     their role. The environmental variables and para-
to test for significant seasonal differences in absolute    meters used in the CCA were salinity (cf. Matthijs et al.
juvenile densities. Seasonal differences in juvenile      1999), light intensity (cf. Smith 1992), herbivorous crab
distribution are important when comparing their distri-    abundance and snail abundance (cf. Dahdouh-Guebas
bution patterns with that of adult or young trees,       et al. 1998). Monte-Carlo tests (randomisation tests)
the distribution of which are not seasonally variable.     were used to test the significance of the eigenvalues
Sampling in a single season may therefore give an       and the species-environment correlations.
erroneous idea of the dispersal potential of mangrove
juveniles. Nevertheless, we emphasise a priori that
seasonal differences between the wet and dry season                   RESULTS
recorded in this study may still have been subject to
annual variability, which can only be determined by           Aerial photography and field surveys
long-term, monthly sampling. Following the distribu-
tion of juveniles is important for understanding the       Fig. 4 shows the vegetation maps of 1972 and 1992
ability of propagules to reach a certain area, and to     and shows that the total mangrove vegetation area
reveal hazards such as propagule predation.          decreased during this period (see also Table 1) with-
  To investigate the similarity in distribution between    out, however affecting the relative floristic composi-
the species in the different vegetation layers and to     tion. The unvegetated sandy area in the upper right
explore the links between vegetation and environmen-      of the vegetation map of 1972 (Fig. 4a) corresponds to
tal factors (salinity, light intensity, propagule predator   bare sand and low beach vegetation that once formed
abundances), species ordinations (Kent & Coker 1992)      a dense forest with the surrounding mangrove, as
were performed on the adult, young and juvenile man-      reported by the interviewees. In fact, locals marked
grove distributional data. It is important to emphasise    this very area as the starting point for mangrove cut-
that, contrary to classical statistical analyses (e.g. rank  ting in 1964 because it was the thinnest and most easily
correlation tests), ordinations are the only way to effec-   accessible area in the forest, and in addition could be
tively analyse large data matrices with information on     used to spot incoming ‘dhows’ (traditional trade boats).
samples versus species (or in this case species-layers),    Clusters of remaining mangrove stumps with clear
to cope with multiple absences (many matrix cells with     signs of cutting in the field as well as the interviews
zero values), and to indirectly or directly integrate     revealed that in general Rhizophora mucronata, Ceriops
environmental factors as explanatory factors (Dah-       tagal and Bruguiera gymnorrhiza are the preferred
douh-Guebas et al. 2002a).                   species for cutting. After 1972, the sandy area
  The samples-species input matrices were generated      expanded at the expense of the mangrove stand (see
based on presence/absence data (0 or 1) and on abun-      Fig. 4b). This was still on-going in 1999, with visible
dance data (0 to 4) for each of the 99 sample points,     effects to trees adjacent to the sandy area in the field
whereby each sample in the matrix included the data      between the surveys of 1997 and 1999, as evidenced
from the 4 original PCQM quadrants. Using PC ORD
for Windows (McCune & Mefford 1997), detrended         Table 1. Changes in areal extension (area in m2) between
correspondence analysis (DCA) (Hill & Gauch 1980) and     1972 and 1992 for some land-cover classes and mangrove
canonical correspondence analysis (CCA) (Ter Braak       vegetation assemblages identified from aerial photographs.
                                Georeferencing was based on 5 GPS points with an average
1986, 1994) were applied and the percentage of vari-      accuracy of 10.5 m. The increase in some of the mangrove
ance in the matrix that is explained by each axis was     vegetation assemblages was due to the existence of a large
calculated. Although this combined ordination approach     ‘mixed’ zone where no particular species dominated in 1972
was suggested by Ter Braak (1995) to provide an
added value with respect to the choice of environmen-      Assemblage/Class     1972   1992  Change (%)
tal variables, the DCA and CCA essentially served
different aims here. DCA was used to interpret the        Avicennia marina     23310  9292    –60.14
                                 Ceriops tagal      4579   3118    –31.91
similarity in the distribution of age classes (adult,      Rhizophora mucronata   1897   2304    + 21.45
young and juvenile trees) within and among species.       Sonneratia alba      0   2652   +100.000
In contrast, CCA was used to test the hypothesis that a     Mixed mangrove      15454  4613    –70.15
selection of environmental factors identified in earlier     Total mangrove area   45239  21978   –51.42
studies as playing a major role in or as determining the     Cocos nucifera plantation
establishment of mangrove zonation patterns (based        adjacent to old fishermen’s
                                 place (Fig. 3)      6377  915    –85.65
on univariate studies), are indeed responsible for the
                                 Unvegetated (sandy)    2078   7057   + 239.61
observed vegetation structure when analysed in a
                 Dahdouh-Guebas et al.: Human-impacted mangroves                   83




by the presence of entire (i.e. not cut) dead adult trees  that exists in all seaward or creekward areas, and the
at the margins of the sandy stretch (Fig. 5a,b). The    multiple topographic ridges and depressions present in
Sonneratia alba assemblage in the upper right corner    more landward zones. Roughly these 2 areas are sepa-
of the 1992 vegetation map (Fig. 4b) is another remain-   rated by a series of sandy ridges running parallel to the
der of the originally larger seaward mangrove belt. It is  creek (Fig. 4b). The Avicennia marina fringe on the
physically unable to stop the high-energy effects of    landward side is reduced in area (Fig. 4; Table 1).
waves, which penetrate the S. alba assemblage, fall     However, within the mixed zone, individual A. marina
onto the beach, and are eroding terrestrial coconut     trees present in 1972 (Fig. 4a) have formed a pure
plantations (Fig. 5c).                   patch (see photograph and vegetation map of 1992:
 During the ground-truth surveys, a difference was     Figs. 3b & 4b). Besides the seaward A. marina assem-
noted between the regular slope of the intertidal zone   blage, within the remainder of the mangrove forest




                                            Fig. 4. Vegetation maps of study
                                            site in (a) 1972 and (b) 1992. Ap-
                                            proximate orientation of photo-
                                            graphs with respect to each other
                                            is indicated by dashed rectangle.
                                            Long black lines: Transects A
                                            and B; black rectangle: locations
                                            of 5 transects shown in Fig. 6. As
                                            shown in the key above, a dotted
                                            pattern represents an area with an
                                            open character, a hatched pattern
                                            an area that is degraded and a
                                            pattern with small lines the area
                                            outside our study field, all of which
                                             are independent of the colour
84                       Mar Ecol Prog Ser 272: 77–92, 2004




   (a)




Fig. 5. (a) (b) Margins of expanding sandy area indicated by arrows in middle left of Fig. 3b; arrows indicate entire Rhizophora
mucronata trees that died naturally; in the foreground and background are also stumps of the same trees chopped down by
villagers. (c) Degraded Sonneratia alba assemblage at low tide (left) in 1992, and erosive effects on terrestrial coconut plantations
      (right) in location indicated by arrow on top right of Fig. 3b. Here, coconut trees are uprooted by wave action



the landward Ceriops tagal assemblage of 1972 seems         tide, zonation is less evident and species composition
to be the only other assemblage that is expanding          seems restricted initially to C. tagal, A. marina and
(Fig. 4). However, since the seaward C. tagal assem-         Lumnitzera racemosa and more landwards to A.
blage of 1972 has disappeared, this is not reflected in       marina and L. racemosa as was evident from Transects
the total surface covered by C. tagal (Table 1).           A and B (not shown). In these areas other species are
                                   occasionally present, but R. mucronata, which is often
                                   found near regularly flooded, small creeklets (as small
Mangrove zonation — horizontal vegetation structure         as 1 m wide), is almost totally absent. Within this land-
                                   ward area the border between the vegetation zone
 The forest section studied in the seaward transects,        dominated by R. mucronata and that dominated by C.
which gradually slopes down towards the creek, is          tagal corresponds with the high water limit of neap
(partially) zoned from land to sea with a landward          tide (F.D.G. pers. obs.).
Avicennia marina-dominated zone, an A. marina /Bru-           Although Lumnitzera racemosa was occasionally
guiera gymnorrhiza /Ceriops tagal /Rhizophora mucro-         observed along Transects A and B, these transects did
nata mixed zone, a R. mucronata-dominated and/or           not contain zones or patches other than those observed
seaward A. marina-dominated zone and finally a Son-         from the aerial photographs, and were therefore not
neratia alba-dominated zone (Fig. 6). However, in the        shown. The most landward species was Avicennia
more landward areas, where topography is irregular          marina, but some local inhabitants recall Rhizophora
with many small emergent areas even at spring high          mucronata as the first mangrove tree species encoun-
                   Dahdouh-Guebas et al.: Human-impacted mangroves                    85




                                          tered on the way from the village to the
                                          fishermen’s new place (Fig. 3). Locals
                                          identified R. mucronata as the species
                                          preferred for house building, with A.
                                          marina being less sought after.


                                           Mangrove stratification — vertical
                                              vegetation structure

                                           The quantified PCQM vegetation data
                                          from the zoned forest section are given
                                          in Tables 2 & 3. It can be seen that Rhizo-
                                          phora mucronata is ubiquitously im-
                                          portant species in all vegetation classes
                                          identified on the aerial photograph and
                                          in all vegetation layers (see Fig. 6a). In
                                          the landward assemblages (R. mucro-
                                          nata, mixed mangrove) the understorey
                                          is well represented, particularly by Ceri-
                                          ops tagal, as can be seen in the overlay
                                          of young trees (Fig. 6b) and the distribu-
                                          tion of juveniles (Fig. 7). The absolute
                                          densities of the juveniles differed sig-
                                          nificantly between the 2 seasons over
                                          the entire study site (χ2 = 25.526; df = 2;
                                          p < 0.001) and for the different forest
                                          patches, except for the Sonneratia alba
                                          zone (G-test = 4.194; df = 2; not sig-
                                          nificant). We repeat that seasonal
                                          differences or correspondences may still
                                          have been subject to annual variability.
                                           The results of the G-test (Table 4) show
                                          that the observed differences between the
                                          abundance of adult and young trees in
Fig. 6. Close-up of overlay of visualised point-centred quarter method (PCQM)   each vegetation class were not significant,
data from 5 parallel transects on vegetation map of 1992 for (a) adult and (b)   except for the most seaward zones (Avi-
young mangrove trees. The sixth orthogonal (overlapping) transect is not shown.  cennia marina and Sonneratia alba).
Each block of 4 squares represents 4 PCQM quadrants with the PCQM sample
point in the centre; 1 block represents 5 m × 5 m. Each colour represents a dif-
                                           Outliers do not dominate the DCA
ferent tree species, with the tree nearest to the sample point being considered  ordination graph and species clusters are
dominant (for detailed PCQM description see Cintrón & Schaeffer Novelli 1984)   easily distinguishable (Fig. 7). Although it
Table 2. Relative density (DEr), relative dominance (DOr), relative frequency (FRr) and importance value (IV: Curtis 1959) for the major vegetation classes identified
                                                                                                 86


                                    on the aerial photograph

 Vegetation class     Avicennia marina zone Rhizophora mucronata zone  Sonneratia alba zone                       Mixed zone     Sparsely vegetated (sandy)
              DEr DOr FRr IV Rank   DEr DOr FRr IV Rank   DEr DOr FRr IV Rank                      DEr DOr FRr IV Rank   DEr DOr FRr IV Rank

 Adult trees
 Avicennia marina     15.8 65.5 20.8 1020 2                   9.3 14.7 13.6 38   3  12.0 5.8   13.5 31   3  3.9 7.9 7.5 19   4  10.7 30.3 18.2 59   2
 Bruguiera gymnorrhiza   3.0 0.2 5.7   9 5                   0.0 0.0 0.0   0  5  0.0 0.0   0.0  0  4  3.9 0.7 5.7 10   5   0.0 0.0 0.0   0   5
 Ceriops tagal      15.8 5.8 18.9 41  3                   5.3 0.7 6.8 13    4  0.0 0.0   0.0  0  4  22.3 7.3 24.5 54  3   7.1 9.6 9.1 26    4
 Rhizophora mucronata   58.4 21.0 43.4 1230 1                  69.3 10.3 54.5 1340  1  21.7 5.0   32.4 59   2  66.0 29.2 56.6 1520 1  75.0 46.9 63.6 1850  1
 Sonneratia alba      6.9 7.5 11.3 26  4                  16.0 74.3 22.7 1130  2  66.3 89.20  48.6 2040  1  3.9 54.9 5.7 64  2   7.1 13.2 9.1 29    3
 Nil                 0.0                           2.3             5.4            0.0           9.1
 Young trees
 Avicennia marina     1.1    1.9  3 4                  3.9   6.3 10 4     9.0     15.2 24 2    2.0   3.5  5 3   7.4    8.3 16   2
 Bruguiera gymnorrhiza   3.2    5.7  9 3                  1.3   3.1  4 5    0.0     0.0  0 4    2.0   3.5  5 3   0.0    0.0  0   3
 Ceriops tagal      21.3    26.4 48  2                  5.3   6.3 12 3     0.0     0.0  0 4   34.7   33.3 68  2   0.0    0.0  0   3
 Rhizophora mucronata   73.4    54.7 1280 1                  85.5   75.0 1610 1   80.8     63.6 1440 1   59.4   54.4 1140 1   92.6   75.0 1680  1
 Sonneratia alba      1.1    1.9  3 4                  3.9   9.4 13 2    10.3     12.1 22 3    2.0   3.5  5 3   0.0    0.0  0   3
 Nil                 9.4                          0.0              9.1           1.8           16.7
 Juvenile trees 1997
 Avicennia marina     25.8    16.9    43              3  12.2   17.8 30 2     0.1     4.8  5 3    0.9   9.7  11  3  42.8   10.5  53  2
 Bruguiera gymnorrhiza   0.0    2.6     3              5  0.0   4.4  4 4    0.1     2.4  2 5    0.0   1.4  1  4  0.0    0.0  0  4
 Ceriops tagal      26.3    19.5    46              2  3.6   13.3 17 3     0.1     4.8  5 3   44.3   29.2  73  2  3.8   21.1  25  3
 Rhizophora mucronata   47.8    42.9    91              1  84.1   48.9 1330 1   98.7     35.7 1340 1   54.7   41.7  96  1  53.4   42.1  96  1
 Sonneratia alba      0.1    5.2     5              4  0.0   0.0  0 5    0.9     7.1  8 2    0.0   1.4  1  4  0.0    0.0  0  4
 Nil                 13.0                          15.6             45.2           16.7           26.3
 Juvenile trees 1999
 Avicennia marina     64.1    30.8    95              1  30.0   15.6 46 2     5.4     12.8 18 2    39.5   21.6  61  2  83.1   22.2 1050  1
 Bruguiera gymnorrhiza   0.0    1.5     2              5  0.0   2.2  2 4    0.1     4.3  4 4    0.0   1.4  1  4  0.0    0.0  0   4
 Ceriops tagal       1.5    18.5    20              3  1.4   20.0 21 3     0.8     6.4  7 3   14.6   25.7  40  3  0.4   11.1 12   3
 Rhizophora mucronata   34.4    41.5    76              2  68.6   46.7 1150 1   93.8     29.8 1240 1   45.9   36.5  82  1  16.5   22.2 39   2
 Sonneratia alba      0.0    1.5     2              4  0.0   0.0  0 5    0.0     4.3  4 5    0.0   0.0  0  5  0.0    0.0  0   4
                                                                                                 Mar Ecol Prog Ser 272: 77–92, 2004




 Nil                 6.2                          15.6             42.6           14.9           44.4




(Table 5).
                                          from the matrices.




                        structure and environment
                       Relationships between vegetation
                                          elongated position along the first axis




                                          have the widest spread along the first
                                          Bruguiera gymnorrhiza, Ceriops tagal




                                          ability of A. marina to widely disperse




surements show an expected lower
                                          nile distributional data are omitted
                                          in a bimodal pattern, forming land-




value for densely forested zones
sis. The 1999 light intensity mea-
fore averaged between seasons




rain) and were omitted from analy-
surements in 1997 suffered under
when entered in the multivariate
                                          not change significantly when juve-
                                          apart. This trend is less obvious for
                                          whereas those for juveniles lie
                                          viduals tend to lie closer together,
                                          for these species clusters, the data
                                          and R. mucronata) and, particularly
                                          proximately 20 Units for C. tagal
                                          axis (> 40 Units compared to ap-
                                          marina, B. gymnorrhiza and S. alba
                                          species clusters represented by A.
                                          values for the first and second axis
                                          field survey. A. marina is distributed
                                          ing to ‘zones’ from land to sea in the
                                          Sonneratia alba cluster, correspond-
                                          marina – R. mucronata cluster and a
                                          and Rhizophora mucronata), an A.
                                          mixed cluster (Avicennia marina,




                                          the other species. The variance does
                                          deduced from Fig. 7, which shows a
                                          to confirm zonation, the latter can be




erratic weather conditions (clouds,
analysis matrix. Light-intensity mea-
(<1.25 ‰, Table 5), which was there-
little seasonal variability for salinity
  The environmental variables show
                                          points for adult and young indi-
                                          axes are 44 and 8% respectively. The
                                          zonation pattern, together with the
                                          ward and the most seawards zone




                                          the variance explained by the same
                                          are 0.404 and 0.194 respectively and
                                          in the ordination plot. The eigen-
                                          its seeds, contributes to its central
                                          (seaward together with S. alba). This
                                          was not the purpose of this ordination
                  Dahdouh-Guebas et al.: Human-impacted mangroves                              87




Table 3. Relative density (DEr), relative dominance (DOr), relative frequency (FRr) and importance value (IV) for landward
            and seaward A. marina fringe separately and for all vegetation classes together


Vegetation class      Landward A. marina zone       Seaward A. marina zone               All classes
              DEr DOr FRr IV Rank         DEr DOr FRr IV Rank              DEr DOr FRr IV Rank

Adult trees
Avicennia marina      23.5 73.9 19.0 116    1     13.2 28.8 21.9 64          2      10.3 16.5 14.5 41     3
Bruguiera gymnorrhiza    5.9 0.4 9.5 16     4     1.5 0.1 3.1   5         4      1.8 0.1 6.1   8     5
Ceriops tagal       47.1 19.4 47.6 114    2     0.0 0.0 0.0   0         5      11.5 2.0 11.7 25      4
Rhizophora mucronata    20.6 5.6 19.0 45     3     76.5 51.3 59.4 187         1      55.9 11.6 44.7 112     1
Sonneratia alba       2.9 0.7 4.8   8    5     8.8 19.8 15.6 44          3      20.5 69.9 21.2 112     2
Nil                  0.0                0.0                       1.7
Young trees
Avicennia marina      0.0    0.0   0  4     1.6        3.3  5     3      4.0    7.6 12     3
Bruguiera gymnorrhiza    9.7    13.0  23  3     0.0        0.0  0     4      1.6    4.7  6    5
Ceriops tagal       51.6    43.5  95  1     6.3        13.3 20     2      15.7    18.8 35     2
Rhizophora mucronata    38.7    34.8  73  2     92.1        73.3 165     1      75.0    55.3 130    1
Sonneratia alba       0.0    0.0   0  4     0.0        0.0  0     4      3.7    7.1 11     4
Nil                  8.7                    10.0                   6.5
Juvenile trees 1997
Avicennia marina      43.0    25.0  68  2     0.2        7.1  7     4      68.3    10.8   79  1
Bruguiera gymnorrhiza    0.0    0.0   0  4     0.1        4.8  5     5      0.0    2.7    3  5
Ceriops tagal       44.7    34.4  79  1     0.8        9.5 10     2      7.5    18.8   26  3
Rhizophora mucronata    12.3    37.5  50  3     98.7        50.0 149     1      24.2    39.9   64  2
Sonneratia alba       0.0    0.0   0  4     0.3        9.5 10     3      0.0    4.0    4  4
Nil                  3.1                    19.0                  23.8
Juvenile trees 1999
Avicennia marina      87.5    34.8 122   1     42.7        26.2 69     2      64.6    21.7   86  1
Bruguiera gymnorrhiza    0.0    0.0  0   4     0.0        2.4  2     5      0.0    2.4    2  4
Ceriops tagal        5.4    26.1 31    3     0.2        14.3 14     3      2.4    18.9   21  3
Rhizophora mucronata    7.1    30.4 38    2     57.1        50.0 107     1      32.9    36.3   69  2
Sonneratia alba       0.0    0.0  0   4     0.0        2.4  2     4      0.0    1.4    1  5
Nil                  8.7                    4.8                  19.3



  In the direct ordination (CCA) the environmental                    Detrended correspondence analysis
variables and parameters used failed to explain most of
the observed variation in the vegetation data of the          80                       AmYT
adult, young and juvenile mangroves. However, when                                    SaJT99
the juvenile vegetation data are omitted from the                         BgJT99    AmAT
                                        BgYT
matrices, the remaining adult and young distributional         60     BgAT
                                                  AmJT99
data clearly separate according to species, and to a cer-                AmJT97
tain extent according to environmental factors (Fig. 8).                      BgJT97                   SaYT
                               Axis 2




Adult and young Avicennia marina (AmAT and AmYT)                CtAT
                                                                  SaAT
                                    40                RmJT97
and adult Sonneratia alba (SaAT) are weakly nega-
tively correlated with the first axis, whereas young S.              CtYT  CtJT99         RmYT
alba (SaYT) and Ceriops tagal (CtYT) are strongly pos-                           RmJT99    SaJT97
itively correlated with it (Fig. 8). Adult and young          20
Bruguiera gymnorrhiza (BgAT and BgYT) are strongly                   CtJT97

                                                       RmAT
Table 4. Results of G-test for differences in species pro-       0
portions of adult tree and young tree individuals in each          0             40           80
       vegetation class. ns: not significant                               Axis 1

Vegetation zone        G     df       p       Fig. 7. Results of indirect species ordination (DCA) of pres-
                                    ence/absence vegetation data for 99 sample points along 6
Avicennia marina      30.363    5    < 0.001      transects (each sample point consisted of 4 measurements,
Rhizophora mucronata    10.250    5      ns       1 in each quadrant) for adult (AT), young (YT) and juvenile
Sonneratia alba       66.830    3    < 0.001      (JT) trees recorded during field field study in 1997 (97) or in
Mixed            7.089    5      ns       1999 (99). Am: Avicennia marina; Bg: Bruguiera gym-
Unvegetated (sand)      6.822    4      ns       norrhiza; Ct: Ceriops tagal; Rm: Rhizophora mucronata;
                                               Sa: Sonneratia alba
   88                               Mar Ecol Prog Ser 272: 77–92, 2004




                   Table 5. Mean (± SD) for the environmental factors in each of the vegetation zones

     Vegetation zone                      Salinity             Light 1999        Abundance
                              1997        1999          (lux)    Crab (burrows m–2) Snails (ind. m–2)

     Avicennia marina                35.24 ± 2.39  34.11 ± 2.26         3480.00      5.00 ± 4.17    0.73 ± 1.58
     Rhizophora mucronata              34.29 ± 1.86  33.06 ± 1.00         5905.00      5.39 ± 4.65    1.83 ± 2.64
     Sonneratia alba                 33.47 ± 2.37  33.00 ± 0.85        23440.00      3.17 ± 4.36    1.57 ± 3.09
     Mixed                      35.38 ± 3.30  34.73 ± 1.98         5880.00      5.64 ± 3.84    4.52 ± 5.14
     Sparsely vegetated (sandy area)         34.00 ± 1.87  35.00 ± 1.00        38415.00      4.00 ± 8.94    0.20 ± 0.45



   positively correlated with the second axis (Fig. 8). The                  2.8% for the first 2 axes respectively, which means that
   canonical coefficients are highest for snail abundance                   even though some differences are significant, other
   with respect to the first axis (0.639) and for light condi-                environmental factors contribute more to the observed
   tions with respect to the second axis (–0.944). These                   vegetation structure.
   canonical coefficients are conceptually similar to the
   usual regression coefficients and represent the unique
   contribution of individual variables as opposed to the                       DISCUSSION AND CONCLUSIONS
   simple correlation coefficient between a variable and
   an ordination axis. The correlation coefficients that                       Aerial photography and field surveys
   correspond to the abovementioned canonical coeffi-
   cients are 0.703 for snail abundance with respect to the                   Analysis of aerial photographs revealed a decrease
   first axis and –0.727 for light conditions with respect to                 in mangrove area, and mangrove remains in the field,
   the second axis. Monte Carlo tests showed that for the                   and interviews with local inhabitants indicated that
   first axis the species-environment correlations were                    human activity is the cause of the mangrove decline in
   not significant (pMonte Carlo test = 0.101), whereas for the                Gazi, mainly overharvesting. This is in line with other
   second axis they were (pMonte Carlo test = 0.013). However,                literature indicating that Gazi is a site with a long-
   the total amount of variability in the species data that                  standing history of mangrove trade and human impact
   could potentially be ‘explained’ by the environmental                   (e.g. Beeckman et al. 1989, Gallin et al. 1989, Vanhove
   factors in this direct ordination was only 4.2 and                     et al. 1992, Kairo 1995a, Schrijvers et al. 1995, Fondo
                                                & Martens 1998, Aloo 2000, Dahdouh-Guebas et al.
                                                2000a, Hoorweg et al. 2000, Abuodha & Kairo 2001,
              Canonical correspondence analysis                  Kairo et al. 2001). The trees reported to be preferred
                                                for cutting were also listed in an in-depth survey of cut-
     80
                          BgAT                     ting preferences for mangrove species (Dahdouh-Gue-
                         BgYT                     bas et al. 2000a). For Mida Creek, further north along
                                                the Kenyan coast, a similar anthropogenic cause for
     60
                                                the quality and spontaneous regeneration of man-
                                                groves was reported (Kairo 2001). However, occasional
Axis 2




                                                natural hazards may further contribute to the reduc-
     40        Sal97/99                               tion in mangrove area; the El-Niño rains of 1997 in
                       AmYT
                                                Kenya for instance caused siltation and a subsequently
                      AmAT
                 SaAT                     CtAT
                                                massive die-off of adult and young trees within a small
     20                                           Rhizophora mucronata stand in Gazi Bay (J.G.K. pers.
                                      CtYT
         SaYT              RmYT                     obs.). This R. mucronata stand, located in the upper
                                            Snails
                         RmAT
                                                right part of the vegetation map of 1992 (Fig. 4b), is
     0                                           however rejuvenating at present as a result of a
       0            40            80
                                                rehabilitation programme (Kairo & Dahdouh-Guebas
                      Axis 1
                                      Crabs        in press).
                                                  Despite the deteriorating status of the forest adjacent
                   Fig. 8. Results of direct species ordina-         to the village, and despite the general decrease in
                   tion (CCA) of presence/absence vege-           area, some mangrove assemblages have expanded.
                   tation data for 99 sample points along
             Lux99    6 transects for adult and young trees.
                                                The low spatial dynamics of the sand banks in the
                   Sal: salinity; Lux: light; other abbrevia-        creek over time (e.g. major sand banks of the creek in
                          tions as in Fig. 7              the aerial photographs of 1972 and 1992 are roughly on
                  Dahdouh-Guebas et al.: Human-impacted mangroves                  89




the same spot) has enabled Sonneratia alba to estab-     in the case of Gazi Bay it is mainly due to canopy over-
lish itself on these banks, very close to the S. alba zone  growth of R. mucronata (with more stems recorded in
of the adjacent forest (at about 150 m). The low desir-   the field study) by huge seaward A. marina (with few,
ability of Avicennia marina for mangrove cutters, as in   thick stems and high canopies).
Mida Creek (Dahdouh-Guebas et al. 2000a), allowed
this species to expand towards the seaward zone.
The landward Ceriops tagal assemblage of 1972 has      Multivariate mangrove vegetation structure analysis
expanded into the mixed zone.
  Unlike our previous report for a disturbed forest in     The distribution of young individuals is more closely
Sri Lanka, using a similar remote sensing and ground-    related to that of adult trees than to the distribution of
truth approach (Dahdouh-Guebas et al. 2000b), com-      juveniles, particularly for Avicennia marina, Bruguiera
parisons between the cover of adult or young man-      gymnorrhiza and Sonneratia alba (Fig. 7). The juve-
grove trees and the dominant canopy species did not     niles are thus generally spread over a wider area than
show a significant difference in Gazi Bay, except      adult trees, but young trees only survive near adult
for the Avicennia marina and Sonneratia alba zone      trees. This is as expected for A. marina and S. alba,
(Table 3). The predominantly seaward position of these    which are pioneering species (e.g. Osborne & Berjak
zones in Kenya (the landward fraction of the disjunct    1997). The same observation was made for B. gymnor-
A. marina zonation pattern is negligible in our study    rhiza in Sri Lankan mangroves (Dahdouh-Guebas et
site: Fig. 4b) implies strong tidal currents that probably  al. 2000b). Although there was no difference rank for
lead to the rare establishment of species compared to    the importance of mangrove juveniles for the most
more landward zones. This is reflected in the high pro-   common species, the 2 sampling years displayed a
portion of ‘nil’ (empty) data points for juveniles during  difference in numbers (Table 2), and further annual
both field surveys (Table 2), and in the absence of a    variability remains possible.
significant seasonal difference between mangrove        The clusters of adult, young and juvenile individuals
juveniles in the S. alba zone (regardless of season, the   of one species can easily be distinguished from those of
tidal currents remain strong in this most seaward      other species (Fig. 7), meaning that the distributions
zone). This rare establishment is also reflected by the   of these individuals are rather similar within species,
low adult tree density in the S. alba zone (28.8 stems    and that there is no high degree of species mixing
ha–1 only) compared to the density in the Rhizophora     (except perhaps for pioneer species). At least for the
mucronata zone (95.3 stems ha–1) and that in the mixed    landward zones, this is corroborated by the lack of sig-
zone (205.2 stems ha–1). However, this is less evident    nificant differences between the distribution of adult
for the seaward A. marina zone (156.5 stems ha–1)      and young trees (Table 4).
because of the R. mucronata-dominated understorey        The omission of juveniles from the ordination matri-
(1 ha = 1000 m2), the complex root system of which      ces emphasised the separation between species clus-
typically facilitates the trapping of propagules (pers.   ters (Fig. 8). Sonneratia alba and Avicennia marina
obs. by F.D.G. & J.G.K.).                  adult trees clearly overlap in distribution, whereas this
  Since Rhizophora mucronata is a ubiquitously im-     is less obvious for the young trees; young A. marina
portant species, irrespective of vegetation class or     trees can be found all along the landward side, but S.
layer, it is emphasised that data from field surveys do   alba young trees are restricted to the S. alba vegetation
not always correspond with remotely sensed data (see     zone. An interesting observation on the understorey
overgrowth of one species by another below). The       of the landward assemblages is that it may be well-
opposite is also true, and fieldwork alone does not     represented in some mangrove assemblages (e.g.
always give a complete picture. For instance, Beeck-     Ceriops tagal). This is contrary to the general claim
man et al. (1989) and Gallin et al. (1989) did not report  that an understorey, whether composed of mangrove
the presence of Avicennia marina trees within the most    or of non-mangrove species, is absent from mangrove
seaward R. mucronata and Sonneratia alba-dominated      stands (Janzen 1985, Snedaker & Lahmann 1988). In
zones in the same study area (see Fig. 4a for the      fact, it is the very presence of C. tagal in the under-
past and Fig. 4a,b for the present situation). On the    storey of assemblages dominated by other mangroves
other hand, species distribution has been reported      that may camouflage a dynamic shift (e.g. Kairo et al.
to be strikingly variable (Dahdouh-Guebas et al.       2002), either imminent or incipient. In the light of
2002b). The discrepancy between remotely sensed and     succession, the present study seems to support the
ground-truth data was reported earlier by Verheyden     hypothesis that mangrove stands should not be con-
et al. (2002) for Sri Lankan mangroves and was attrib-    sidered as intermediate communities preceding terres-
uted to the difficulties of detection within a stand due   trial forests (Johnstone 1983), at least not for high tidal
to interference of different image tonalities. However,   amplitude areas such as Gazi Bay. Whether the results
90                     Mar Ecol Prog Ser 272: 77–92, 2004




of this study favour the hypothesis that mangrove for-     on the sandy area and that are comparable to those in
est comprises a community with its own successional      the landward A. marina zone. Although A. marina is
stages (leading to zonation) and with its own climax      known to thrive in a disjunct zone, it must be deter-
(Snedaker 1982, Johnstone 1983) is not clear; primarily    mined to which extent trees in the seaward zone, with
because of the anthropogenic impact in Gazi Bay,        a different morphology than specimens in the land-
which has been too intensive to speculate on natural      ward zone (Dahdouh-Guebas et al. in press), are able
succession sensu stricto.                   to withstand conditions of the landward zone.
 An overall striking facet of the CCA ordination         A ‘mangrove cutting scenario’ following the current
analysis is the extremely low explanatory power of       cutting preferences is likely to follow the same trend as
environmental variables claimed to play a major role in    the ‘no impact scenario’, but is expected to speed up
the shaping of mangrove vegetation structure, as con-     the expansion of the sandy area, and the degradation
cluded from univariate studies. This emphasises that      of Rhizophora mucronata. Besides the replacement of
mangrove zonation is the result of a complex inter-      the R. mucronata zone by Avicennia marina, another
action and synergism between spatial and temporal       problem is the inadequacy of the frontal A. marina and
factors that extend beyond those investigated in this     Sonneratia zone to stop the effects of waves (Fig. 5c),
paper.                             and the slow washing away of the sand.
                                 Under the above scenarios the landward mangroves
                                on the vegetation map of 1992 (Fig. 4) are not likely
    Predictions based on vegetation history         to change substantially. Scenarios with greater im-
                                pact, however, such as the future alteration or dis-
  In Gazi Bay, anthropogenic influences (e.g. cutting)    appearance of the small topographic differences on
have first led to a direct loss of mangroves (e.g. Beeck-   the landward side at Gazi, will lead to a major re-
man et al. 1989, Gallin et al. 1989, Vanhove et al. 1992,   organisation of the mangrove vegetation structure,
Kairo 1995a, Schrijvers et al. 1995, Fondo & Martens      as well as of the terrestrial vegetation in that area.
1998, Aloo 2000, Dahdouh-Guebas et al. 2000a, Hoor-      Transformation of the multiple topographic ridges and
weg et al. 2000, Abuodha & Kairo 2001, Kairo et al.      depressions to a regular slope in this more landward
2001, this study), and second to further natural degra-    mangrove would be expected to lead to an enlarge-
dation (expanding sandy area). Yet, because this site is    ment of the mangrove assemblages and landward
subject to a pronounced tidal regime, zonation seems      species shifts or species extensions. Terrestrial assem-
to be imposed, in contrast for instance to disturbed      blages on the ridges would be expected to disappear in
sites in Sri Lanka where tidal influence is almost       favour of mangroves tolerant of high salinities, such as
absent (Dahdouh-Guebas et al. 2000b).             Avicennia marina and Lumnitzera racemosa. From the
  The similar distribution of adult and younger vegeta-    species–environment correlations (CCA) it is unclear
tion layers (Table 4; Fig. 7), leads us to predict that    to which environmental variable the distribution of
under a ‘no-impact scenario’ (i.e. in the absence of      Rhizophora mucronata is most correlated. However, as
further human impact or natural catastrophes), in an      observed in the field (see ‘Results’), the border be-
approximately 10–20 yr period (roughly the same        tween R. mucronata and Ceriops tagal assemblages
period in the future as that analysed herein), no major    may be linked to the high water limit at neap tide, and
spontaneous dynamic shifts in the natural distribution     consequently to the amount of time the vegetation is
of the species are to be expected in Gazi Bay, except     submerged daily. Such regularly flooded areas, e.g.
for a possible further expansion of the sandy area       newly formed creeklets, would be expected to become
due to synergism between the former selective cutting     fringed with R. mucronata and C. tagal. This prediction
of trees (primarily Rhizophora mucronata) and the       is valid provided there are no regenerative constraints
subsequent intrusion of sand into the mangroves        such as propagule predation. From the CCA (Fig. 8),
(Fig. 5a,b). Even when the incidence of mangrove        it can be seen that C. tagal positively correlates with
cutting is lower, siltation is able to kill these trees, as  propagule predator density, which is in line with ear-
evidenced by the presence of dead standing trees in      lier reports regarding propagule predation (Dahdouh-
the sandy area (see ‘Results’). This may lead to the dis-   Guebas et al. 1997, 1998), and may also explain the
appearance of the R. mucronata zone, which may be       shift from a mixed mangrove area in 1972 to a C. tagal
replaced by Avicennia marina, a minor shift consider-     area in 1992 (see Kairo et al. 2002).
ing that these 2 species currently form adjacent assem-      Independent of the scenario, human interference
blages and are also present as individuals in each       may be needed to prevent the sandy area from
other’s zones. A. marina can cope with more arid con-     expanding. It might be preferable to replant large, but
ditions (e.g. dry soils) and high light intensity (low for-  protected, areas with the more vulnerable mangrove
est cover), environmental conditions that are present     species such as Rhizophora mucronata. Small-scale
                   Dahdouh-Guebas et al.: Human-impacted mangroves                     91




reforestation programmes in Gazi Bay (Kairo 1995a,b)        Cormier-Salem MC (1999) The mangrove: an area to be
have been shown to successfully (re)convert denuded          cleared… for social scientists. Hydrobiologia 413:135–142
                                  Cottam G, Curtis JT (1956) The use of distance measures in
or degraded mangrove areas into functional forest
                                    phytosociological sampling. Ecology 37:451–460
units (Bosire et al. 2003, 2004). This implies that both      Curtis JT (1959) The vegetation of Wisconsin. An ordina-
floral secondary succession and faunal recruitment           tion of plant communities. University of Wisconsin Press,
occur within the regeneration plots. Continuous moni-         Madison
toring of past and present mangrove vegetation, in nat-      Dahdouh-Guebas F (2002) The use of remote sensing and GIS
                                    in the sustainable management of tropical coastal ecosys-
ural forests as well as rehabilitation areas, will enable       tems. Environ Dev Sustainability 4:93–112
us to better understand mangrove dynamics and per-         Dahdouh-Guebas F, Verneirt M, Tack JF, Koedam N (1997)
haps predict the future vegetation structure in more          Food preferences of Neosarmatium meinerti the Man
detail. An in-depth study of the introgression between         (Decapoda: Sesarminae) and its possible effect on the
                                    regeneration of mangroves. Hydrobiologia 347:83–89
natural and artificially planted mangroves will be be a
                                  Dahdouh-Guebas F, Verneirt M, Tack JF, Van Speybroeck D,
new research challenge.                        Koedam N (1998) Propagule predators in Kenyan man-
                                    groves and their possible effect on regeneration. Mar
                                    Freshw Res 49:345–350
Acknowledgements. The first author is a Postdoctoral        Dahdouh-Guebas F, Mathenge C, Kairo JG, Koedam N
Researcher from the Fund for Scientific Research (FWO,         (2000a) Utilization of mangrove wood products around
Vlaanderen). The research was also financed with a speciali-      Mida Creek (Kenya) amongst subsistence and commercial
sation fellowship of the Institute for the Promotion of Innova-    users. Econ Bot 54:513–527
tion by Science and Technology in Flanders (IWT), and by the    Dahdouh-Guebas F, Verheyden A, De Genst W, Hettiarach-
European Commission (Contract No. IC18-CT96-0065), and is       chi S, Koedam N (2000b) Four decade vegetation dynam-
published with the support of the University Foundation of       ics in Sri Lankan mangroves as detected from sequential
Belgium. We thank the staff of the Kenya Belgium Project and      aerial photography: a case study in Galle. Bull Mar Sci
the Kenya Marine and Fisheries Research Institute (KMFRI)       67:741–759
for providing logistic support. Much gratitude is due to all the  Dahdouh-Guebas F, Verheyden A, Jayatissa LP, Koedam N
inhabitants of Gazi, in particular Latifa Salim and family S.     (2001) A note on the identification of mangroves from aer-
Ba’alawy for hosting us, Abdulhakim Abubakr Ali Jilo and        ial photography in Kenya and Sri Lanka. In: Dahdouh-
Fatuma M. Saidi for their practical help in the field, and       Guebas F (ed) Mangrove vegetation structure dynamics
Abdulbasit M. Daghar, Fatuma M. Saidi, Samir Abubakr and        and regeneration, PhD thesis, Vrije Universiteit Brussel,
Omari Juma Kisasi for their translation assistance during the     Brussels, p 73–83
interview surveys. Chris Gordon (Centre for African Wet-      Dahdouh-Guebas F, Kairo JG, Jayatissa LP, Cannicci S,
lands, University of Ghana) is gratefully acknowledged for       Koedam N (2002a) An ordination study to view vegetation
scientific and style comments on the paper. We also thank       structure dynamics in disturbed and undisturbed man-
3 referees for their constructive comments.              grove forests in Kenya and Sri Lanka. Plant Ecol 161:
                                    123–135
                                  Dahdouh-Guebas F, Verneirt M, Cannicci S, Kairo JG, Tack
          LITERATURE CITED                  JF, Koedam N (2002b) An exploratory study on grapsid
                                    crab zonation in Kenyan mangroves. Wetlands Ecol
Abuodha PAW, Kairo JG (2001) Human-induced stresses on         Manag 10:179–187
  mangrove swamps along the Kenyan coast. In: Harper D,      Dahdouh-Guebas F, De Bondt R, Abeysinghe PD, Kairo JG,
  Boar R, Everard M, Hickley P (eds) Science and the sus-       Triest L, Koedam N (in press) Comparative study of the
  tainable management of shallow tropical waters. Hydro-       disjunct zonation pattern of the grey mangrove Avicennia
  biologia 458:255–265                        marina (Forsk.) Vierh. in Gazi Bay (Kenya). Bull Mar Sci
Aloo P (2000) Marine resources. In: Hoorweg J, Foeken D,      Duke NC (1991) A systematic revision of the mangrove genus
  Obudho RA (eds) Kenya coast handbook. Culture, resources      Avicennia (Avicenniaceae) in Australasia. Aust Syst Bot
  and development in the East African littoral. LIT Verlag,      4:299–324
  Münster, p 43–60                        Duke NC, Jackes BR (1987) A systematic revision of the man-
Beeckman H, Gallin E, Coppejans E (1989) Indirect gradient       grove genus Sonneratia (Sonneratiaceae) in Australasia.
  analysis of the mangal formation of Gazi Bay (Kenya).        Blumea 32: 277–302
  Silva Gandavensis 54:57–72                   Farnsworth EJ, Ellison AM (1997) The global conservation
Bosire J, Dahdouh-Guebas F, Kairo JG, Koedam N (2003)         status of mangroves. Ambio 26:328–334
  Colonisation of non-planted mangrove species into re-      Fondo EN, Martens EE (1998) Effect of mangrove deforesta-
  stored mangrove stands in Gazi Bay, Kenya. Aquat Bot        tion on macrofaunal densities, Gazi Bay, Kenya. Man-
  76:267–279                             groves Salt Marshes 2:75–83
Bosire J, Dahdouh-Guebas F, Kairo JG, Cannicci S, Koedam      Gallin E, Coppejans E, Beeckman H (1989) The mangrove
  N (2004) Spatial variations in macrobenthic fauna re-        vegetation of Gazi bay (Kenya). Bull Soc R Bot Belg 122:
  colonisation in a tropical mangrove bay. Biodiv Conserv       197–207
  13:1059–1074                          Hill MO, Gauch HG (1980) Detrended correspondence analy-
Brokaw N, Thompson J (2000) The H for DBH. For Ecol          sis: an improved ordination technique. Vegetatio 42:47–58
  Manag 129:89–91                         Hoorweg J, Foeken D, Obudho RA (eds) (2000) Kenya coast
Cintrón G, Schaeffer Novelli Y (1984) Methods for studying       handbook. Culture, resources and development in the
  mangrove structure. In: Snedaker SC, Snedaker JG (eds)       East African littoral, LIT Verlag, Münster
  The mangrove ecosystem: research methods. UNESCO,        Janzen DH (1985) Mangroves: where’s the understory? J Trop
  Paris, France, p 91–113                       Ecol 1:89–92
92                       Mar Ecol Prog Ser 272: 77–92, 2004




Jeathold R, Smidt H (1976) Farm management handbook          remarkable pioneering role played by seeds of Avicennia
  of Kenya, Volume II. Part C, Ministery of Agriculture,       marina. Endeavour 21:143–147
  Nairobi                             Schrijvers J, Van Gansbeke D, Vincx M (1995) Macrobenthic
Johnstone IM (1983) Succession in zoned mangrove commu-        infauna of mangroves and surrounding beaches at Gazi
  nities: where is the climax? In: Teas HJ (ed) Tasks for      Bay, Kenya. Hydrobiologia 306:53–66
  vegetation science 8. W. Junk, The Hague, p 131–139       Smith TJ (1987) Seed predation in relation to tree dominance
Kairo JG (1995a) Artificial regeneration and sustainable yield    and distribution in mangrove forest. Ecology 68:266–273
  management of mangrove forests in Gazi Bay, Kenya,       Smith TJ III (1992) Forest structure. In: Robertson AI, Alongi
  MSc thesis botany, University of Nairobi, Nairobi         DM (eds) Tropical mangrove ecosystems. American Geo-
Kairo JG (1995b) Community participatory forestry for reha-      physical Union Washington, DC, p 101–136
  bilitation of deforested mangrove areas of Gazi Bay       Snedaker SC (1982) Mangrove species zonation: why? In: Sen
  (Kenya). A first approach. Final technical report. Univer-     DN, Rajpurohit KS (eds) Tasks for vegetation science 2,
  sity of Nairobi, Department of Botany, Nairobi           W. Junk, The Hague, p 111–125
Kairo JG (2001) Ecology and restoration of mangrove systems    Snedaker SC, Lahmann EJ (1988) Mangrove understorey ab-
  in Kenya, PhD dissertation, Vrije Universiteit Brussel,      sence: a consequence of evolution? J Trop Ecol 4:311–314
  Brussels                            Sokal RR, Rohlf FJ (1981) Biometry. The principles and prac-
Kairo JG, Dahdouh-Guebas F (in press) Mangroves of Kenya.       tice of statistics in biological research, 2nd edn. WH Free-
  In: Ashton E (ed) Code of conduct for sustainable manage-     man & Co, New York
  ment of mangrove ecosystems. Internation Society of       Ter Braak CJF (1986) Canonical correspondence analysis: a
  Mangrove Ecosystems (ISME), Okinawa, Japan             new eigenvector technique for multivariate direct gradi-
Kairo JG, Dahdouh-Guebas F, Bosire J, Koedam N (2001)         ent analysis. Ecology 67:1167–1179
  Restoration and management of mangrove systems — a       Ter Braak CJF (1994) Canonical community ordination. Part I:
  lesson for and from the East African region. S Afr J Bot 67:    Basic theory and linear methods. Ecoscience 1:127–140
  383–389                             Ter Braak CJF (1995) Ordination. In: Jongman RHG, Ter
Kairo JG, Dahdouh-Guebas F, Gwada PO, Ochieng C, Koedam        Braak CJF, Van Tongeren OFR. Data analysis in commu-
  N (2002) Regeneration status of mangrove forests in Mida      nity and landscape ecology, Cambridge University Press,
  Creek, Kenya: a compromised or secured future? Ambio        Cambridge, p 91–173
  31:562–568                           Tomlinson CB (1986) The botany of mangroves. Cambridge
Kent M, Coker P (1992) Vegetation description and analysis.      tropical biology series, Cambridge University Press, New
  A practical approach, John Wiley & Sons, Chichester        York
Lieth H, Berlekamp J, Fuest S, Riediger S (1999) Climate      Vandeput S (1999) De haalbaarheid van een kleinschalig
  diagram world atlas, Backhuys, Leiden               ecotoeristisch project in Gazi. MSc thesis human ecology,
Matthijs S, Tack J, Van Speybroeck D, Koedam N (1999)         Vrije Universiteit Brussel, Brussel (in Dutch)
  Mangrove species zonation and soil redox state, sulphide    Vanhove S, Vincx M, Van Gansbeke D, Gijselinck W, Schram
  concentration and salinity in Gazi Bay (Kenya), a prelimi-     D (1992) The meiobenthos of five mangrove vegetation
  nary study. Mangroves Salt Marshes 3:243–249            types in Gazi Bay, Kenya. Hydrobiologia 247:99–108
McCune B, Mefford MJ (1997) Multivariate analysis of eco-     Verheyden A, Dahdouh-Guebas F, De Genst W, Hettiarach-
  logical data, Version 3.09, MjM Software, Gleneden         chi S, Koedam N (2002) High-resolution vegetation data
  Beach, OR                             for mangrove research as obtained from aerial photo-
Osborne DJ, Berjak P (1997) The making of mangroves: the       graphy. Envion Dev Sustainability 4:113–133

Editorial responsibility: Victor de Jonge (Contributing Editor),  Submitted: November 28, 2002; Accepted: December 1, 2003
Haren, The Netherlands                       Proofs received from author(s): April 22, 2004