Coral Reefs (2007) 26:641–653
DOI 10.1007/s00338-007-0230-1

REPORT



Phase shifts and the role of herbivory in the resilience
of coral reefs
M. H. Ledlie Æ N. A. J. Graham Æ J. C. Bythell Æ
S. K. Wilson Æ S. Jennings Æ N. V. C. Polunin Æ
J. Hardcastle




Received: 7 November 2006 / Accepted: 22 March 2007 / Published online: 17 May 2007
Ó Springer-Verlag 2007

Abstract Cousin Island marine reserve (Seychelles) has       structure. Analysis of the feeding habits of six abundant
been an effectively protected no-take marine protected area     and representative herbivorous fish species around Cousin
(MPA) since 1968 and was shown in 1994 to support a         Island in 2006 demonstrated that epilithic algae were the
healthy herbivorous fish assemblage. In 1998 Cousin Island      preferred food resource of all species and that macroalgae
reefs suffered extensive coral mortality following a coral     were avoided. Given the current dominance of macroalgae
bleaching event, and a phase shift from coral to algal       and the apparent absence of macroalgal consumers, it is
dominance ensued. By 2005 mean coral cover was <1%,         suggested that the increasing abundance of macroalgae is
structural complexity had fallen and there had been a        reducing the probability of the system reverting to a coral
substantial increase in macroalgal cover, up to 40% in       dominated state.
some areas. No clear trends were apparent in the overall
numerical abundance and biomass of herbivorous fishes        Keywords Recovery Á Coral bleaching Á Seychelles Á
between 1994 and 2005, although smaller individuals be-       Marine protected areas Á Coral reef fishes Á Feeding
came relatively scarce, most likely due to the loss of reef     observations


Communicated by Ecology Editor P.J. Mumby.             Introduction
M. H. Ledlie (&) Á N. A. J. Graham Á S. K. Wilson Á
N. V. C. Polunin                          The interaction between natural and anthropogenic distur-
School of Marine Science and Technology,              bance has undermined the resilience of coral reefs and led
Newcastle University, Newcastle-upon-Tyne                                   ¨
                                  to their worldwide degradation (Nystrom et al. 2000;
NE1 7RU, UK
                                  Gardner et al. 2003; Hughes et al. 2003). Coral bleaching is
e-mail: maryledlie@hotmail.com
                                  one such disturbance which, through potential enhance-
J. C. Bythell                            ment by anthropogenic global warming (Reaser et al.
School of Biology, Newcastle University,              2000), poses a great challenge to coral reef management.
Newcastle-upon-Tyne NE1 7RU, UK
                                  While the immediate effects of coral bleaching on reef fish
S. K. Wilson                            assemblages are largely restricted to species which depend
Australian Institute of Marine Science,               on live coral for habitat or food (reviewed by Wilson et al.
Townsville MC, Townsville, QLD 4810, Australia           2006), there is evidence that reefs can support abundant
                                  and diverse fish assemblages after bleaching as long as reef
S. Jennings
Lowestoft Laboratory, Centre for Environment,            structure is maintained (Lindahl et al. 2001). However,
Fisheries and Aquaculture Science,                 the longer-term loss of structural complexity can affect
Lowestoft NR33 0HT, UK                       recruitment, competition and predation (Buchheim and
                                                       ¨
                                  Hixon 1992; Hixon and Beets 1993; Ohman et al. 1998),
J. Hardcastle
Nature Seychelles, The Centre for Environment and Education,    leading to declines in species richness (Graham et al. 2006)
                  ´
P.O. Box 1310, Roche Caiman, Mahe, Seychelles            and numerical abundance (Garpe et al. 2006).


                                                           123
642                                                    Coral Reefs (2007) 26:641–653


  The loss of live coral cover following disturbances such       greatest coral mortalities on record worldwide (Hoegh-
as bleaching is often accompanied by a proliferation of         Guldberg 1999; Goreau et al. 2000) but the Indian Ocean
macroalgae (McClanahan et al. 2001; Graham et al. 2006).        was most severely affected (Sheppard 2003). In the Cousin
Although there has been some speculation regarding the         Island no-take MPA (Inner Seychelles), there was a massive
causality and mechanisms of competitive interactions be-        decline in live coral cover in 1998, followed by an ongoing
tween algae and corals (McCook et al. 2001), high algal         decline in structural complexity. The aim of this study was
biomass has been demonstrated to have a detrimental effect       to describe changes in the composition of benthic commu-
on coral health (e.g. Birkeland 1977; Tanner 1995; Smith        nities following the 1998 bleaching event and to assess the
et al. 2006). Through their role in algal removal, herbivo-       potential role of herbivorous fishes in promoting reef
rous fishes are considered to play an important role in         recovery. This was achieved by (1) analysing benthic
promoting reef resilience and in reef recovery to coral         community and herbivorous fish biomass and numerical
                          ¨
dominated states if disturbance has occurred (Nystrom and        abundance data from before, immediately after and seven
Folke 2001; Bellwood et al. 2004).                   years after the bleaching event and (2) assessing the feeding
  The role of herbivorous fishes in promoting coral           habits of six abundant and representative herbivorous fish
recovery and enhancing resilience suggests that relatively       species to determine whether they had the potential to create
small-scale (km to 10s km) variation in the structure and        suitable conditions for successful coral settlement.
abundance of their assemblages will contribute to small-
scale variation in rates of reef recovery. Management
measures that locally reduce fishing mortality and increase       Materials and methods
the abundance of herbivorous fishes, such as marine pro-
tected areas (MPAs), are therefore expected to play an         Study area
important role in promoting recovery and resilience
(Hughes et al. 2003). The extent to which different groups       The small granitic island of Cousin is situated in the Inner
of herbivorous fishes promote recovery will depend on          Seychelles (Fig. 1) (4°20¢S, 55°40¢E). The entire island
their functional role and the algae that they graze. Three       including the surrounding reef was declared a ‘Special
functional groups have been recognised within the her-         Reserve’ in 1968 and 1.2 km2 of sea, extending 400 m
bivorous fish guild—grazers, scrapers and bioeroders—and         seaward from the high-water mark, is legally protected
these have different and complementary roles in precon-         (Jennings et al. 1996). Effective and continuous policing by
ditioning reefs to permit coral recovery following distur-       resident Seychellois wardens has ensured the reef has not
bance (Bellwood et al. 2004). Furthermore, in terms of         been fished in recent years (Jennings 1998).
enhancing resilience it is not only the species diversity
within functional groups which is important but the re-         Temporal assessment of herbivorous fish assemblage
sponse diversity of these species to environmental change        and benthic composition
(Elmqvist et al. 2003).
  Although MPAs may be an effective means of reducing         Three sites were used in the temporal monitoring of the
local disturbance (Jennings et al. 1996; Halpern 2003), they      herbivorous fish assemblage and benthic composition
are not immune to the effects of large-scale external dis-       (Fig. 1). Site M1 was a fringing reef with a carbonate
turbances such as bleaching (Reaser et al. 2000; Jones et al.      framework, site M2 was characterised by coral growth on a
2004). The global bleaching event of 1998 led to the          granitic substrate and site M3 was patch reef on a base of

Fig. 1 The Inner Seychelles                                  Cousin Island
                        4°10'S




and Cousin Island showing
                                                                  N
location of monitoring sites                                      M2
(M1, M2 and M3) and feeding                                      FA
                        4°20 'S




                                   Cousin
study sites (FA, FB and FC)
(adapted from Jennings et al.                         Praslin                      FB
1996)
                        4°30'S




                                                M1
                        4°40'S




                                    Mahé                        M3
                        4°50'S




                                                              FC

                                                          50m
                             55°20'E  55°30'E 55°40'E 55°50'E




123
Coral Reefs (2007) 26:641–653                                               643


sand, rock or rubble (Jennings et al. 1995). The sites were   boulders covered with epilithic algae from 2–5 m and
surveyed at the same time of year in 1994, 1998 (imme-      macroalgae and sand patches below 5 m.
diately after the bleaching event) and in 2005, although
site M2 was not surveyed in 1998. At each site, fishes were    Study animals
counted in sixteen 7-m-radius point counts. Large mobile
species were counted first, before the count area was ac-     For the feeding study, species were chosen to represent the
tively searched for territorial species. Point counts were    three functional groups of reef herbivores identified by
considered appropriate because spearfishing is banned in     Bellwood et al. (2004); bioeroders, scrapers and grazers.
the Seychelles and recreational diving and fish feeding,     This study focused on the most abundant species in each
which can result in fish gathering around divers, do not     functional group. Chlorurus sordidus and Chlorurus
take place in Cousin MPA. The size and numerical abun-      strongylocephalus were chosen to represent the bioeroders
dance of herbivorous fishes were recorded as part of a      (e.g., Bellwood 1995). C. sordidus was abundant at all three
larger study in which the individual size and numerical     sites but C. strongylocephalus was only observed at sites FB
abundance of 134 reef-associated fish species were        and FC. However, the scarcity of any other bioeroding fishes
documented. The time taken to complete the census was      in the study area limited the choice of species. Scarus niger
not standardised and varied according to the number and     and Scarus rubroviolaceus were chosen to represent the
diversity of species present. The accuracy of length       scrapers (Bellwood and Choat 1990). Observations of scar-
estimation was maintained by practising with objects of     ids were restricted to terminal phase individuals. While it is
known size (from 8–35 cm) throughout the survey period      possible that the feeding preferences of such individuals may
and mean errors were 2.2–3.1%. Size estimates of fishes      have been influenced by reproductive activity, evidence
were converted to biomass using published length-weight     from the Caribbean has indicated that the diet of initial and
relationships (Letourneur 1998; Letourneur et al. 1998;     terminal phase scarids does not differ (Bruggemann et al.
Froese and Pauly 2006).                     1994; McAfee and Morgan 1996). Acanthurus leucosternon
  When a fish count was complete the percentage cover      and Acanthurus triostegus were chosen to represent the
(based on plan view) of sand, rock, rubble, macroalgae,     grazers (e.g., Barlow 1974).
dead and live branching coral, massive, tabulate, encrusting
and soft coral was estimated. The topographic structural     Behavioural observations
complexity of the reef inside each count area was described
using a six-point scale (Polunin and Roberts 1993). Visual    All feeding observations were undertaken whilst snorkel-
estimates of these benthic parameters provide comparable     ling. The prohibition of spearfishing in the Seychelles, and
values to line intercept transects for benthic composition (no  the fact that Cousin has been protected from any other
significant difference P = 0.639) and linear versus contour    fishing or tourist diving activity since 1968, meant that
chain distance for structural complexity (significantly cor-   fishes generally did not show any obvious behavioural re-
related, P < 0.001) (Wilson et al. 2007).            sponse to an observer at distances > 2–3 m, and would
                                 often swim underneath the observer and continue feeding.
Feeding habits                          However some species such as A. triostegus were more
                                 skittish, especially in shallow water, and observations
Data relating to the feeding habits of fishes were collected   could only be undertaken when visibility was sufficient to
from 22 April to 31 May 2006. Three study sites (FA, FB     allow observations from a greater distance. In all cases
and FC) were selected as representative of different reef    observations were discontinued if the fishes appeared to
habitats around the island (Fig. 1). These locations were    have been disturbed by the observer.
slightly different from those used in monitoring of the fish
assemblage and benthic composition and were therefore      Diet composition
treated separately. Site FA (northern side of Cousin)
was structurally non-complex and from 1–5 m depth was      Upon arrival at the site the observer swam in a rough zig-
dominated by dense macroalgae growing on a granitic       zag pattern from the reef flat to the reef slope (from depths
substrate with intermittent sand and coral rubble patches.    of approximately 3–6 m) until a target individual was lo-
Below 5 m macroalgal cover was sparse and sand was the      cated (Bellwood 1995). After a short acclimation period the
abundant substrate type. Site FB (NE side of Cousin) was a    fish was followed for a period of 5 minutes during which
reef slope environment ranging from 5–8 m in depth and      the number of bites on different substrates was recorded, in
consisting mainly of dead coral rubble covered in epilithic   addition to the time of day and depth.
algae and sand patches. Site FC (SE side of Cousin) was a      Within each survey period every effort was made to
structurally complex reef consisting of large granitic      select different individuals for observation; for some


                                                          123
644                                               Coral Reefs (2007) 26:641–653


species their abundance ensured a degree of independence    that had observations for all factors were analysed sepa-
but for those which were less abundant the size and any     rately (Quinn and Keough 2002). This involved one
distinctive markings were noted to ensure that the same     ANOVA with data from 1994 and 2005 at sites M1, M2
individual was not observed twice. All observations of     and M3 and one with data from 1994, 1998 and 2005 at
feeding behaviour were conducted by the same observer      sites M1 and M2. Normality of data was examined using
(M.H.L.) between 09.00 and 16.00. A total of 168 fishes     probability plots of the residuals and homogeneity of
were observed and as time of day has previously been      variances was tested using Levene’s test (P < 0.05). Data
shown to affect the feeding rates of herbivorous fishes     required log10 transformation to meet the assumptions
(Polunin and Klumpp 1989) observations were split as      of the analysis. In the case of macroalgal cover the
evenly as possible between morning and afternoon.        assumption of homogeneity of variance could not be met
                                due to the presence of outliers. The removal of these
Substratum availability                     outliers allowed the assumption of homogeneity of vari-
                                ance to be met but did not affect the inference of the test
To determine whether feeding preferences were influenced     and the original results are reported. Where differences
by the availability of potential food resources, substratum   were significant, Tukey’s test was used to identify sig-
availability was quantified in situ. This was possible only at  nificant pairwise differences.
sites FA and FB as adverse sea conditions restricted access
to site FC at the end of the study period. At sites FA and FB  Temporal fish numerical abundance and biomass data
twenty randomly placed replicate 5 m transects were sur-
veyed using the line point intercept method. Substrate type   A 2-way ANOVA, with year and monitoring site as fixed
was recorded at 20 cm intercepts. For site FC, analysis     factors, was used to assess changes in numerical abun-
of digital photographs was used to quantify substratum     dance and biomass of the whole herbivorous fish assem-
availability by estimating the percentage cover of the     blage and of the three functional groups of herbivorous
different substrates beneath lines % 5 m overlain on the    fishes. The absence of data for site M3 in 1998 was
photographs. Algal vegetation was categorised as: epilithic   overcome using the approach described for benthic data.
algae [multispecies assemblage of diminutive algae with a    Probability plots of residuals were used to assess normality
canopy height of less than 1 cm (Steneck 1988)], macro-     of the data and homogeneity of variances were tested using
algae [large fleshy algae with canopy heights greater than    Levene’s test (P < 0.05). Numerical abundance data were
1 cm (Steneck 1988), in this case dominated by Sargassum    log10 transformed and biomass data were square root
and Turbinaria], crustose coralline algae (encrusting cal-   transformed to meet the assumptions of normality and
careous algae occurring as a hard, smooth pavement on the    homogeneity of variance. In several cases the assumption
substratum) and other algae (predominantly Chlorodesmis).    of homogeneity of variance could not be met due to the
Non-algal categories were live coral, sand and dead coral    presence of outliers. Again, the removal of these outliers
(coral which had recently died and had not been colonised    allowed the assumption of homogeneity of variance to be
by algae).                           met but did not affect the inference of the tests and
                                the original results are reported. Where differences were
Data analysis                          significant, Tukey’s test was used to identify significant
                                pairwise differences.
Temporal benthic data
                                Feeding selectivity
Changes in benthic composition were analysed using a
correlation-based principal components analysis. Drafts-    Two separate 2-way crossed ANOSIMs with site and
man plots were used to indicate any skewness in the data    species, and with site and functional group, based on Bray-
and variables were log10 transformed in order to improve    Curtis similarity matrices were undertaken in PRIMER
the spread. In order to place the data on a common scale    (Plymouth Routines in Multivariate Ecological Research)
they were normalised by subtracting the mean and        to assess differences in the proportion of bites on different
dividing by the standard deviation (Clarke and Gorley      substrates (Clarke and Gorley 2006). As most bites were on
2006). Changes in live coral and macroalgal cover        epilithic algae, data were square root transformed to give
between 1994, 1998 and 2005 were assessed using a 2-      greater relative weighting to bites on other substrates.
way ANOVA with year and monitoring site as fixed          Ivlev’s electivity index was used to quantify feeding
factors. Since monitoring sites were located in different    selectivity, by comparing the proportion of bites taken on
reef habitats they could not be pooled for analysis and as   different substrates with respect to their availability (Ivlev
site M3 had not been surveyed in 1998, subsets of the data   1961). The index is defined as


123
Coral Reefs (2007) 26:641–653                                                                                        645


E ¼ ðri À pi Þ=ðri þ pi Þ                                               Phaeophyta from the genera Sargassum and Turbinaria,
                                                            with some Padina) and bare rock (Fig. 2). Coral cover
where ri is the proportion of bites taken on food type i and                              declined from 1994 to 1998 and again from 1998 to 2005
pi is the average percent occurrence of food type i in the                               with similarly low levels at all three sites in 2005 (Fig. 3a).
environment. The index ranges from +1.0 to –1.0 with a                                 Although there was no increase in macroalgal cover be-
positive value indicating preference, a negative value                                 tween 1994 and 1998, there had been a substantial increase
indicating avoidance and zero representing no selection                                by 2005 (Fig. 3b). However there were marked differences
(Ivlev 1961).                                                     in macroalgal cover among monitoring sites in 2005, with
                                                            site M2 having much lower cover than sites M1 and M3
                                                            (Fig. 3b). There was a significant interaction between year
Results                                                        and monitoring site, due to smaller declines in coral cover
                                                            and smaller increases in macroalgal cover at site M2 than
Changes in benthic composition 1994–2005                                        at the other sites over time. There was also a significant
                                                            difference in coral and macroalgal cover among years and,
In 1994 the reefs were dominated by live (massive and                                 in most cases, among monitoring sites (Table 1).
branching) coral and were structurally complex (Fig. 2).
The bleaching event of 1998 resulted in a massive reduc-
tion in live coral cover but the structural complexity was                               a             60
maintained (Fig. 2). By 2005 this complexity had been lost                                                                      1994



                                                            Mean % coral cover ± SE
                                                                                                   1998
and the reefs were dominated by macroalgae (mainly                                                                          2005
                                                                         40


              4
                                                   M1 1994
                                                   M2 1994                  20
                                                   M3 1994
                                                   M1 1998
              2                                     M2 1998                                           X
PC2 (23.2% of variation)




                                                   M1 2005                   0
                                      macroalgae
                                                   M2 2005                        M1       M2          M3
                   rubble                             M3 2005
                                    live                      b             60
                   live              encrusting
              0
                 branching
                                                            Mean % macroalgae ± SE




                            live
                            plating        rock
                                     sand                                  40


              -2         live massive
                            structural     dead branching
                            complexity
                                                                         20


              -4
                                                                                                 X
                 -4       -2        0        2      4       6                   0
                               PC1 (25.6% of variation)                                    M1       M2          M3


Fig. 2 Principal components analysis of change in benthic compo-                            Fig. 3 a Mean percent coral cover, and b mean percent macroalgal
sition and structure of Cousin reefs over period 1994–1998–2005.                            cover at monitoring sites in Cousin Marine Protected Area in 1994,
Eigenvectors of each benthic variable are overlaid                                   1998 and 2005. X indicates a lack of data for site M3 in 1998


Table 1 Results of two-way ANOVAs on percentage cover of coral and macroalgae
Variate                         Year                          Monitoring site                       Year · monitoring site
                            df       F ratio       P value       df                F ratio     P value  df     F ratio    P value

Comparison of 1994 and 2005 at monitoring sites M1, M2 and M3
Coral cover                       1,90      1131.87       <0.001       2,90                 3.02     0.054  2,90    11.72     <0.001
Macroalgal cover                    1,90       575.85      <0.001       2,90               55.77      <0.001  2,90    68.75     <0.001
Comparison of 1994, 1998 and 2005 at monitoring sites M1 and M2
Coral cover                       2,90       188.48      <0.001       1,90               18.43      <0.001  2,90    9.93     <0.001
Macroalgal cover                    2,90       210.89      <0.001       1,90               68.52      <0.001  2,90    74.88     <0.001



                                                                                                 123
646                                                                 Coral Reefs (2007) 26:641–653


Numerical abundance and biomass of herbivorous                  abundance of smaller individuals and an increase in larger
fishes                                       individuals (Graham et al. 2007). It is likely to have been
                                         driven by a combination of increased growth and/or sur-
No consistent trends were apparent in numerical abundance             vivorship of larger fishes due to improved dietary resources
or biomass of grazers, scrapers or bioeroders among                and an increase in the mortality of smaller fishes due to a
monitoring sites or years (Fig. 4; Table 2). Total numerical           loss of reef structure (Graham et al. 2007). Such changes
abundance remained stable from 1994–1998–2005 at site               were especially notable in the case of grazers and scrapers
M1 but decreased at sites M2 and M3, while total biomass             whereas the numerical abundance and biomass of the
increased at site M1 but showed little change at sites M2             bioeroders remained relatively more stable over time
and M3 (Fig. 4). Changes in the numerical abundance and              (Fig. 4). There were significant interactions between year
biomass of the three functional groups were very variable             and monitoring site for both numerical abundance and
and differed among monitoring sites but in the majority of            biomass (Table 2); in many cases this was due to greater
cases numerical abundance declined or remained stable               declines in numerical abundance at sites M2 and M3 than
whereas biomass increased or declined to a lesser extent             at site M1. Numerical abundance of grazers and bioeroders
(Fig. 4). This trend reflects a change in the size structure of          differed significantly among years, as did biomass of bio-
the fish assemblage due to a decline in the numerical               eroders (Table 2). There was also a significant difference




Fig. 4 a Numerical abundance,  a                                 b
and b biomass of herbivorous                                 1994
                             10                               2000
                 Mean abundance ± SE




fishes at monitoring sites in                     Total                                Total




                                                  Mean biomass ± SE
                                               1998
Cousin Marine Protected Area               8
                                               2005            1500
in 1994, 1998 and 2005
                             6
                                                            1000
                             4
                                                             500
                             2

                             0                                0
                               M1       M2     M3                   M1      M2     M3

                             16                               2000      Grazers
                 Mean abundance ± SE




                                  Grazers
                                                  Mean biomass ± SE




                             12                               1500

                             8                               1000

                             4                                500

                             0                                 0
                               M1    M2    M3                      M1      M2     M3

                             5                               2000
                  Mean abundance ± SE




                                  Scrapers                              Scrapers
                                                  Mean biomass ± SE




                             4                               1500
                             3
                                                            1000
                             2
                                                            500
                             1

                             0                                 0
                               M1    M2     M3                     M1      M2     M3

                             10                               2000
                 Mean abundance ± SE




                                  Bioeroders                             Bioeroders
                                                  Mean biomass ± SE




                             8                               1500
                             6
                                                            1000
                             4
                                                             500
                             2

                             0                                 0
                               M1    M2    M3                      M1      M2     M3



123
Coral Reefs (2007) 26:641–653                                                     647


Table 2 Results of two-way ANOVAs on numerical abundance and biomass of herbivorous fish species
Variate       Year                     Monitoring site             Year · monitoring site
           df      F ratio    P value     df     F ratio    P value    df     F ratio    P value

Comparison of 1994 and 2005 at monitoring sites M1, M2 and M3
Numerical abundance
Grazers       1,83     32.65     <0.001*     2,83     0.96     0.387    2,83     2.56     0.083
Scrapers      1,76     1.30      0.259     2,76     3.97     0.023*    2,76     3.20     0.046*
Bioeroders     1,65     0.78      0.381     2,65     1.45     0.243    2,65     2.16     0.123
All         1,87     29.17     <0.001*     2,87     1.64     0.199    2,87     7.70     0.001*
Biomass
Grazers       1,90     0.00      0.986     2,90     0.26     0.769    2,90     3.20     0.046*
Scrapers      1,90     1.17      0.282     2,90     9.12    <0.001*    2,90     6.35     0.003*
Bioeroders     1,90     0.12      0.733     2,90     3.76     0.027*    2,90     6.23     0.003*
All       1,90     1.13     0.290     2,90         5.13     0.008*    2,90    10.51     <0.001*
Comparison of 1994, 1998 and 2005 at monitoring sites M1 and M2
Numerical abundance
Grazers       2,85     12.95     <0.001*     1,85     2.51     0.117    2,85     4.71     0.012*
Scrapers      2,79     0.75      0.475     1,79     0.08     0.774    2,79    13.47     <0.001*
Bioeroders     2,70     4.91      0.010*     1,70    10.04     0.002*    2,70     4.65     0.013*
All         2,87     8.93     <0.001*     1,87     1.34     0.250    2,87    10.52     <0.001*
Biomass
Grazers       2,90     1.33      0.269     1,90     0.04     0.833    2,90     3.07     0.051
Scrapers      2,90     1.36      0.261     1,90     3.98     0.049*    2,90    16.12     <0.001*
Bioeroders     2,90     6.11      0.003*     1,90    29.99     <0.001*    2,90    10.66     <0.001*
All         2,90     2.91      0.060     1,90     5.78     0.018*    2,90    13.04     <0.001*
* Indicate statistically significant results at a significance level of 0.05




among monitoring sites in numerical abundance and bio-            was a significant pairwise difference between grazers and
mass of bioeroders and scrapers (Table 2).                  bioeroders (P = 0.029).

Feeding selectivity
                                       Discussion
There was a significant difference among species (ANO-
SIM, P = 0.024) in terms of proportion of bites on different         The three functional groups of herbivores studied here all
substrates but not among feeding study sites and the only          appeared to prefer epilithic algae over other algal resources
significant pairwise difference was between A. triostegus           and frequently avoided macroalgae. Such preferences have
and C. sordidus (P = 0.001).                         also been noted for Caribbean herbivores (Bruggemann
  Epilithic algae and macroalgae (predominantly Sargas-           et al. 1994; Paddack et al. 2006), for Indo-Pacific scarids
sum and Turbinaria) were the most abundant substrate             (Bellwood and Choat 1990) and for herbivorous reef fishes
types at the feeding study sites although more than 80%           in general (Russ and St. John 1988; Choat 1991; Bellwood
of bites for all fish species were from epilithic algae            et al. 2006). Furthermore, analysis of the gut contents of
(Table 3). Only A. triostegus and S. rubroviolaceus con-           some of the species studied here support the finding that
sumed macroalgae in any notable proportion although bites          macroalgae are rarely consumed by these fishes (Hiatt and
from this substrate represented less than 10% of their total         Strasburg 1960; Robertson et al. 1979; Sano et al. 1984;
bites (Table 3). Electivity indices confirmed that most            Choat et al. 2002). The avoidance of macroalgae has been
fishes consumed only epilithic algae and avoided macro-            attributed to the presence of chemical and physical deter-
algae and other substrata (Fig. 5). There was no significant         rents which renders macroalgae less palatable and digest-
difference overall in the proportion of bites taken on dif-         ible to herbivores (e.g., Hay et al. 1987), although intense
ferent substrates among functional groups although there           herbivorous grazing of epilithic algal turfs has been shown


                                                                  123
648                                                                                      Coral Reefs (2007) 26:641–653

Table 3 Availability of
                  Substrate                         Epilithic         Macro-          Crustose         Live      Other        Sand        Dead
substrate types averaged over all
                                                algae           algae          coralline         coral     algae                 coral
three feeding study sites and
                                                                          algae
mean percentage of bites taken
on these substrates by each fish   Mean availability (%)                   53.00           27.58          0.28           1.06      0.98        17.00       0.08
species
                  Mean percentage of bites
                  Chlorurus sordidus                     98.16            1.46          0.03           0.12      0.00         0.12       0.12
                  Chlorurus strongylocephalus                100             0.00          0.00           0.00      0.00         0.00       0.00
                  Acanthurus leucosternon                  98.79            1.00          0.17           0.03      0.00         0.00       0.00
                  Acanthurus triostegus                   85.08            7.96          0.00           0.50      0.13         0.61       0.00
                  Scarus rubroviolaceus                   95.26            4.69          0.00           0.05      0.00         0.00       0.00
                  Scarus niger                        98.01            0.00          1.91           0.08      0.00         0.00       0.00



to limit the establishment and growth of macroalgae (Lewis                      these algae (Paddack et al. 2006). Coral cover has been
1986; Williams et al. 2001; Paddack et al. 2006).                          implicated as an important factor in determining the impact
  Several sites on Cousin are dominated by macroalgae                        of herbivorous fishes on algae, and on high coral-cover
and macroalgal cover has increased through time. This                        reefs the impact of herbivorous fishes on algae will be
highlights the limited capacity of the herbivorous fish guild                     greater since there will be stronger competition for the
to exclude macroalgae. Unfished herbivorous fish guilds in                       limited algal resources. Conversely, on low coral cover
the Caribbean could only maintain 40–60% of reef sub-                        reefs the large amount of space occupied by algae limits
stratum in a cropped state (Williams and Polunin 2001) and                      the ability of herbivorous fishes to keep it cropped down
in the Florida Keys the high biomass of herbivorous fishes                      (Williams et al. 2001). On Cousin Island coral cover
restricted macroalgal spread, but did not entirely exclude                      declined dramatically and algal cover increased, yet the


Fig. 5 Mean values of Ivlev                                                                    GRAZERS
(1961)’s electivity indices for                                       A. leucosternon (31)                              A. triostegus (35)
                                             1                                       1
Acanthurus leucosternon,
Acanthurus triotegus, Scarus                               0.5                                      0.5
rubroviolaceus, Scarus niger,
Chlorurus sordidus and                                  0                                       0
Chlorurus strongylocephalus,
                                            -0.5                                     -0.5
averaged over all feeding study
sites. Numbers in brackets after                             -1                                      -1
species refer to number of
individuals observed                                                                        SCRAPERS
                                                      S. rubroviolaceus (21)                                 S. niger (16)
                              Electivity index ± 95% CI




                                             1                                       1

                                            0.5                                      0.5

                                             0                                       0

                                            -0.5                                     -0.5

                                             -1                                      -1

                                                                              BIOERODERS
                                                       C. sordidus (53)                            C. strongylocephalus (12)
                                              1                                      1

                                             0.5                                     0.5

                                              0                                      0

                                            -0.5                                     -0.5

                                             -1                                      -1
                                                                                                                  Other algae
                                                                           Other algae




                                                                                                           Live coral
                                                                   Live coral




                                                                                               Macroalgae
                                                       Macroalgae




                                                                                        Epilithic




                                                                                                     Coralline
                                                Epilithic




                                                             Coralline




                                                                                        algae
                                                algae




                                                                                                     algae
                                                             algae




123
Coral Reefs (2007) 26:641–653                                               649


numerical abundance and biomass of algal feeding fishes      (2005) also indicated that predation by scarids may have
generally did not increase sufficiently to control and restrict  retarded post-bleaching recovery of coral transplants in
macroalgal development.                     Kenya. One of the most prolific feeders on live coral is
  The recovery of coral reefs following disturbance is     Bolbometopon muricatum, which take nearly half of all
complex and there is evidence to suggest that once a phase    their bites from this substrate on the Great Barrier Reef
shift has been initiated, hysteresis effects can inhibit     (Bellwood et al. 2003). Schools of B. muricatum can often
reversal (Scheffer and Carpenter 2003). Around Cousin, it    be seen around Cousin (Jennings 1998; M. H. Ledlie,
is likely that following the decline in coral cover in 1998,   personal observation) although their role in reef recovery is
epilithic algae became dominant and colonised the avail-     largely undefined, as the feeding response of this species
able space. Not only can epilithic algal turfs reduce coral   when coral cover is low is unknown.
settlement in their own right (Birrell et al. 2005) but once    The role of herbivorous fishes in promoting reef
they develop into macroalgae they become increasingly      recovery and resilience is likely to depend not only on their
resistant to perturbations (McManus and Polsenberg 2004)     feeding preferences but also on their numerical abundance
and the community becomes more stable. High algal bio-      and biomass, which may be affected by changes in
mass has been shown to negatively affect coral recruitment    the benthos. With an increase in algal availability, the
(Birkeland 1977; Kuffner et al. 2006) and growth rates      numerical abundance and biomass of herbivorous fishes
(Tanner 1995) and algae may also have an indirect effect     might be expected to increase and to control algal prolif-
on corals through the release of dissolved compounds       eration. However, no consistent positive or negative trends
which have been shown to enhance microbial activity and     were detected in the numerical abundance or biomass of
lead to an increased occurrence of coral mortality (Smith    herbivores on Cousin, over time, or at the different moni-
et al. 2006). Mumby (2006) described this cycle of events    toring sites. Other studies have also found little evidence to
based on the results of model simulations of Caribbean      indicate that the numerical abundance and biomass of
reefs, whereby coral mortality led to an increase in the     herbivorous fishes increased following increases in turf
amount of space available for algae and a decrease in the    algae (Hart et al. 1996; Spalding and Jarvis 2002). How-
grazing intensity on any given patch of reef. Reduced      ever, Russ (2003) found a significant positive correlation
grazing intensity resulted in an increase in macroalgal     between grazer biomass and algal productivity on the Great
cover and therefore an increase in juvenile coral mortality   Barrier Reef. and other studies have shown that an increase
(Mumby 2006). The changes on Cousin probably provide       in the abundance and productivity of algal resources may
evidence for a similar positive feedback loop in which an    result in an increase in grazing rates and in the numerical
initial decline in coral cover due to bleaching provided     abundance and biomass of herbivores if they were formerly
more space for epilithic algae, which developed into       food limited (Carpenter 1990; Robertson 1991; McClana-
macroalgae and dominated the benthos due to insufficient     han et al. 2000; Lindahl et al. 2001; Williams et al. 2001;
herbivory.                            Sheppard et al. 2002; Mumby et al. 2005; Garpe et al.
  There are several other factors which may also have      2006). Clearly not all herbivorous fish populations will be
impeded coral recovery on Cousin and caused further       food limited and other changes in the reef benthos, such as
declines in coral cover noted from 1998 to 2005. While      the loss of habitat complexity following coral mortality can
Seychelles’ reefs are located on a shallow continental      also influence the numerical abundance and biomass of
plateau, they are relatively isolated from other reef sys-    fishes. This has certainly been the case on Cousin and may
tems, suggesting coral populations are largely reliant on    help to explain why herbivores have not increased in
self-recruitment. The small and disconnected nature of      numerical abundance, despite the observed increase in
many coral brood stocks post-1998 is likely to have re-     epilithic algae.
duced the supply of coral larvae (Hughes and Tanner 2000;      Habitat complexity can reduce competition and preda-
Ayre and Hughes 2004). In addition, while the role of      tion (Buchheim and Hixon 1992; Hixon and Beets 1993)
coral-feeding fishes in reef recovery is largely unknown,     and there are locations where the maintenance of reef
they have been implicated in undermining reef resilience     structure following coral mortality has sustained abundant
(West and Salm 2003). While some reef fishes are obligate     and diverse fish populations (Lindahl et al. 2001; Riegl
coral feeders (Kokita and Nakazono 2001; Pratchett et al.    2002). Yet other studies have been less conclusive; Almany
2004) others, notably several species of scarids, consume    (2004) found the effect of habitat complexity on predation
live coral in addition to other resources (Bellwood and     and competition was more variable and depended on fac-
Choat 1990; Bythell et al. 1993; Rotjan and Lewis 2005).     tors such as the availability of appropriate shelter sites and
Rotjan et al. (2006) demonstrated that scarid grazing had    the behavioural attributes of predators. Habitat complexity
the potential to reduce coral fitness and retard the recovery                  ¨
                                 can also affect recruitment (Ohman et al. 1998) and the loss
of coral colonies following bleaching. McClanahan et al.     of complexity may have contributed to the increased


                                                          123
650                                                Coral Reefs (2007) 26:641–653


dominance of larger fishes on Cousin as smaller size        dimension to the complexity of coral reef recovery and
classes were limited by the availability of recruitment sites.  resilience following phase shifts.
This trend is concerning as it is likely to result in declines    While the importance of MPA networks and connec-
in numerical abundance and biomass of herbivorous fishes      tivity has been recognised (Lubchenco et al. 2003;
in the future, and further restrain their role in algal removal  Palumbi 2003; Ayre and Hughes 2004) the applicability
(Graham et al. 2006, 2007).                    of these concepts to remote reefs has yet to be deter-
  That the dramatic phase shift from coral to macroalgal     mined. Moreover, the global scale of disturbance events
dominance on Cousin took place in an established and well     such as climate change, and the large dispersal distances
enforced MPA with a fully protected herbivorous fish guild     of many larvae, mean that even the largest MPAs may not
highlights some questions regarding the role of spatial                   ¨
                                 be self-sustaining (Nystrom and Folke 2001; Bellwood
closures in buffering the effects of external disturbance     et al. 2004). Consequently, the localised benefits of small
events. Several studies have highlighted the fact that      MPAs may become ineffective if such areas do not have
grazing by herbivorous fishes influences competitive        the resilience to recover from global disturbance events.
interactions between corals and macroalgae (Lirman 2001;     As a means of mitigating biodiversity losses from coral
Williams and Polunin 2001; Mumby et al. 2006a; Paddack      bleaching, West and Salm (2003) suggest that areas where
et al. 2006) and Mumby et al. (2006b) demonstrated that      environmental conditions enhance resistance and resil-
MPAs can enhance grazing and reduce macroalgal cover.       ience to bleaching are incorporated into MPA networks.
However, the complexity of trophic interactions on coral     Indeed, coral recovery in the Seychelles has been shown
reefs and the variable role of predation in structuring reef   to be highly site specific and influenced by local factors
fish communities mean that spatial closures do not have a     such as water quality and upwelling (Engelhardt 2004). In
consistent effect on the numerical abundance and biomass     the Seychelles as a whole, granitic reef habitats appear to
of herbivores (Jennings and Polunin 1997; Graham et al.      have recovered most successfully from the 1998 bleach-
2003, 2005; Dulvy et al. 2004; Mumby et al. 2006b). The      ing (Graham et al. 2006) and the protection of such areas
size of MPAs is likely to be an important factor deter-      may represent a means of preserving coral biodiversity
mining their role in the promotion of resilience, although    (Engelhardt 2004) and enhancing resilience at a local
the numerical abundance and biomass of herbivorous fishes     scale.
has been shown to be enhanced even in very small MPAs        With the frequency of coral bleaching in the Indian
(Halpern 2003), probably reflecting the strong site attach-    Ocean predicted to increase in the future (Sheppard 2003)
ment of many herbivorous fishes (Chapman and Kramer        the prospects for reef recovery to a coral dominated state
2000). Indeed, surveys in 1994 indicated that the effective    on Cousin are not encouraging. Reef recovery following
protection of Cousin MPA from fishing had maintained a       disturbance can no longer be taken for granted (Nystrom ¨
higher biomass and diversity of herbivores than in fished     et al. 2000) and hypothesised outcomes of increases in the
areas (Jennings et al. 1996).                   occurrence of coral bleaching include changes in coral
  The phase shift observed in the 1.2 km2 Cousin MPA       community structure or set backs to early successional
must be considered in the context of concurrent declines in    stages or alternate states (Done 1999; Hoegh-Guldberg
coral reef resilience in the Seychelles as a whole (Graham    1999). The lack of resilience in Cousin MPA and the
et al. 2006). Even if the protection of Cousin had afforded    consequent phase shift from coral to macroalgal dominated
more resilience to this small area by maintaining a healthy    reefs would appear to support such predictions and even if
herbivorous fish guild, the degradation of other reef habi-    coral recruitment does increase, the abundance of algae in
tats within the Seychelles may have caused declines once a    this community may retard coral recovery. The MPA ad-
critical threshold was passed (Hughes et al. 2005). Fur-     hered to many of the current paradigms regarding effective
          ¨
thermore, it is naıve to assume that the protection of her-    coral reef management; local anthropogenic stressors were
bivorous species which can prevent phase shifts by keeping    virtually non-existent and effective enforcement of the
epilithic algae cropped down will result in the reversal of a   MPA had led to healthy populations of herbivorous fishes
phase shift once macroalgae have become established        (Jennings et al. 1996). The fact that a dramatic phase shift
(Bellwood et al. 2006). Most herbivorous fishes avoid       accompanied by a collapse in reef structure still took place
macroalgae and a recent study on the Great Barrier Reef      could be taken as an indication that such small-scale pro-
found a ‘sleeping functional group’, represented by a single   tected areas may not be successful on their own. Perhaps a
species of batfish, that usually consumes benthic inverte-     larger scale approach involving networks of appropriately
brates or plankton, was almost solely responsible for       located MPAs (Lubchenco et al. 2003) based on an
removing macroalgae and facilitating reef recovery (Bell-     appreciation of the complexities inherent in the dynamics
wood et al. 2006). It is possible that other functional groups  of coral reef recovery following disturbance (Bellwood
may be present in the Seychelles, imparting an additional     et al. 2006) would be more appropriate.


123
Coral Reefs (2007) 26:641–653                                                      651

Acknowledgments Funding for this work was provided by Nature      Done TJ (1999) Coral community adaptability to environmental
Seychelles under the Global Environment Facility project ‘Improving     change at the scales of regions, reefs and reef zones. Am Zool
management of NGO and privately-owned islands of high biodiver-       39:66–79
sity value in Seychelles’, Newcastle University, the British Overseas  Dulvy NK, Polunin NVC, Mill AC, Graham NAJ (2004) Size
Development Administration (now Department for International         structural change in lightly exploited coral reef fish communi-
Development), the Leverhulme Trust, the Western Indian Ocean         ties: evidence for weak indirect effects. Can J Fish Aquat Sci
Marine Science Association (WIOMSA) and the Fisheries Society of       61:466–475
the British Isles. We thank Nature Seychelles for logistical support                ¨
                                    Elmqvist T, Folke C, Nystrom M, Peterson G, Bengtsson J, Walker B,
and V.R. Johnson and the wardens of Cousin Island for assistance in     Norberg J (2003) Response diversity, ecosystem change, and
the field.                                  resilience. Front Ecol Environ 1:488–494
                                    Engelhardt U (2004) The status of scleractinian hard coral and reef
                                       fish communities 6 years after the 1998 mass coral bleaching
                                       event GEF SEYMEMP final report. Reefcare International Pty
References                                  Ltd, Townsville
                                    Froese R, Pauly D (2006) Fishbase. http://www.fishbase.org
Almany GR (2004) Does increased habitat complexity reduce                ˆ ´
                                    Gardner TA, Cote IM, Gill JA, Grant A, Watkinson AR (2003) Long-
   predation and competition in coral reef fish assemblages? Oikos     term region-wide declines in Caribbean corals. Science 301:958–
   106:275–284                               960
Ayre DJ, Hughes TP (2004) Climate change, genotypic diversity and                        ¨
                                    Garpe KC, Yahya SAS, Lindahl U, Ohman MC (2006) Long-term
   gene flow in reef-building corals. Ecol Lett 7:273–278          effects of the 1998 coral bleaching event on reef fish assem-
Barlow GW (1974) Contrasts in social behaviour between Central        blages. Mar Ecol Prog Ser 315:237–247
   American cichlid fishes and coral-reef surgeon fishes. Am Zool    Goreau T, McClanahan T, Hayes R, Strong A (2000) Conservation of
   14:9–34                                 coral reefs after the 1998 global bleaching event. Conserv Biol
Bellwood DR (1995) Direct estimate of bioerosion by two parrotfish      14:5–15
   species, Chlorurus gibbus and C. sordidus, on the Great Barrier  Graham NAJ, Evans RD, Russ GR (2003) The effects of marine
   Reef, Australia. Mar Biol 121:419–429                  reserve protection on the trophic relationships of reef fishes on
Bellwood DR, Choat JH (1990) A functional analysis of grazing in       the Great Barrier Reef. Environ Conserv 30:200–208
   parrotfishes (family Scaridae): the ecological implications.    Graham NAJ, Dulvy NK, Jennings S, Polunin NVC (2005) Size-
   Environ Biol Fish 28:189–214                      spectra as indicators of the effects of fishing on coral reef fish
Bellwood DR, Hoey AS, Choat JH (2003) Limited functional           assemblages. Coral Reefs 24:118–124
   redundancy in high diversity systems: resilience and ecosystem   Graham NAJ, Wilson SK, Jennings S, Polunin NVC, Bijoux JP,
   function on coral reefs. Ecol Lett 6:281–285              Robinson J (2006) Dynamic fragility of oceanic coral reef
Bellwood DR, Hughes TP, Hoey AS (2006) Sleeping functional          systems. Proc Natl Acad Sci USA 103:8425–8429
   group drives coral reef recovery. Curr Biol 16:2434–2439      Graham NAJ, Wilson SK, Jennings S, Polunin NVC, Robinson J,
                      ¨
Bellwood DR, Hughes TP, Folke C, Nystrom M (2004) Confronting        Bijoux JP, Daw TM (2007) Lag effects in the impacts of mass
   the coral reef crisis. Nature 429:827–833                coral bleaching on coral reef fish, fisheries, and ecosystems.
Birkeland C (1977) The importance of rate of biomass accumu-         Conserv Biol ‘in press’
   lation in early successional stages of benthic communities to   Halpern BS (2003) The impact of marine reserves: do reserves work
   the survival of coral recruits. Proc 3rd Int Coral Reef Symp      and does reserve size matter? Ecol Appl 13:S117–S137
   1:15–21                              Hart AM, Klumpp DW, Russ GR (1996) Response of herbivorous
Birrell CL, McCook LJ, Willis BL (2005) Effects of algal turfs and      fishes to crown-of-thorns starfish Acanthaster planci outbreaks.
   sediments on coral settlement. Mar Pollut Bull 51:408–414        II: Density and biomass of selected species of herbivorous fish
Bruggemann JH, van Oppen MJH, Breeman AM (1994) Foraging by         and fish-habitat correlations. Mar Ecol Prog Ser 132:21–30
   the stoplight parrotfish Sparisoma viride. I. Food selection in   Hay ME, Fenical W, Gustafson K (1987) Chemical defense against
   different, socially determined habitats. Mar Ecol Prog Ser       diverse coral-reef herbivores. Ecology 68:1581–1591
   106:41–55                             Hiatt RW, Strasburg DW (1960) Ecological relationships of the fish
Buchheim JR, Hixon MA (1992) Competition for shelter holes in the      fauna on coral reefs on the Marshall Islands. Ecol Monogr
   coral-reef fish Acanthemblemaria spinosa Metzlaar. J Exp Mar       30:65–127
   Biol Ecol 164:45–54                        Hixon MA, Beets JP (1993) Predation, prey refuges, and the structure
Bythell JC, Gladfelter EH, Bythell M (1993) Chronic and catastrophic     of coral-reef fish assemblages. Ecol Monogr 63:77–101
   natural mortality of three common Caribbean reef corals. Coral   Hoegh-Guldberg O (1999) Climate change, coral bleaching and the
   Reefs 12:143–152                            future of the world’s coral reefs. Mar Freshw Res 50:839–866
Carpenter RC (1990) Mass mortality of Diadema antillarum. II:      Hughes TP, Tanner JE (2000) Recruitment failure, life histories, and
   Effects on population densities and grazing intensity of        long-term decline of Caribbean corals. Ecology 81:2250–2263
   parrotfishes and surgeonfishes. Mar Biol 104:79–86          Hughes TP, Bellwood DR, Folke C, Steneck RS, Wilson J (2005)
Chapman MR, Kramer DL (2000) Movements of fishes within and          New paradigms for supporting the resilience of marine ecosys-
   among fringing coral reefs in Barbados. Environ Biol Fish        tems. Trends Ecol Evol 20:380–386
   57:11–24                              Hughes TP, Baird AH, Bellwood DR, Card M, Connolly SR, Folke C,
Choat JH (1991) The biology of herbivorous fishes on coral reefs. In:     Grosberg R, Hoegh-Guldberg O, Jackson JBC, Kleypas J, Lough
   Sale PF (ed) The ecology of fishes on coral reefs. Academic                   ¨
                                       JM, Marshall P, Nystrom M, Palumbi SR, Pandolfi JM, Rosen B,
   Press, London, pp 120–155                        Roughgarden J (2003) Climate change, human impacts, and the
Choat JH, Robbins WD, Clements KD (2002) The trophic status of        resilience of coral reefs. Science 301:929–933
   herbivorous fishes on coral reefs. II. Food processing modes and  Ivlev VS (1961) Experimental ecology of the feeding of fishes. Yale
   trophodynamics. Mar Biol 145:445–454                  University Press, New Haven
Clarke KR, Gorley RN (2006) PRIMER v6 User Manual/Tutorial.       Jennings S (1998) Cousin Island, Seychelles: a small, effective and
   PRIMER-E Ltd, Plymouth, UK                       internationally managed marine reserve. Coral Reefs 17:190


                                                                 123
652                                                      Coral Reefs (2007) 26:641–653

Jennings S, Polunin NVC (1997) Impacts of predator depletion by       trophic cascades, and the process of grazing on coral reefs.
   fishing on the biomass and diversity of non-target reef fish       Science 311:98–101
   communities. Coral Reefs 16:71–82                    ¨
                                    Nystrom M, Folke C (2001) Spatial resilience of coral reefs.
Jennings S, Grandcourt EM, Polunin NVC (1995) The effects of         Ecosystems 4:406–417
   fishing on the diversity, biomass and trophic structure of        ¨
                                    Nystrom M, Folke C, Moberg F (2000) Coral reef disturbance and
   Seychelles’ reef fish communities. Coral Reefs 14:225–235        resilience in a human-dominated environment. Trends Ecol Evol
Jennings S, Marshall SS, Polunin NVC (1996) Seychelles’ marine        15:413–417
   protected areas: comparative structure and status of reef fish   ¨
                                    Ohman MC, Munday PL, Jones GP, Caley MJ (1998) Settlement
   communities. Biol Conserv 75:201–209                  strategies and distribution patterns of coral reef fishes. J Exp Mar
Jones GP, McCormick MI, Srinivasan M, Eagle JV (2004) Coral         Biol Ecol 225:219–238
   decline threatens fish biodiversity in marine reserves. Proc Natl  Paddack MJ, Cowen RK, Sponaugle S (2006) Grazing pressure of
   Acad Sci USA 101:8251–8253                       herbivorous coral reef fishes on low coral-cover reefs. Coral
Kokita T, Nakazono A (2001) Rapid response of an obligately         Reefs 25:461–472
   corallivorous filefish Oxymonacanthus longirostris (Monacan-     Palumbi SR (2003) Population genetics, demographic connectivity,
   thidae) to a mass coral bleaching event. Coral Reefs 20:155–158     and the design of marine reserves. Ecol Appl 13:S146–S158
Kuffner IS, Walters LJ, Becerro MA, Paul VJ, Ritson-Williams R,     Polunin NVC, Klumpp DW (1989) Ecological correlates of foraging
   Beach KS (2006) Inhibition of coral recruitment by macroalgae      periodicity in herbivorous reef fishes of the Coral Sea. J Exp Mar
   and cyanobacteria. Mar Ecol Prog Ser 323:107–117            Biol Ecol 126:1–20
Letourneur Y (1998) Length-weight relationship of some marine fish    Polunin NVC, Roberts CM (1993) Greater biomass and value of
         ´
   species in Reunion Island, Indian Ocean. Naga 21:37–49         target coral-reef fishes in two small Caribbean marine reserves.
Letourneur Y, Kulbicki M, Labrosse P (1998) Length-weight          Mar Ecol Prog Ser 100:167–176
   relationships of fishes from coral reefs and lagoons of New     Pratchett MS, Wilson SK, Berumen ML, McCormick MI (2004)
   Caledonia – an update. Naga 21:39–46                  Sublethal effects of coral bleaching on an obligate coral feeding
Lewis SM (1986) The role of herbivorous fishes in the organisation of     butterflyfish. Coral Reefs 23:352–356
   a Caribbean reef community. Ecol Monogr 56:183–200         Quinn G, Keough M (2002) Experimental design and data analysis for
       ¨
Lindahl U, Ohman MC, Schelten CK (2001) The 1997/1998 mass          biologists. Cambridge University Press, Cambridge
   mortality of corals: effects on fish communities on a Tanzanian   Reaser JK, Pomerance R, Thomas PO (2000) Coral bleaching and
   coral reef. Mar Pollut Bull 42:127–131                 global climate change: scientific findings and policy recommen-
Lirman D (2001) Competition between macroalgae and corals: effects      dations. Conserv Biol 14:1500–1511
   of herbivore exclusion and increased algal biomass on coral    Riegl B (2002) Effects of the 1996 and 1998 positive sea-surface
   survivorship and growth. Coral Reefs 19:392–399             temperature anomalies on corals, coral diseases and fish in the
Lubchenco J, Palumbi SR, Gaines SD, Andelman S (2003) Plugging a       Arabian Gulf (Dubai, UAE). Mar Biol 140:29–40
   hole in the ocean: the emerging science of marine reserves. Ecol  Robertson DR (1991) Increases in surgeonfish populations after mass
   Appl 13:S3–S7                              mortality of the sea urchin Diadema antillarum in Panama     ´
McAfee ST, Morgan SG (1996) Resource use by five sympatric          indicate food limitation. Mar Biol 111:437–444
   parrotfishes in the San Blas Archipelago, Panama. Mar Biol     Robertson DR, Polunin NVC, Leighton K (1979) The behavioural
   125:427–437                               ecology of three Indian Ocean surgeonfishes (Acanthurus
McClanahan TR, Muthiga NA, Mangi S (2001) Coral and algal          lineatus, A. leucosternon and Zebrasoma scopas): their feeding
   changes after the 1998 coral bleaching: interaction with reef      strategies, and social and mating systems. Environ Biol Fish
   management and herbivores on Kenyan reefs. Coral Reefs         4:125–170
   19:380–391                             Rotjan RD, Lewis SM (2005) Selective predation by parrotfishes on
McClanahan TR, Maina J, Starger CJ, Herron-Perez P, Dusek E         the reef coral Porites astreoides. Mar Ecol Prog Ser 305:193–
   (2005) Detriments to post-bleaching recovery of corals. Coral      201
   Reefs 24:230–246                          Rotjan RD, Dimond JL, Thornhill DJ, Leichter JJ, Helmuth B, Kemp
McClanahan TR, Bergman K, Huitric M, McField M, Elfwing T,          DW, Lewis SM (2006) Chronic parrotfish grazing impedes coral
      ¨
   Nystrom M, Nordemar I (2000) Response of fishes to algae         recovery after bleaching. Coral Reefs 25:361–368
   reduction on Glovers Reef, Belize. Mar Ecol Prog Ser 206:273–   Russ GR (2003) Grazer biomass correlates more strongly with
   282                                   production than with biomass of algal turfs on a coral reef. Coral
McCook LJ, Jompa J, Diaz-Pulido G (2001) Competition between         Reefs 22:63–67
   corals and algae on coral reefs: a review of evidence and     Russ GR, St.John J (1988) Diets, growth rates and secondary
   mechanisms. Coral Reefs 19:400–417                   production of herbivorous reef fishes. Proc 6th Int Coral Reef
McManus JW, Polsenberg JF (2004) Coral-algal phase shifts on coral      Symp 2:37–43
   reefs: ecological and environmental aspects. Prog Oceanogr     Sano M, Shimizu M, Nose Y (1984) Food habits of teleostean reef
   60:263–279                               fishes in Okinawa Island, Japan. The University Museum of the
Mumby PJ (2005) Patch dynamics of coral reef macroalgae under        University of Tokyo 25:1–128
   chronic and acute disturbance. Coral Reefs 24:681–692       Scheffer M, Carpenter SR (2003) Catastrophic regime shifts in
Mumby PJ (2006) The impact of exploiting grazers (Scaridae) on the      ecosystems: linking theory to observation. Trends Ecol Evol
   dynamics of Caribbean coral reefs. Ecol Appl 16:747–769         18:648–656
Mumby PJ, Hedley JD, Zychaluk K, Harborne AR, Blackwell PG       Sheppard CRC (2003) Predicted recurrences of mass coral mortality
   (2006a) Revisiting the catastrophic die-off of the urchin        in the Indian Ocean. Nature 425:294–297
   Diadema antillarum on Caribbean coral reefs: Fresh insights    Sheppard CRC, Spalding M, Bradshaw C, Wilson S (2002) Erosion
   on resilience from a simulation model. Ecol Model 196:131–148                               ˜
                                       vs. recovery of coral reefs after 1998 El Nino: Chagos reefs,
Mumby PJ, Dahlgren CP, Harborne AR, Kappel CV, Micheli F,          Indian Ocean. Ambio 31:40–48
   Brumbaugh DR, Holmes KE, Mendes JM, Broad K, Sanchirico      Smith JE, Shaw M, Edwards RA, Obura D, Pantos O, Sala E, Sandin
   JN, Buch K, Box S, Stoffle RW, Gill AB (2006b) Fishing,         SA, Smriga S, Hatay M, Rohwer FL (2006) Indirect effects of




123
Coral Reefs (2007) 26:641–653                                                      653

   algae on coral: algae-mediated, microbe-induced coral mortality.  Williams ID, Polunin NVC, Hendrick VJ (2001) Limits to grazing by
   Ecol Lett 9:835–845                           herbivorous fishes and the impact of low coral cover on
Spalding MD, Jarvis GE (2002) The impact of the 1998 coral          macroalgal abundance on a coral reef in Belize. Mar Ecol Prog
   mortality on reef fish communities in the Seychelles. Mar Pollut     Ser 222:187–196
   Bull 44:309–321                          Wilson SK, Graham NAJ, Pratchett MS, Jones GP, Polunin NVC
Steneck RS (1988) Herbivory on coral reefs: a synthesis. Proc 6th Int    (2006) Multiple disturbances and the global degradation of coral
   Coral Reef Symp 1:37–49                         reefs: are reef fishes at risk or resilient? Global Change Biol
Tanner JE (1995) Competition between scleractinian corals and        12:2220–2234
   macroalgae: an experimental investigation of coral growth,     Wilson SK, Graham NAJ, Polunin NVC (2007) Appraisal of visual
   survival and recruitment. J Exp Mar Biol Ecol 190:151–168        assessments of habitat complexity and benthic composition on
West JM, Salm RV (2003) Resistance and resilience to coral          coral reefs. Mar Biol 151:1069–1076
   bleaching: implications for coral reef conservation and manage-
   ment. Conserv Biol 17:956–967
Williams ID, Polunin NVC (2001) Large-scale associations between
   macroalgal cover and grazer biomass on mid-depth reefs in the
   Caribbean. Coral Reefs 19:358–366




                                                                 123