CHAPTER 4:
GLOBAL ALGAE-CORAL-HERBIVORE
LITERATURE COMPILATION AND COMPARISONS
The contents of this chapter are to be submittedfor publication
(planned submission September 2015)
4.1 Introduction
Coral reefs are among the most threatened ecosystems worldwide (Bellwood et al. 2004,
Wilkinson 2008) and are experiencing increasing multiple stresses. Not all reefs have been
affected by these stressors to the same degree (i.e. some have been more resilient than
others). For example, after the very severe coral bleaching in 1998, which was followed by
high mortality rates, some reefs recovered relatively rapidly, while others have failed to
return to coral dominance even after many years (Arthur et al. 2006, Ledlie et al. 2007).
Resilience is the ability of an ecosystem to absorb shocks, regenerate after disturbances, and
adapt to change without fundamentally changing the ecosystem structure and functions
(Nystrom et al. 2000). Understanding why some reefs are more resilient than others less has
become increasingly important for ensuring proper reef management and reef "health"
(Hughes et al. 2003).
Less resilient reefs have often shifted to an alternate state dominated by macroalgae
(Hughes 1994, Hunter and Evans 1995, McClanahan and Muthiga 1998, Graham et al.
2006). Herbivory therefore plays a critical role in coral reef resilience by controlling algal
communities and lowering competition between coral and benthic algae which may
otherwise out-compete corals (Lirman 2001, Jompa and McCook 2002). Herbivory also
affects the coral-algal balance by producing available substrate for the recruitment of new
coral larvae. In the face of climate change and increasing coral mortalities, herbivory has
thus become an extremely important factor in preventing shifts from coral dominated reefs
to macroalgal dominated reefs.
59
Many experimental studies show that when herbivores are removed, macroalgal cover
and biomass rapidly increases (see chapters 2 and 3). Thus a logical extension of these
results would be that the higher the levels of herbivore biomass present on a reef the lower
the cover and abundance of macroalgae, and vice versa. A recent review of the resilience of
Indo-Pacific and Atlantic coral reefs found such a negative linear relationship between
macroalgae cover and herbivore biomass (Roff et al. 2012). Their analysis also shows that
this linear relationship differs between the Atlantic and the Indo-Pacific region. However,
only few sites were included in the study, and the Indo-Pacific was only represented by sites
from Hawaii and the Philippines. To see if their results and conclusions withhold when more
sites are included, and to get a better understanding of the overall picture we present an
extensive overview and analysis of the herbivore-macroalgae relationships from a much
wider geographic range of coral reefs.
4.2 Material and Methods
Herbivore fish biomass, percent macroalgae cover, percent coral cover, habitat type
(forereef, reef crest, reef flat, back-reef, patch reef, lagoon, bank reef, channel), and depth
data (0-6.5 m, 6.5-10 m, 10-15 m, 15-20 m, 20-25 m, 25-30 m) were compiled from 991
sites from 31 studies from three different oceans (Table CI). To examine the relationship
between herbivore fish biomass and macroalgae percent cover a generalized linear model
(GLM) was used due to the non-normal distribution, and heteroscedasticity in the data
(GLMqb); with oceanic region, coral cover, habitat, and depth as covariates. Because
macroalgae cover data were expressed as percentages and high overdispersion was present, a
GLM with a logistic link function and a quasibinomial distribution of errors was used to fit
the data. Herbivore biomass was log-transformed (using loglp) and coral cover square root
transformed to reduce the influence of outliers. We used a stepwise forward selection
approach to select the model with optimal fit to the data. To estimate model fit we calculated
each model's quasi-Akaike information criterion (qAIC) as the quasibinomial regression
60
does not produce Akaike information criterion (AIC) values. An F-test with a critical
significance level established as 0.05 was used for assessing changes in residual deviance. In
cases where habitat and/or depth were significant factors, their levels which were not
significant were combined and the more complex model was tested against the simplified
model (with the merged levels) in each step.
Coral cover is closely associated with macroalgae via space availability and therefore we
did a second series of regression analyses where we pooled the data into 3 groups based on
coral cover: <20%, 20^-0%, and >40% coral cover, and analyzed these groups separately
for herbivore biomass and ocean effects. Adjusted McFadden's pseudo R-squared was
calculated to quantify the effect of the explanatory variables. All graphs and statistical
analyses were done using R software.
Due to the heterogeneous variances in the data we also used quantile regression analysis,
which was performed on the upper boundary of the conditional distribution of responses
(95th
and 99t h
percentile) for sites that had <20% coral cover. This method provides better
estimates of the limiting effect of the independent variable compared to the ordinary mean
regression estimates. We used the quantreg package and rq function to build the quantile
regressions.
4.3 Results
Macroalgal cover was negatively and significantly related to herbivore biomass and
exhibited triangular relationship (Fig. 4.1). High macroalgal cover was consistently
associated with low herbivore biomass, however low macroalgae cover was associated with
high herbivore biomass macroalgal cover (GLMqb: coral, Fi83i=162.6, p<<0.001; region,
^2,830=149,8, p«0.001; habitat, F2,828=7.74, p«0.001; depth, F2,826=6.5, p<0.01; herbivore
biomass, Fijg25=33.8, p<<0.001). The five variables combined explained 32.3% of
variability in macroalgae cover with coral cover explaining 15.0% (gross effect), region
61
9.7% (pure effect), habitat 4.1% (pure effect), depth 0.6% (pure effect) and herbivory (pure
effect), explaining only 2.9% of variability. In the full GLMqb model the Indian Ocean and
Pacific Ocean were not significantly different and were thus grouped analysed as the IndoPacific
region. All three oceans however differed significantly in macroalgae cover with the
highest values observed in Atlantic Ocean (mean 32.4%) and much lower in the IndoPacific,
but still significantly (GLMqb: p = 0.002) higher in Pacific (4.8%) than in Indian
Ocean (5.0%). The Atlantic Ocean also had lower levels of herbivore biomass; the Indian
Ocean despite having variable levels of herbivory had the lowest macroalgae cover, and the
Pacific Ocean was the most variable in both macroalgae cover and herbivore biomass (Fig.
4.2). Reef habitats were also significant predictors of macroalgae. The forereef and backreef
were similar (and thus merged into one habitat level in the overall GLMqb model) and
the reef crest and lagoon were similar (thus also merged into one habitat level in the overall
GLMqb model). These two habitat levels differed significantly from each other as well as
from the reef flat habitat. The depth of the reef was also a significant predictor and shallow
reefs (less than 6.5m) significantly differed from reefs deeper than 6.5m. For herbivore
biomass < 56 g m" macroalgal cover was extremely variable (Fig. 4.1) but when biomass
exceeded 50 to 80 g m" , macroalgal cover was low and independent of the region, coral
cover, habitat, or depth.
62
Fig. 4.1. Relationship between herbivore biomass and macroalgae cover for 1000 sites from
various places around the globe. The various colours represent different densities of sites.
Notice that many sites have both low herbivore biomass and low macroalgae cover. See Table
CI in the appendix material for the compiled data set.
The analysis based on the data pooled into the 3 coral cover groups (Fig. 4.3) found that
in the group with coral cover > 40%, macroalgae was consistently low and none of the
remaining factors except depth was significant (GLMqb: depth, F2 ,i5i=2.9, p<0.01). Depth
explained 8% of the variability. In the 20-40% coral cover group macroalgae was more
variable and only the region factor was significant (GLMqb: region, F\^i9= 64J, p<<0.001)
and explained 20% of the variability. Finally, in the low < 20% coral cover group, region,
habitat, and herbivore biomass were all significantly associated with macroalgae cover
(GLMqb: region, Fi44i=60.9, p<<0.001; habitat 7*3,438=14.2, p<<0.001, herbivore biomass,
63
^1,437=46.5, p<<0.001)) and all three factors explained 25.9% with region, habitat, and
herbivore biomass accounting for 10.5%, 7.4%, and 8% of variability respectively. The
slopes of the 95t h
and 99t h
quantile regressions were both significant and show that herbivory
does function as a limiting factor (Fig. 4.4).
100
80
60
40
20
0
g 100
o5 80 H
8 60
<D
a 40
ro
o
o
(0
20
0
100
80
60
40
20
0
Atlantic Ocean
Indian Ocean
*»
•
Pacific Ocean
1" I I I I I
0 50 100 150 200 250 300
Herbivore biomass (g m~2
)
40-100
3040
20-30
10-20
0-10
40-100
3040
20-30
10-20
0-10
40-100
3040
20-30
10-20
0-10
Atlantic Ocean
-UZh
-a.
Indian Ocean
I
- m
Pacific Ocean
• • *
• • • •
T 1 1 1 1 1 —
0 25 50 75 100 125
Herbivore biomass (g m~2
)
Fig. 4.2. Relationships between fish herbivore biomass and macroalgal cover in different
oceans displayed as scatterplots (on the left) and boxplots (on the right) seperately for the three
different oceans. Note the different scales. The black dots in the boxplot graph represent
outliers (not all outliers are depicted for the Pacific Ocean).
64
4.4 Discussion
Indo-Pacific reefs with low levels of herbivore biomass have been described as having
relatively low macroalgae cover compared to Atlantic reefs of the same herbivore levels
(Roff and Mumby 2012). The authors, therefore, suggested Indo-Pacific reefs are less
susceptible to shifts from coral to macroalgal dominance. While Atlantic reefs tend to have
less coral and more macroalgae than Indo-Pacific reefs and may be less resilient to
macroalgal dominance, our larger global data compilation indicates variable and complex
coral-macroalgae-herbivore relationships both within and between oceans (Fig. 4.2). In
contrast to the study of Roff and Mumby (2012) we found that, at the same low herbivore
levels, some reef sites in the Pacific Ocean have macroalgal cover similarly high to those in
the Atlantic Ocean. In some cases Pacific reefs even exceeded Atlantic reefs in macroalgae
cover (Fig. 4.2). This shows that not all Indo-Paific reefs are as resilient as assumed.
The Atlantic reefs, on the other hand, never reached such high herbivore levels as found
in the Indo-Pacific, which was expected given the history of overfishing in the Atlantic. The
low herbivore levels in the Atlantic are often presented as one of the main reasons for the
high macroalgae cover. Our analysis shows that on low coral cover reefs herbivory is indeed
an important limiting factor (Fig. 4.4). However, the compiled data showed great variability
in macroalgal cover across different oceans and herbivory levels (Fig. 4.1 and 4.2). This
variability indicates that the relationship between macroalgae cover and herbivory cannot be
described by a single rate of change and thus that the amount of macralgae present on reef is
likely governed by a large set of factors and their complex interactions.
65
a
100 •
>40% coral cover
80-
60-
40-
20 H
• o
>_o
n 1 1 1 —
0 100 200 300
100 -
21-40% coral cover
£ 80s
—
<D
O 60-
0
<D
CO
U} 40-
TO
1 20-
0 O i
o „ o
- 1 1 1 1 1
0 50 100 150 200
100 •
80-
60-
40-
20-
0<20%
coral cover
gpo o
0 50 100 150
Herbivore biomass (g rrf2
)
Fig. 4.3. Relationship between herbivore biomass and macroalgae cover for compiled data
from reefs in the Pacific (turquoise squares), Indian Ocean (blue circles), and Atlantic Ocean
(black triangles). Because macroalgae cover is closely associated with coral cover, the sites
66
were divided into three groups based on coral cover: coral cover over 40% (a), coral cover
between 2(M10% (b), coral cover less than 20% (c). See table CI in the appendix material for
the compiled data set.
Fig. 4.4. Relationship between herbivore biomass and macroalgae cover on sites with coral
cover <20%. Solid line represents ordinary least square regression; dashed line and dotted line
represent the 95th
and 99th
quantile regression, respectively. Shaded areas indicate the 95%
confidence intervals for each of the regression lines. 99th
percentile: p<<0.001, 95th
percentile:
p«0.001, OLS: p«0.001, R2
=0.04.
Although many Indo-Pacific sites with low herbivore biomass had a lot of macroalgae,
majority of the sites had very low macroalgae cover (Fig. 4.1), which was rather curious.
67
Given the extreme competition between coral and macroalgae and the fact that coral cover
was shown to be a significant factor in our regression analysis, the low macroalgal cover on
these reefs may be due to being constrained by coral cover, purely through spatial exclusion
(Williams et al. 2001). Thus we created 3 coral cover groups, in order to see whether these
low-herbivore low-macrolagae sites were only on reefs with high coral cover. High coral
cover sites really did have low macroalgae as expected, which was probably due to spatial
exclusion given that herbivory was not a significant controlling factor for these reefs
(GLMqb: 7*5158=2.9, p>0.05, Fig. 4.3). But with decreasing coral cover the variability in
macroalgae increased and thus some reefs with low herbivore biomass and low macroalgae
cover had low coral cover as well. How these low-coral low-herbivore reefs maintain low
macroalgae cover requires study beyond the descriptive work done here. Cheal et al. (2010)
suggest that herbivore fish diversity and the functional make-up of the herbivore community
are more important than herbivore biomass, especially where bare substrate is available.
Other possible governing factors may include: sea urchin abundance/biomass, nutrient
accessibility; intensity of ocean currents, waves and storms; and probably many other less
obvious factors - for example a study from the Line Islands (Sandin et al. 2008) found that
macroalgae was more abundant on islands with higher fishing pressure and low biomass of
predatory fishes and sharks. Unfortunately, none of these factors were available for the
compiled reef sites in this study. Another two possible factors, which were available for the
compiled sites, were depth and reef habitat type. From the two, habitat resulted as a
significant factor in our regression analyses where the forereef and back-reef habitats were
significantly different from the reef flat and also from the merged crest and lagoon habitat
level, whereas depth was significant mainly for the high coral cover reefs (i.e. reefs with
coral cover >40%). However, both helped explain only a very small portion of the
macroalgal variability. Another possible explanation for the low-coral low-herbivore sites
having such low macroalgae cover, is that they were dominated by other benthic organisms
such as turf, crustose coralline algae, soft coral, or other invertebrates. Although this was
probably true in some cases, it is less likely to be an explanation for all of the sites, given
their vast number.
This study shows that herbivore biomass, coral cover, habitat type, as well as
geographical affinity (i.e. Atlantic, Indian, and Pacific ocean), are all important factors
68
predicting the presence of macroalgae, however they all explained only a little over one third
of macroalgal variability, thus other factors than the ones examined here are likely to be
important. This high macroalgal variability was, however, only present for herbivore
-2 -2
biomass levels <50g m" . When herbivore biomass exceeded 50 g m" , reefs were (with few
exceptions) consistently associated with low macroalgal cover independent of the region,
coral cover or habitat (Fig. 4.1). Nevertheless, only 9.4% of compiled reefs reported
herbivore levels >50g m" , which brings into question the pervasiveness of what might be
considered baseline conditions.
Nevertheless, our global data compilation challenges generalizations about herbivoremacrolgae-coral
relationships by the high variability and low explanatory levels of the
commonly reported factors of herbivore biomass and coral cover. The compilation indicates
that there is a complex and variable relationship between macroalgae cover, geography,
coral cover, habitat, depth and herbivores, but many other poorly studied factors must also
be contributing to the variability. This stresses the importance of context when studying
coral-macroalgal shifts and when preparing management plans.
69