Fong et al 06
Ecology, 87(5), 2006, pp. 1162–1168
Ó 2006 by the Ecological Society of America
EPIPHYTIC CYANOBACTERIA MAINTAIN SHIFTS TO MACROALGAL
DOMINANCE ON CORAL REEFS FOLLOWING ENSO DISTURBANCE
PEGGY FONG,1,3 TYLER B. SMITH,2,4 AND MATTHEW J. WARTIAN1
1
Department of Ecology and Evolutionary Biology, University of California Los Angeles, 621 Young Drive South, Los Angeles,
California 90095-1606 USA
2
Division of Marine Biology and Fisheries, Rosenstiel School of Marine and Atmospheric Science, University of Miami,
4600 Rickenbacker Causeway, Miami, Florida 33149 USA
Abstract. Macroalgal dominance of some tropical reef communities in the Eastern Pacific
after coral mortality during the 1997–1998 El Nino Southern Oscillation (ENSO) was
˜
facilitated by protection from herbivory by epiphytic cyanobacteria. Our results do not
support that reduction in number of herbivores was a necessary precursor to coral reef decline
and shifts to algal reefs in this system. Rather, macroalgae dominated the community for
several years after this pulse disturbance with no concurrent change in herbivore populations.
While results of microcosm experiments identified the importance of nutrients, especially
phosphorus, in stimulating macroalgal growth, nutrient supply alone could not sustain
macroalgal dominance as nutrient-stimulated growth rates in our in situ experiments never
exceeded consumption rates of unprotected thalli. In addition, thalli with nutrient-enriched
tissue were preferentially consumed, possibly negating the positive effects of nutrients on
growth. These tropical reefs may be ideal systems to conduct experimental tests distinguishing
phase shifts from alternative stable states. Shifts were initiated by a large-scale disturbance
with no evidence of a changing environment except, perhaps, dilution in herbivory pressure
due to increased algal cover. Community establishment was most likely stochastic, and the
community was likely maintained by strongly positive interaction between macroalgal hosts
and cyanobacterial epiphytes that uncoupled consumer control of community structure.
Key words: alternative stable states; coral reefs; facilitation; herbivory; macroalgal bloom; nutrients;
phase shifts.
INTRODUCTION shifts to algal domination, the underlying mechanisms
Investigating mechanisms causing and maintaining causing and maintaining these shifts are hotly debated
shifts among different states of ecological communities (e.g., Jackson et al. 2001, Aronson et al. 2003, Hughes et
has been a central focus of research for many years (e.g., al. 2003, Pandolfi et al. 2003). Are they alternative stable
Estes and Palmisano 1974, Sutherland 1974). Current states where two communities exist under the same set of
debate centers on whether these shifts are between environmental conditions, or phase shifts from one state
alternative stable states or are phase shifts due to to another caused by a threshold response to change in
underlying differences in environmental conditions, and an environmental parameter? Most current evidence
under what conditions shifts are stochastic or determin- points to the latter for coral reefs (reviewed in Petraitis
istic processes (e.g., Beisner et al. 2003, Bertness et al. and Dudgeon 2004b), with some claiming that reduction
2004a, Petraitis and Dudgeon 2004a). Despite much in consumer populations via harvesting of herbivorous
debate, community shifts initiated by disturbance, fishes is a necessary first step in this phase shift (e.g.,
frequently mediated by strong biotic interactions, and Jackson et al. 2001). They reason that the force of
often sustained by accompanying changes in abiotic herbivory is so strong in coral reef ecosystems with
factors such as supply of nutrients or changes in physical intact herbivore populations that herbivores virtually act
habitat structure, have been documented in freshwater, as lawnmowers, limiting algal biomass accumulation
terrestrial, and marine communities (reviewed in and negating the impact of any other factor that may
Schroder et al. 2005).
¨ enhance algal productivity and growth, thus maintain-
While most coral reef ecologists agree coral commun- ing the competitive dominance of slower growing corals.
ities are undergoing worldwide declines with concurrent Once herbivores are removed, faster-growing algae
become competitively dominant and a phase-shift
Manuscript received 16 May 2005; revised 4 October 2005; occurs. Others argue that reduced herbivory is not
accepted 22 November 2005. Corresponding Editor: P. D. always needed to initiate a phase shift attributable to
Steinberg. altered outcomes of competition. One alternative
3
E-mail: pfong@biology.ucla.edu
4
Present address: Center for Marine and Environmental
mechanism is for increased nutrient supplies to stimulate
Studies, University of the Virgin Islands, 2 John Brewers rapid growth and allow algae to escape control by
Bay, St. Thomas, U.S. Virgin Islands 00802 USA. herbivores, biomass to accumulate, a change in com-
1162
May 2006 CYANOBACTERIA MAINTAINS SHIFT TO ALGAE 1163
petitive dominance from coral to algae, and ultimately a of nutrient-rich bottom water, but persisted for years
phase shift from coral to algae (Lapointe 1997). past this environmental fluctuation. Herbivore popula-
A plethora of evidence supports the role of herbivores in tions have been monitored for over 25 years on Uva
limiting algal biomass on coral reefs; in contrast, the role Island reef, with no overall trend of reduction in
of nutrients has been less well established (reviewed in numbers (Glynn 1990, 2004). In this study, we quantified
Szmant 2002). Overall, results of laboratory and micro- benthic dominance of the A. spicifera–cyanobacterial
cosm experiments demonstrate that reef algae are nutrient association on our two study reefs, and investigated
limited (e.g., Fong et al. 2003), while field experiments several factors that may sustain their dominance after
often have mixed results (e.g., Hatcher and Larkum 1983). ENSO disturbance, including rates of herbivory, pro-
Interpretation of these experiments is limited, in part, by tection from herbivory by cyanobacterial epiphytes, and
the difficulty of relating results of laboratory or micro- the potential for nutrient stimulation.
cosm studies of the effects of nutrient addition to natural
growth in high-energy, high-flow environments typical of METHODS
reefs, and the related methodological challenge of Benthic cover of the A. spicifera–cyanobacterial
effectively enriching in situ experiments in these same association, total coral, and total algae on our study
environments (reviewed in McCook 1999). In addition, a reefs in the Panamanian Gulf of Chiriquı´ was measured
recent study suggests that, on the local scale, herbivores on Uva Island reef in 1999 and 2000 and on the Cavada
may selectively consume more opportunistic, fast-growing Island reef in 2001 and 2002. The cyanobacterial
macroalgae that accumulate nutrient reserves in their epiphyte was putatively in the genus Lyngbya. Total
tissues when enriched (Boyer et al. 2004), further algae was the sum of macroalgae, crustose corallines,
confounding in situ enrichment experiments. There is and algal turfs. On Cavada reef, 1-m2 quadrats were
similar evidence for selective consumption of enriched placed at regular intervals (5 m) along cross-reef-zone
seagrass (McGlathery 1995). Thus, studies assessing the transects. On Uva reef, 1-m2 quadrats were placed
roles of both herbivores and nutrients in sustaining algal randomly along 30-m transects within reef zones in the
communities on coral reefs are important. southern portion of the reef. Percentage of cover was
An alternative view from changes in top-down or calculated by determining the cover category underlying
bottom-up processes leading to phase shifts is that 81 points within the 1-m2 quadrat.
complex multiple stresses to coral communities may act We assessed the importance of herbivory on biomass
alone or in concert to cause shifts to alternative accumulation of A. spicifera by conducting two field
community states on tropical reefs (Knowlton 1992, experiments on our two study reefs. In February 2001 on
Aronson et al. 2003, 2004). For these shifts to be Cavada reef, three areas (3–5 m apart) on the reef crest
considered alternative stable states, theory requires that were haphazardly selected. Within areas, paired plots
causative factors initiating changes must be pulse were assigned to either open or caged treatments. Within
disturbances (Connell and Sousa 1983, Peterson 1984, each plot, two thalli were marked with cable ties and
Petraitis and Dudgeon 2004b; but see Beisner et al. height measured to the nearest millimeter. We did not
2003); coral reef examples include El Nino Southern
˜ remove epiphytes during this experiment. Cages were 20
Oscillation (ENSO)-related mortality (Glynn et al. 2001) 3 20 cm (height 3 diameter) cylinders of high-density
and destructive fishing practices (Fox et al. 2005). Press polyethylene with a mesh size of 1.4 cm (Memphis Net
forces that may act in concert with pulse disturbances to and Twine, Memphis, Tennessee, USA) attached to
sustain algal dominance may include reduction in three rebar stakes driven into the substrate. The experi-
herbivory pressure and increases in nutrient supplies, ment lasted six days, when marked thalli were remea-
resulting in the proliferation of opportunistic, fast- sured. Percent increase in height was averaged for the
growing algae on newly opened space and/or dominance two thalli per plot, and the averages used in a paired t
by algae that is physically or chemically defended from test to detect differences in means between caged and
herbivory (reviewed in Knowlton 1992, Szmant 2002). uncaged plots. A mark was lost from only one thallus,
Clearly, there is a need for studies that investigate the and in this case one change in height was used in the
complex and potentially interacting ecological forces analysis. Another experiment was conducted on Uva
that maintain shifts to algal dominated tropical reefs Island reef in May 2000. A. spicifera thalli were collected
after large-scale pulse disturbances. from the bloom area, cleaned of macroepiphytes, placed
Blooms of Acanthophora spicifera (Rhodophyta) in mesh bags, spun for 1 min in a salad spinner to
covered by cyanobacterial epiphytes have been observed remove excess water, and weighed. Samples of 5 g wet
on several reefs in the Eastern Tropical Pacific since weight were cable tied together and attached to weighted
widespread coral mortality opened substrate for colo- ropes. Ropes with algae were anchored in seven different
nization during the 1997–1998 ENSO. These reefs were reef areas from the seaward reef base across the reef to
decimated by the 1982–1983 ENSO and suffered further the landward backreef; n ¼ 7 replicates for each area.
coral mortality during the 1997–1998 event (Glynn et al. Herbivory rate was calculated as loss in grams over 24
2001). Bloom initiation coincided with La Nina con- ˜ hours. Data were analyzed with one-way ANOVA (with
ditions, with thermocline shoaling and mixing upwards location as the factor) after being tested to be sure they
1164 PEGGY FONG ET AL. Ecology, Vol. 87, No. 5
met ANOVA assumptions. Fisher’s protected least from the Cavada Island reef in May 2000, we conducted a
significant difference (PLSD) tests were conducted after 2 3 2 factorial experiment with five replicates. Exper-
a significant ANOVA. imental treatments were ambient seawater with addition
Two experiments were conducted to assess the effects of 20 lmol/L NO3, 2 lmol/L PO4, 20 lmol/L NO3 þ 2
of epiphytic cyanobacteria on growth of and herbivory lmol/L PO4, and an ambient seawater control (abbre-
rates on A. spicifera in the bloom area of Uva Island viated as þN, þP, þNþP, and C, respectively). Exper-
reef. In March of 2003, we chose 20 plots (25 3 25 cm) imental units were plastic jars containing 500 mL of
within the visually homogeneous rubble area dominated treatment seawater and 5 g wet weight of algae. Algal
by algae covered in epiphytes. Epiphytes were gently thalli were collected, cleaned of macroepiphytes, weighed
removed (brushing by hand) from thalli in 10 randomly as before, and trimmed to the correct size to include
chosen plots, and left intact on algae in the remaining 10 several apical growing tips. Jars were placed in a random
plots. Percentage of cover of algae was estimated using array in an outdoor flow-through water table to maintain
the point–intercept method (determining the percentage ambient temperature. Because experimental units were
of points in a grid that overlaid the A. spificera– static and not flow-through, this experiment should be
cyanobacteria association) initially and after five days considered a ‘‘bioassay,’’ or a simple attempt to determine
and the change in percentage of cover was calculated. if N or P is limiting, rather than an attempt to predict
Heights of 16 randomly selected thalli in each plot were growth responses of algae in the field. Thus, growth rates
measured and averages calculated. Initial averages were can be compared within the experiment, but should not be
subtracted from final averages to calculate the change in extrapolated to natural systems. Jars were screened to
height per plot after five days. T tests were used to reduce ambient light by approximately 30%. The experi-
compare the mean change in percentage of cover and ment lasted three days, then algal biomass was again
height between cleaned and epiphytized thalli. measured and percentage of change calculated. A two-
To compare the potential cost of epiphytism (e.g., way ANOVA tested for differences in means after data
reduced growth due to shading, nutrient competition) to were tested to be sure they met ANOVA assumptions.
benefits of epiphytism as protection from herbivory for To determine if added nutrients would enhance
A. spicifera, in March of 2004 we conducted a 2 3 2 growth of in situ algae on Uva Island reef in March
factorial experiment varying protection from herbivory 2002, we exposed algal thalli to ambient and enriched
(with vs. without cages) and removal of epiphytes conditions across four reef zones. Algae were collected
(cleaned vs. epiphytized). Treatments were randomly and weighed (as already described), placed in mesh bags
assigned to plots within the rubble area where the bloom (nylon window screening) to protect them from herbi-
occurred. Because the rubble benthos was topographi- vores, and 10 replicates were anchored in each of four
cally complex, open-bottomed cages were not considered reefs zones (slope, base, crest, and flat). Five grams of
adequate to protect thalli from herbivores. Thus, cages slow-release fertilizer (Osmocote [Scots Miracle-Gro
(constructed as above) were 15 3 10 cm (height 3 Company, Marysville, Ohio, USA], 15% N þ 5% P)
diameter) complete cylinders with both tops and were placed in nylon stockings and attached to five of
bottoms, while open plots consisted of bottoms only; the replicates in each reef zone. Samples were anchored
all experimental units were held to the benthos with rebar at least 2 m apart to reduce cross-contamination by
stakes. Three to four rubble pieces with attached algae nutrients. Preliminary data (P. Fong, unpublished data)
were placed in each cage or on each open plot base, and demonstrated that our method of nutrient enrichment
heights of four or five algal thalli were measured to the was undetectable (measured as storage in algal tissue) at
nearest millimeter. After five days, the change in height .1 m distance. Growth after four days was calculated as
was calculated as described in the last paragraph. Data above and subjected to a two-way mixed-model
were analyzed with two-way ANOVA after being tested ANOVA (with the fixed factor of ambient vs. enriched
to be sure they met ANOVA assumptions. Comparison and the random factor reef zone).
of algal growth of cleaned vs. epiphytized thalli within To investigate whether herbivores consumed more
cages provided an estimate of the costs of being enriched algae, in May 2000 we performed the above
epiphtytzed without the confounding effect of herbivory. experiment with three variations. First, after collecting
Comparison of algal net growth in cleaned vs. epiphy- the algae we cultured half for three days with slow-
tized open plots subject to herbivores estimated the release fertilizer in batch culture. The other half were
benefits of protection provided by the epiphytes. Cage kept in ambient seawater. Initial tissue N was greater in
controls consisting of cages with no tops for the enriched (2.60% 6 0.04% [mean 6 SE]) than in ambient
epiphytized treatment only demonstrated that cage (2.40% 6 0.04%) algae (t test, df ¼ 8, P ¼ 0.006), but
effects were not significant (t test comparing epiphytized differences in P (0.14% 6 0.004% enriched vs. 0.13% 6
open to cage control, df ¼ 13, t value ¼ 0.738, P ¼ 0.4737). 0.003% ambient) were not significant. Second, algae
We explored the role of nutrients in sustaining algal were exposed to herbivores as they were anchored
growth and biomass accumulation using three experi- without being enclosed in mesh bags. Finally, instead of
mental approaches. To determine if nitrogen (N) or testing effects across reef zones, we chose four locations
phosphorus (P) limited growth of A. spicifera collected within the reef slope, where herbivory rates measured
May 2006 CYANOBACTERIA MAINTAINS SHIFT TO ALGAE 1165
TABLE 1. Benthic cover of the Acanthophora spicifera–cyano- A. spicifera maintained net biomass accumulation in
bacterial association, total coral, and total algae on study the presence of herbivores only when it was covered with
reefs.
cyanobacterial epiphytes on Uva Island reef (Fig. 2A and
Cover (%) B). When comparing algae in cleaned vs. epiphytized
plots in 2003, epiphytized algae increased both cover by
A. spicifera– Total Total
Reef, year, and zone cyaonobacteria coral algae ;5% (t test; df ¼ 18, t ¼À2.859, P ¼ 0.0104) and thallus
height by ;2.5 cm (t test; df ¼ 18, t ¼À7.603, P , 0.0001)
Cavada Island 2001 over five days. When epiphytes were removed, cover was
All zones 59.5 6 4.0 19.8 6 3.5 74.6 6 3.7 reduced by almost 10% and thalli did not grow taller.
Cavada Island 2002 The effect of epiphytes on biomass accumulation of A.
All zones 66.1 6 5.8 3.9 6 1.7 nd spicifera thalli, measured as change in height, differed
Uva Island 1999 between caging treatments, resulting in interaction
Base 22.2 6 9.9 14.8 6 5.4 65.2 6 7.4 (Table 2A, Fig. 2C). Within cages, epiphytes reduced
Slope 0.8 6 0.5 26.5 6 4.6 57.3 6 6.4
Flat 1.3 6 0.7 14.3 6 4.6 77.54 6 8.1 thallus growth by ;50%. This estimated the cost in
growth to the algae of bearing epiphytes. In contrast,
Uva Island 2000
Base 77.2 6 6.8 9.3 6 5.1 89.8 6 4.9
only thalli that were epiphytized maintained positive
Slope 0.3 6 0.3 4.0 6 3.6 74.0 6 5.5 biomass accumulation in open plots exposed to herbi-
Flat 0.0 6 0.0 9.9 6 3.2 84.6 6 3.8 vores. This comparison between cleaned and epiphytized
Notes: Total algae is the sum of macroalgae, crustose thalli exposed to herbivores estimated the benefit of
corallines, and algal turfs. Data are means 6 SE; nd ¼ no data protection from herbivores provided by the cyanobac-
collected. All sample sizes are n ¼ 10 quadrats, except for teria. As biomass accumulation was similar between
Cavada Island 2002 (all zones), n ¼ 19 quadrats.
caged/cleaned and open/epiphytized plots, these results
suggest that epiphytism was as efficacious in reducing
previously were highest. Herbivory rates over 24 h were herbivory as were our cages.
calculated as loss from initial biomass and subjected to a Microcosm bioassays using algae from the Cavada
two-way mixed-model ANOVA with the fixed factor of Island bloom demonstrated primary limitation by P (Fig.
ambient vs. enriched and the random factor of location. 3A, Table 2B). Addition of N and P together appeared to
RESULTS increase growth more than addition of P alone; however,
the interaction between N and P was not significant,
The Acanthophora spicifera–cyanobacterial associa- suggesting that, if present, secondary N limitation was
tion observed first in the southern portion of the reef
weak. Exposure to N and P in situ increased algal growth
base of Uva Island reef in 1999 dominated benthic cover
in this area by 2000 (Table 1) and persisted for at least
six years. On Cavada Island reef, the A. spicifera
association dominated all reef zones in both years
measured, covering ;60% of the benthos. In general,
A. specifera co-occurred with an ‘‘understory‘‘ of algal
turfs and crustose coralline algae; only rarely were other
upright macroalgae found within blooms. Throughout
the six years we observed A. spicifera domination of a
whole reef (Cavada) or a reef zone (Uva), thalli were
always coated with filaments of cyanobacteria.
The A. spicifera association sustained net biomass
accumulation in both uncaged and caged treatments on
the Cavada reef crest (Fig. 1A). However, algal thalli
increased in height faster when protected from herbi-
vores (paired t test; df ¼ 2, t ¼ 5.730, P ¼ 0.0291),
demonstrating the importance of herbivory in limiting
algal standing-stock during blooms. Thalli grew nearly
twice as high when large herbivores were excluded, but
still increased in height by over 40% in six days with
herbivores. On Uva Island reef, herbivory rates varied FIG. 1. In situ experiments quantified the effects of
across the reef (Fig. 1B, one-way ANOVA, F6,42 ¼ 5.367, herbivory on two reefs where Acanthophora spicifera dominated
P ¼ 0.0003), with highest rates on the slope and lowest on community structure: (A) Cavada Island in March 2001, caged
the reef base where the algal bloom occurred. Rates (herbivores excluded) vs. uncaged growth as increase in thallus
height over six days; (B) spatial patterns in herbivory across
varied from 18% to 73% removal of the initial 5-g sample Uva Island reef in May 2000. Bars with shared letters are not
across all reef zones, demonstrating spatial variability in significantly different by Fisher’s protected least significant
the importance of herbivory on Uva Island reef. difference (PLSD) tests. Error bars represent 6SE.
1166 PEGGY FONG ET AL. Ecology, Vol. 87, No. 5
secondary compounds that deter herbivory (reviewed in
Hay 1996). However, A. spicifera has long been
considered highly palatable (Lewis 1986), and our results
support that undefended algal thalli are rapidly con-
sumed. In contrast, filamentous cyanobacteria in the
genus Lyngbya have a wide diversity and abundance of
chemical defenses that effectively deter consumption by
herbivorous fishes and invertebrates (Nagle and Paul
1998, 1999). Other positive interactions where physically
or chemically defended organisms have provided pro-
tection to palatable organisms by proximity or crypsis
have been documented (reveiwed in Hay 1996). However,
to our knowledge, ours is the first documentation that
this form of facilitation was responsible for maintaining a
change in community structure up to the scale of an
entire reef. Since cyanobacterial blooms on tropical reefs
appear to be increasingly common (Nagle and Paul
1999), further research into community effects is needed.
Our results do not support the contention that
reduction in numbers of herbivores was a necessary
precursor to coral reef decline and shifts to algal
dominance. Rather, at least on some reefs of the Eastern
Pacific, macroalgae dominated benthic communities for
several years after the pulse disturbances of ENSO-
related coral mortality, even in the presence of abundant
herbivores. The conclusion that reductions in herbivory
were not necessary rests on two assumptions. The first is
that the 25-yr monitoring data showing no decrease in
abundance of herbivorous fishes (Glynn 1990, 2004)
represents ‘‘natural’’ herbivore abundance. Certainly it is
possible that historical (.25 yr ago) numbers of
TABLE 2. Results of statistical tests for two-way ANOVAs.
FIG. 2. In situ experiments demonstrated protection from Source df F P
herbivory by cyanobacterial epiphytes: (A) change in percent-
A) Effects of herbivory and
age of cover; (B) change in thallus height of cleaned plots and epiphytes on algal growth
plots with epiphytes on Uva Island reef in March 2003; (C)
change in thallus height for an experiment varying epiphyte and Caged vs. uncaged 1 8.648 0.0058
Epiphytized vs. cleaned 1 2.782 0.1043
herbivore presence in March 2004 on Uva Island reef. Error
Interaction 1 19.139 0.0001
bars represent 6SE. Residual 32
B ) Effect of nutrients on algal growth
across all reef zones tested (Fig. 3B, Table 2C). In some
N addition 1 1.626 0.2205
zones, growth more than doubled in the presence of slow- P addition 1 14.634 0.0015
release fertilizer, approaching 30% over four days Interaction 1 4.163 0.0582
(maximum, 7.5% per day). Herbivores selectively grazed Residual 16
algal tissue enriched in nutrients (Fig. 3C, Table 2D) C) Effects of nutrients and reef zone
on algal growth
across all locations on the reef slope. Herbivory was up
Reef zone 3 1.409 0.2581
to four times greater on enriched algae. Although Nutrient treatment 1 17.517 0.0002
herbivory rates appeared spatially variable, differences Interaction 3 0.246 0.8634
among locations were not significant. Residual 32
D) Effects of nutrients and location
DISCUSSION on reef slope on algal growth
Algal dominance of some tropical reef communities of Location 3 2.55 0.0727
Nutrients 1 18.450 0.0002
the Eastern Pacific that was initiated by ENSO dis- Interaction 3 0.354 0.7868
turbance of coral was facilitated by the protection from Residual 32
herbivory afforded by epiphytic cyanobacteria. Numer- Notes: Panels (C) and (D) are mixed models with reef zone
ous studies have established that many tropical macro- and location, respectively, being random factors. Levels of
algae are expert in chemical warfare, accumulating factors are described in Methods.
May 2006 CYANOBACTERIA MAINTAINS SHIFT TO ALGAE 1167
functionally decrease herbivory pressure even with
constant numbers of herbivores (Williams et al. 2001);
this would be a phase shift in response to changing
conditions. However, unless herbivore populations are
regulated by factors other than food resources, one
would expect compensatory increases in population size
to eliminate the algal bloom over time, and no such
compensatory increases occurred. While it was clear that
herbivory was an important force on the study reefs as
unepiphytized macroalgal biomass was diminished or
canopy heights reduced by herbivory across all reef
zones, herbivory did not eliminate algal dominance in
these reef systems over the six years of study.
We believe it was unlikely that a change in nutrient
supply alone such as occurs during La Nina conditions
˜
could sustain macroalgal dominance in this system.
Clearly nutrient supply was important to primary
productivity. Our bioassay experiments showed that
macroalgae from the bloom that dominated all zones of
Cavada Island reef was primarily limited by P and may
be secondarily limited by N. Furthermore, elevated
nutrient supply to algae transplanted across all reefs
zones on Uva Island reef increased growth. These results
suggested that La Nina conditions may have facilitated
˜
the initiation of the macroalgal blooms. However,
nutrient-stimulated growth rates (maximum of 7% per
day) never exceeded consumption rates of unprotected
thalli of A. spicifera (minimum of 18% per day). In
addition, thalli with nutrient enriched tissue were
preferentially consumed, presumably negating the pos-
itive effects of nutrients on growth of opportunistic, fast-
growing species that can store nutrients.
Our results suggested that this tropical reef system may
be an ideal area to test experimentally whether this large-
FIG. 3. Experiments investigating the role of nutrients. (A) scale community shift was a phase shift as is commonly
A microcosm ‘‘bioassay’’ experiment tested for nitrogen vs. thought to occur on other reef systems, or represents
phosphorus limitation in May 2000 with algae from Cavada alternative stable states. To be a phase shift, there must be
Island; treatments are þN, nitrogen enrichment; þP, phospho- an underlying change in an environmental parameter,
rus enrichment; þNþP, nitrogen and phosphorus enrichment;
and C, ambient seawater (control). (B) An in situ experiment
such as herbivory pressure, that triggered a threshold
tested the effects of enriched vs. ambient nutrient supplies on response in the community. It is possible that widespread
the growth of algae in different zones on Uva reef in March coral mortality changed herbivory pressure without
2002. (C) An in situ experiment quantified herbivory on changing herbivore numbers by diluting the force of
enriched and ambient algal tissue outplanted to Uva reef in
herbivory over a larger grazing area; thus, to test for a
May 2000. Error bars represent 6SE.
phase shift it will be necessary to study the relationship
between herbivore numbers, areal grazing rates, and
herbivores were higher. However, in most places where benthic community structure in this system. To be
herbivore populations have declined historically, the rate alternate stable states, the shift must meet several criteria
of the decline has accelerated during the recent, recorded (reviewed Petraitis and Dudgeon 2004b), including that
past (Pandolfi et al. 2005). This pattern did not occur in each state is self-sustaining over many generations. While
the area of the Eastern Tropical Pacific with benthic our observations certainly exceeded many generations of
community shifts (Glynn 1990, 2004), although decreases the macroalga, they did not cover a single coral
in abundance and size of herbivores have been recorded generation; therefore the jury will remain out on stability
in more populated areas of the region (Smith 2005). The for many years to come. However, several of the criteria
second assumption is that the number of herbivores is a for shifts between alternate stable states were met. First,
good estimate of herbivory rate in this system. While this the shift was in response to a large-scale pulse disturb-
may be true for many coral reefs, large disturbances such ance, the 1997–1998 ENSO that caused widespread coral
as ENSO open large portions of the substrate to algal mortality (Glynn et al. 2001). Second, this shift was most
colonization that may increase resource availability and probably initiated by stochastic rather than deterministic
1168 PEGGY FONG ET AL. Ecology, Vol. 87, No. 5
events as establishment of the algal community must have Glynn, P. W. 1990. Coral mortality and disturbances to coral reefs
been initiated through co-occurrence of a sufficient in the tropical eastern Pacific. Pages 55–126 in P. W. Glynn,
editor. Global ecological consequences of the 1982–83 El Nino–
˜
supply of larval or asexual recruits of both the macroalgal Southern Oscillation. Elsevier, Amsterdam, The Netherlands.
host and the cyanobacterial epiphyte. These were most Glynn, P. W. 2004. High complexity food webs in low-diversity
likely stochastic processes as they occurred across many eastern Pacific reef-coral communities. Ecosystems 7:358–367.
reef zones at least on Cavada Island reef, suggesting ´
Glynn, P. W., J. L. Mate, A. C. Baker, and M. O. Calderon. ´
´
2001. Coral bleaching and mortality in Panama and Ecuador
occurrence across many environmental gradients. Third,
during the 1997–98 El Nino–Southern Oscillation event:
˜
once established, algal dominance was maintained by spatial/temporal patterns and comparisons with the 1982–83
strong positive associations within the algal community event. Bulletin of Marine Science 69:79–109.
that reduced herbivory. Bertness et al. (2004b) suggested Hatcher, B. G., and A.W. D. Larkum. 1983. An experimental
that stochastic events may be more important in areas analysis of factors controlling the standing crop of the
epilithic algal community on a coral reef. Journal of
where strong determinism associated with consumer Experimental Marine Biology and Ecology 69:61–84.
control is relaxed. In our systems, we found that the Hay, M. E. 1996. Marine chemical ecology: what’s known and
positive interaction between macroalga and cyanobacte- what’s next? Journal of Experimental Marine Biology and
rial epiphytes uncoupled consumer control of community Ecology 200:103–134.
Hughes, T. P., et al. 2003. Climate change, human impacts, and
structure. Thus, large-scale and long-term experiments the resilience of coral reefs. Science 301:929–933.
investigating alternate stable states should be a focus of Jackson, J. B. C., et al. 2001. Historical overfishing and the
future study in this tropical reef community. recent collapse of coastal ecosystems. Science 293:629–638.
Knowlton, N. 1992. Threshold and multiple stable states in coral
ACKNOWLEDGMENTS reef community dynamics. American Zoologist 32:674–682.
Funding from this study was provided by the Biological Lapointe, B. 1997. Nutrient thresholds for eutrophication and
Oceanography Program, U.S. NSF grant OCE-0002317 to P. macroalgal overgrowth of coral reefs in Jamaica and south-
W. Glynn and P. Fong and a Smithsonian Tropical Research east Florida. Limnology and Oceanography 42:1119–1131.
Institute Short-Term Predoctoral fellowship to T. B. Smith. We Lewis, S. M. 1986. The role of herbivorous fishes in the
thank Peter Glynn for his advice and encouragement on all organization of a Caribbean reef community. Ecological
aspects of this research; he is a mentor and cherished friend to Monographs 56:183–200.
us all. R. Cohen, C. Hueerkamp, A. Romanski, D. Manzillo, McCook, L. J. 1999. Macroalgae, nutrients and phase shifts on
and the captain and crew of the R.V. Urraca all provided coral reefs: scientific issues and management consequences
enthusiastic help in the field. for the Great Barrier Reef. Coral Reefs 18:357–367.
McGlathery, K. J. 1995. Nutrient and grazing influences on a
LITERATURE CITED subtropical seagrass community. Marine Ecology Progress
Series 122:239–252.
Aronson, R. B., J. F. Bruno, W. F. Precht, P. W. Glynn, C. D. Nagle, D. G., and V. J. Paul. 1998. Chemical defense of a
Harvell, L. Kaufman, C. S. Rogers, E. A. Shinn, and J. F. marine cyanobacterial bloom. Journal of Phycology 34:43.
Valentine. 2003. Causes of coral reef degradation. Science Nagle, D. G., and V. J. Paul. 1999. Production of secondary
302:1502–1504. metabolites by filamentous tropical marine cyanobacteria:
Aronson, R. B., I. G. Macintyre, C. M. Wapnick, and M. W. ecological functions of the compounds. Journal of Phycology
O’Neill. 2004. Phase shifts, alternative states, and the 35:1412–1421.
unprecedented convergence of two reef systems. Ecology Pandolfi, J. M., et al. 2003. Global trajectories of the long-term
85:1876–1891. decline of coral reef ecosystems. Science 301:955–958.
Beisner, B. E., D. T. Haydon, and K. Cuddington. 2003. Pandolfi, J. M., J. B. C. Jackson, N. Baron, R. H. Bradbury, H.
Alternative stable states in ecology. Frontiers in Ecology and M. Guzman, T. P. Hughes, C. V. Kappel, F. Micheli, J. C.
Environmental Science 1:376–382. Odgen, H. P. Possingham, and E. Sala. 2005. Are U. S. coral
Bertness, M. D., G. C. Trussell, P. J. Ewanchuk, and B. R. reefs on the slippery slope to slime? Science 307:1725–1726.
Silliman. 2004a. Do alternate stable community states exist in Peterson, C. H. 1984. Does a rigorous criterion for environ-
the Gulf of Maine rocky intertidal zone? Reply. Ecology 85: mental identity preclude the existence of multiple stable
1165–1167. points? American Naturalist 124:127–133.
Bertness, M. D., G. C. Trussell, P. J. Ewanchuk, B. R. Silliman, Petraitis, P. S., and S. R. Dudgeon. 2004a. Do alternate stable
and C. M. Crain. 2004b. Consumer-controlled community community states exist in the Gulf of Maine rocky intertidal
states on Gulf of Maine rocky shores. Ecology 85:1321–1331. zone? Comment. Ecology 85:1160–1165.
Boyer, K. E., P. Fong, A. R. Armitage, and R. A. Cohen. 2004. Petraitis, P. S., and S. R. Dudgeon. 2004b. Detection of
Elevated nutrient content of macroalgae increases rates of alternative stable states in marine communities. Journal of
herbivory in coral, seagrass, and mangrove habitats. Coral Experimental Marine Biology and Ecology 300:343–371.
Reefs 23:530–538. Schroder, A., L. Persson, and A. M. De Roos. 2005. Direct
¨
Connell, J. H., and W. P. Sousa. 1983. On the evidence needed experimental evidence for alternative stable states: a review.
to judge ecological stability or persistence. American Oikos 110:3–19.
Naturalist 121:789–824. Smith, T. B. 2005. The dynamics of coral reef algae in an
Estes, J. A., and J. F. Palmisano. 1974. Sea otters: their role in upwelling system. Dissertation. University of Miami, Miami,
structuring nearshore communities. Science 185:1058–1060. Florida, USA.
Fong, P., K. E. Boyer, K. Kamer, and K. A. Boyle. 2003. Sutherland, J. P. 1974. Multiple stable points in natural
Influence of initial tissue nutrient status of tropical marine communities. American Naturalist 108:859–873.
algae on response to nitrogen and phosphorus additions. Szmant, A. M. 2002. Nutrient enrichment on coral reefs: is it a
Marine Ecology Progress Series 262:111–123. major cause of coral reef decline? Estuaries 25:743–766.
Fox, H. E., P. J. Mous, J. S. Pet, A. H. Muljadi, and R. L. Williams, I. D., N. V. C. Polunin, and V. J. Hendrick. 2001.
Caldwell. 2005. Experimental assessment of coral reef Limits to grazing by herbivorous fishes and the impact of low
rehabilitation following blast fishing. Conservation Biology coral cover on macroalgal abundance on a coral reef in
19:98–107. Belize. Marine Ecology Progress Series 222:187–196.
Ó 2006 by the Ecological Society of America
EPIPHYTIC CYANOBACTERIA MAINTAIN SHIFTS TO MACROALGAL
DOMINANCE ON CORAL REEFS FOLLOWING ENSO DISTURBANCE
PEGGY FONG,1,3 TYLER B. SMITH,2,4 AND MATTHEW J. WARTIAN1
1
Department of Ecology and Evolutionary Biology, University of California Los Angeles, 621 Young Drive South, Los Angeles,
California 90095-1606 USA
2
Division of Marine Biology and Fisheries, Rosenstiel School of Marine and Atmospheric Science, University of Miami,
4600 Rickenbacker Causeway, Miami, Florida 33149 USA
Abstract. Macroalgal dominance of some tropical reef communities in the Eastern Pacific
after coral mortality during the 1997–1998 El Nino Southern Oscillation (ENSO) was
˜
facilitated by protection from herbivory by epiphytic cyanobacteria. Our results do not
support that reduction in number of herbivores was a necessary precursor to coral reef decline
and shifts to algal reefs in this system. Rather, macroalgae dominated the community for
several years after this pulse disturbance with no concurrent change in herbivore populations.
While results of microcosm experiments identified the importance of nutrients, especially
phosphorus, in stimulating macroalgal growth, nutrient supply alone could not sustain
macroalgal dominance as nutrient-stimulated growth rates in our in situ experiments never
exceeded consumption rates of unprotected thalli. In addition, thalli with nutrient-enriched
tissue were preferentially consumed, possibly negating the positive effects of nutrients on
growth. These tropical reefs may be ideal systems to conduct experimental tests distinguishing
phase shifts from alternative stable states. Shifts were initiated by a large-scale disturbance
with no evidence of a changing environment except, perhaps, dilution in herbivory pressure
due to increased algal cover. Community establishment was most likely stochastic, and the
community was likely maintained by strongly positive interaction between macroalgal hosts
and cyanobacterial epiphytes that uncoupled consumer control of community structure.
Key words: alternative stable states; coral reefs; facilitation; herbivory; macroalgal bloom; nutrients;
phase shifts.
INTRODUCTION shifts to algal domination, the underlying mechanisms
Investigating mechanisms causing and maintaining causing and maintaining these shifts are hotly debated
shifts among different states of ecological communities (e.g., Jackson et al. 2001, Aronson et al. 2003, Hughes et
has been a central focus of research for many years (e.g., al. 2003, Pandolfi et al. 2003). Are they alternative stable
Estes and Palmisano 1974, Sutherland 1974). Current states where two communities exist under the same set of
debate centers on whether these shifts are between environmental conditions, or phase shifts from one state
alternative stable states or are phase shifts due to to another caused by a threshold response to change in
underlying differences in environmental conditions, and an environmental parameter? Most current evidence
under what conditions shifts are stochastic or determin- points to the latter for coral reefs (reviewed in Petraitis
istic processes (e.g., Beisner et al. 2003, Bertness et al. and Dudgeon 2004b), with some claiming that reduction
2004a, Petraitis and Dudgeon 2004a). Despite much in consumer populations via harvesting of herbivorous
debate, community shifts initiated by disturbance, fishes is a necessary first step in this phase shift (e.g.,
frequently mediated by strong biotic interactions, and Jackson et al. 2001). They reason that the force of
often sustained by accompanying changes in abiotic herbivory is so strong in coral reef ecosystems with
factors such as supply of nutrients or changes in physical intact herbivore populations that herbivores virtually act
habitat structure, have been documented in freshwater, as lawnmowers, limiting algal biomass accumulation
terrestrial, and marine communities (reviewed in and negating the impact of any other factor that may
Schroder et al. 2005).
¨ enhance algal productivity and growth, thus maintain-
While most coral reef ecologists agree coral commun- ing the competitive dominance of slower growing corals.
ities are undergoing worldwide declines with concurrent Once herbivores are removed, faster-growing algae
become competitively dominant and a phase-shift
Manuscript received 16 May 2005; revised 4 October 2005; occurs. Others argue that reduced herbivory is not
accepted 22 November 2005. Corresponding Editor: P. D. always needed to initiate a phase shift attributable to
Steinberg. altered outcomes of competition. One alternative
3
E-mail: pfong@biology.ucla.edu
4
Present address: Center for Marine and Environmental
mechanism is for increased nutrient supplies to stimulate
Studies, University of the Virgin Islands, 2 John Brewers rapid growth and allow algae to escape control by
Bay, St. Thomas, U.S. Virgin Islands 00802 USA. herbivores, biomass to accumulate, a change in com-
1162
May 2006 CYANOBACTERIA MAINTAINS SHIFT TO ALGAE 1163
petitive dominance from coral to algae, and ultimately a of nutrient-rich bottom water, but persisted for years
phase shift from coral to algae (Lapointe 1997). past this environmental fluctuation. Herbivore popula-
A plethora of evidence supports the role of herbivores in tions have been monitored for over 25 years on Uva
limiting algal biomass on coral reefs; in contrast, the role Island reef, with no overall trend of reduction in
of nutrients has been less well established (reviewed in numbers (Glynn 1990, 2004). In this study, we quantified
Szmant 2002). Overall, results of laboratory and micro- benthic dominance of the A. spicifera–cyanobacterial
cosm experiments demonstrate that reef algae are nutrient association on our two study reefs, and investigated
limited (e.g., Fong et al. 2003), while field experiments several factors that may sustain their dominance after
often have mixed results (e.g., Hatcher and Larkum 1983). ENSO disturbance, including rates of herbivory, pro-
Interpretation of these experiments is limited, in part, by tection from herbivory by cyanobacterial epiphytes, and
the difficulty of relating results of laboratory or micro- the potential for nutrient stimulation.
cosm studies of the effects of nutrient addition to natural
growth in high-energy, high-flow environments typical of METHODS
reefs, and the related methodological challenge of Benthic cover of the A. spicifera–cyanobacterial
effectively enriching in situ experiments in these same association, total coral, and total algae on our study
environments (reviewed in McCook 1999). In addition, a reefs in the Panamanian Gulf of Chiriquı´ was measured
recent study suggests that, on the local scale, herbivores on Uva Island reef in 1999 and 2000 and on the Cavada
may selectively consume more opportunistic, fast-growing Island reef in 2001 and 2002. The cyanobacterial
macroalgae that accumulate nutrient reserves in their epiphyte was putatively in the genus Lyngbya. Total
tissues when enriched (Boyer et al. 2004), further algae was the sum of macroalgae, crustose corallines,
confounding in situ enrichment experiments. There is and algal turfs. On Cavada reef, 1-m2 quadrats were
similar evidence for selective consumption of enriched placed at regular intervals (5 m) along cross-reef-zone
seagrass (McGlathery 1995). Thus, studies assessing the transects. On Uva reef, 1-m2 quadrats were placed
roles of both herbivores and nutrients in sustaining algal randomly along 30-m transects within reef zones in the
communities on coral reefs are important. southern portion of the reef. Percentage of cover was
An alternative view from changes in top-down or calculated by determining the cover category underlying
bottom-up processes leading to phase shifts is that 81 points within the 1-m2 quadrat.
complex multiple stresses to coral communities may act We assessed the importance of herbivory on biomass
alone or in concert to cause shifts to alternative accumulation of A. spicifera by conducting two field
community states on tropical reefs (Knowlton 1992, experiments on our two study reefs. In February 2001 on
Aronson et al. 2003, 2004). For these shifts to be Cavada reef, three areas (3–5 m apart) on the reef crest
considered alternative stable states, theory requires that were haphazardly selected. Within areas, paired plots
causative factors initiating changes must be pulse were assigned to either open or caged treatments. Within
disturbances (Connell and Sousa 1983, Peterson 1984, each plot, two thalli were marked with cable ties and
Petraitis and Dudgeon 2004b; but see Beisner et al. height measured to the nearest millimeter. We did not
2003); coral reef examples include El Nino Southern
˜ remove epiphytes during this experiment. Cages were 20
Oscillation (ENSO)-related mortality (Glynn et al. 2001) 3 20 cm (height 3 diameter) cylinders of high-density
and destructive fishing practices (Fox et al. 2005). Press polyethylene with a mesh size of 1.4 cm (Memphis Net
forces that may act in concert with pulse disturbances to and Twine, Memphis, Tennessee, USA) attached to
sustain algal dominance may include reduction in three rebar stakes driven into the substrate. The experi-
herbivory pressure and increases in nutrient supplies, ment lasted six days, when marked thalli were remea-
resulting in the proliferation of opportunistic, fast- sured. Percent increase in height was averaged for the
growing algae on newly opened space and/or dominance two thalli per plot, and the averages used in a paired t
by algae that is physically or chemically defended from test to detect differences in means between caged and
herbivory (reviewed in Knowlton 1992, Szmant 2002). uncaged plots. A mark was lost from only one thallus,
Clearly, there is a need for studies that investigate the and in this case one change in height was used in the
complex and potentially interacting ecological forces analysis. Another experiment was conducted on Uva
that maintain shifts to algal dominated tropical reefs Island reef in May 2000. A. spicifera thalli were collected
after large-scale pulse disturbances. from the bloom area, cleaned of macroepiphytes, placed
Blooms of Acanthophora spicifera (Rhodophyta) in mesh bags, spun for 1 min in a salad spinner to
covered by cyanobacterial epiphytes have been observed remove excess water, and weighed. Samples of 5 g wet
on several reefs in the Eastern Tropical Pacific since weight were cable tied together and attached to weighted
widespread coral mortality opened substrate for colo- ropes. Ropes with algae were anchored in seven different
nization during the 1997–1998 ENSO. These reefs were reef areas from the seaward reef base across the reef to
decimated by the 1982–1983 ENSO and suffered further the landward backreef; n ¼ 7 replicates for each area.
coral mortality during the 1997–1998 event (Glynn et al. Herbivory rate was calculated as loss in grams over 24
2001). Bloom initiation coincided with La Nina con- ˜ hours. Data were analyzed with one-way ANOVA (with
ditions, with thermocline shoaling and mixing upwards location as the factor) after being tested to be sure they
1164 PEGGY FONG ET AL. Ecology, Vol. 87, No. 5
met ANOVA assumptions. Fisher’s protected least from the Cavada Island reef in May 2000, we conducted a
significant difference (PLSD) tests were conducted after 2 3 2 factorial experiment with five replicates. Exper-
a significant ANOVA. imental treatments were ambient seawater with addition
Two experiments were conducted to assess the effects of 20 lmol/L NO3, 2 lmol/L PO4, 20 lmol/L NO3 þ 2
of epiphytic cyanobacteria on growth of and herbivory lmol/L PO4, and an ambient seawater control (abbre-
rates on A. spicifera in the bloom area of Uva Island viated as þN, þP, þNþP, and C, respectively). Exper-
reef. In March of 2003, we chose 20 plots (25 3 25 cm) imental units were plastic jars containing 500 mL of
within the visually homogeneous rubble area dominated treatment seawater and 5 g wet weight of algae. Algal
by algae covered in epiphytes. Epiphytes were gently thalli were collected, cleaned of macroepiphytes, weighed
removed (brushing by hand) from thalli in 10 randomly as before, and trimmed to the correct size to include
chosen plots, and left intact on algae in the remaining 10 several apical growing tips. Jars were placed in a random
plots. Percentage of cover of algae was estimated using array in an outdoor flow-through water table to maintain
the point–intercept method (determining the percentage ambient temperature. Because experimental units were
of points in a grid that overlaid the A. spificera– static and not flow-through, this experiment should be
cyanobacteria association) initially and after five days considered a ‘‘bioassay,’’ or a simple attempt to determine
and the change in percentage of cover was calculated. if N or P is limiting, rather than an attempt to predict
Heights of 16 randomly selected thalli in each plot were growth responses of algae in the field. Thus, growth rates
measured and averages calculated. Initial averages were can be compared within the experiment, but should not be
subtracted from final averages to calculate the change in extrapolated to natural systems. Jars were screened to
height per plot after five days. T tests were used to reduce ambient light by approximately 30%. The experi-
compare the mean change in percentage of cover and ment lasted three days, then algal biomass was again
height between cleaned and epiphytized thalli. measured and percentage of change calculated. A two-
To compare the potential cost of epiphytism (e.g., way ANOVA tested for differences in means after data
reduced growth due to shading, nutrient competition) to were tested to be sure they met ANOVA assumptions.
benefits of epiphytism as protection from herbivory for To determine if added nutrients would enhance
A. spicifera, in March of 2004 we conducted a 2 3 2 growth of in situ algae on Uva Island reef in March
factorial experiment varying protection from herbivory 2002, we exposed algal thalli to ambient and enriched
(with vs. without cages) and removal of epiphytes conditions across four reef zones. Algae were collected
(cleaned vs. epiphytized). Treatments were randomly and weighed (as already described), placed in mesh bags
assigned to plots within the rubble area where the bloom (nylon window screening) to protect them from herbi-
occurred. Because the rubble benthos was topographi- vores, and 10 replicates were anchored in each of four
cally complex, open-bottomed cages were not considered reefs zones (slope, base, crest, and flat). Five grams of
adequate to protect thalli from herbivores. Thus, cages slow-release fertilizer (Osmocote [Scots Miracle-Gro
(constructed as above) were 15 3 10 cm (height 3 Company, Marysville, Ohio, USA], 15% N þ 5% P)
diameter) complete cylinders with both tops and were placed in nylon stockings and attached to five of
bottoms, while open plots consisted of bottoms only; the replicates in each reef zone. Samples were anchored
all experimental units were held to the benthos with rebar at least 2 m apart to reduce cross-contamination by
stakes. Three to four rubble pieces with attached algae nutrients. Preliminary data (P. Fong, unpublished data)
were placed in each cage or on each open plot base, and demonstrated that our method of nutrient enrichment
heights of four or five algal thalli were measured to the was undetectable (measured as storage in algal tissue) at
nearest millimeter. After five days, the change in height .1 m distance. Growth after four days was calculated as
was calculated as described in the last paragraph. Data above and subjected to a two-way mixed-model
were analyzed with two-way ANOVA after being tested ANOVA (with the fixed factor of ambient vs. enriched
to be sure they met ANOVA assumptions. Comparison and the random factor reef zone).
of algal growth of cleaned vs. epiphytized thalli within To investigate whether herbivores consumed more
cages provided an estimate of the costs of being enriched algae, in May 2000 we performed the above
epiphtytzed without the confounding effect of herbivory. experiment with three variations. First, after collecting
Comparison of algal net growth in cleaned vs. epiphy- the algae we cultured half for three days with slow-
tized open plots subject to herbivores estimated the release fertilizer in batch culture. The other half were
benefits of protection provided by the epiphytes. Cage kept in ambient seawater. Initial tissue N was greater in
controls consisting of cages with no tops for the enriched (2.60% 6 0.04% [mean 6 SE]) than in ambient
epiphytized treatment only demonstrated that cage (2.40% 6 0.04%) algae (t test, df ¼ 8, P ¼ 0.006), but
effects were not significant (t test comparing epiphytized differences in P (0.14% 6 0.004% enriched vs. 0.13% 6
open to cage control, df ¼ 13, t value ¼ 0.738, P ¼ 0.4737). 0.003% ambient) were not significant. Second, algae
We explored the role of nutrients in sustaining algal were exposed to herbivores as they were anchored
growth and biomass accumulation using three experi- without being enclosed in mesh bags. Finally, instead of
mental approaches. To determine if nitrogen (N) or testing effects across reef zones, we chose four locations
phosphorus (P) limited growth of A. spicifera collected within the reef slope, where herbivory rates measured
May 2006 CYANOBACTERIA MAINTAINS SHIFT TO ALGAE 1165
TABLE 1. Benthic cover of the Acanthophora spicifera–cyano- A. spicifera maintained net biomass accumulation in
bacterial association, total coral, and total algae on study the presence of herbivores only when it was covered with
reefs.
cyanobacterial epiphytes on Uva Island reef (Fig. 2A and
Cover (%) B). When comparing algae in cleaned vs. epiphytized
plots in 2003, epiphytized algae increased both cover by
A. spicifera– Total Total
Reef, year, and zone cyaonobacteria coral algae ;5% (t test; df ¼ 18, t ¼À2.859, P ¼ 0.0104) and thallus
height by ;2.5 cm (t test; df ¼ 18, t ¼À7.603, P , 0.0001)
Cavada Island 2001 over five days. When epiphytes were removed, cover was
All zones 59.5 6 4.0 19.8 6 3.5 74.6 6 3.7 reduced by almost 10% and thalli did not grow taller.
Cavada Island 2002 The effect of epiphytes on biomass accumulation of A.
All zones 66.1 6 5.8 3.9 6 1.7 nd spicifera thalli, measured as change in height, differed
Uva Island 1999 between caging treatments, resulting in interaction
Base 22.2 6 9.9 14.8 6 5.4 65.2 6 7.4 (Table 2A, Fig. 2C). Within cages, epiphytes reduced
Slope 0.8 6 0.5 26.5 6 4.6 57.3 6 6.4
Flat 1.3 6 0.7 14.3 6 4.6 77.54 6 8.1 thallus growth by ;50%. This estimated the cost in
growth to the algae of bearing epiphytes. In contrast,
Uva Island 2000
Base 77.2 6 6.8 9.3 6 5.1 89.8 6 4.9
only thalli that were epiphytized maintained positive
Slope 0.3 6 0.3 4.0 6 3.6 74.0 6 5.5 biomass accumulation in open plots exposed to herbi-
Flat 0.0 6 0.0 9.9 6 3.2 84.6 6 3.8 vores. This comparison between cleaned and epiphytized
Notes: Total algae is the sum of macroalgae, crustose thalli exposed to herbivores estimated the benefit of
corallines, and algal turfs. Data are means 6 SE; nd ¼ no data protection from herbivores provided by the cyanobac-
collected. All sample sizes are n ¼ 10 quadrats, except for teria. As biomass accumulation was similar between
Cavada Island 2002 (all zones), n ¼ 19 quadrats.
caged/cleaned and open/epiphytized plots, these results
suggest that epiphytism was as efficacious in reducing
previously were highest. Herbivory rates over 24 h were herbivory as were our cages.
calculated as loss from initial biomass and subjected to a Microcosm bioassays using algae from the Cavada
two-way mixed-model ANOVA with the fixed factor of Island bloom demonstrated primary limitation by P (Fig.
ambient vs. enriched and the random factor of location. 3A, Table 2B). Addition of N and P together appeared to
RESULTS increase growth more than addition of P alone; however,
the interaction between N and P was not significant,
The Acanthophora spicifera–cyanobacterial associa- suggesting that, if present, secondary N limitation was
tion observed first in the southern portion of the reef
weak. Exposure to N and P in situ increased algal growth
base of Uva Island reef in 1999 dominated benthic cover
in this area by 2000 (Table 1) and persisted for at least
six years. On Cavada Island reef, the A. spicifera
association dominated all reef zones in both years
measured, covering ;60% of the benthos. In general,
A. specifera co-occurred with an ‘‘understory‘‘ of algal
turfs and crustose coralline algae; only rarely were other
upright macroalgae found within blooms. Throughout
the six years we observed A. spicifera domination of a
whole reef (Cavada) or a reef zone (Uva), thalli were
always coated with filaments of cyanobacteria.
The A. spicifera association sustained net biomass
accumulation in both uncaged and caged treatments on
the Cavada reef crest (Fig. 1A). However, algal thalli
increased in height faster when protected from herbi-
vores (paired t test; df ¼ 2, t ¼ 5.730, P ¼ 0.0291),
demonstrating the importance of herbivory in limiting
algal standing-stock during blooms. Thalli grew nearly
twice as high when large herbivores were excluded, but
still increased in height by over 40% in six days with
herbivores. On Uva Island reef, herbivory rates varied FIG. 1. In situ experiments quantified the effects of
across the reef (Fig. 1B, one-way ANOVA, F6,42 ¼ 5.367, herbivory on two reefs where Acanthophora spicifera dominated
P ¼ 0.0003), with highest rates on the slope and lowest on community structure: (A) Cavada Island in March 2001, caged
the reef base where the algal bloom occurred. Rates (herbivores excluded) vs. uncaged growth as increase in thallus
height over six days; (B) spatial patterns in herbivory across
varied from 18% to 73% removal of the initial 5-g sample Uva Island reef in May 2000. Bars with shared letters are not
across all reef zones, demonstrating spatial variability in significantly different by Fisher’s protected least significant
the importance of herbivory on Uva Island reef. difference (PLSD) tests. Error bars represent 6SE.
1166 PEGGY FONG ET AL. Ecology, Vol. 87, No. 5
secondary compounds that deter herbivory (reviewed in
Hay 1996). However, A. spicifera has long been
considered highly palatable (Lewis 1986), and our results
support that undefended algal thalli are rapidly con-
sumed. In contrast, filamentous cyanobacteria in the
genus Lyngbya have a wide diversity and abundance of
chemical defenses that effectively deter consumption by
herbivorous fishes and invertebrates (Nagle and Paul
1998, 1999). Other positive interactions where physically
or chemically defended organisms have provided pro-
tection to palatable organisms by proximity or crypsis
have been documented (reveiwed in Hay 1996). However,
to our knowledge, ours is the first documentation that
this form of facilitation was responsible for maintaining a
change in community structure up to the scale of an
entire reef. Since cyanobacterial blooms on tropical reefs
appear to be increasingly common (Nagle and Paul
1999), further research into community effects is needed.
Our results do not support the contention that
reduction in numbers of herbivores was a necessary
precursor to coral reef decline and shifts to algal
dominance. Rather, at least on some reefs of the Eastern
Pacific, macroalgae dominated benthic communities for
several years after the pulse disturbances of ENSO-
related coral mortality, even in the presence of abundant
herbivores. The conclusion that reductions in herbivory
were not necessary rests on two assumptions. The first is
that the 25-yr monitoring data showing no decrease in
abundance of herbivorous fishes (Glynn 1990, 2004)
represents ‘‘natural’’ herbivore abundance. Certainly it is
possible that historical (.25 yr ago) numbers of
TABLE 2. Results of statistical tests for two-way ANOVAs.
FIG. 2. In situ experiments demonstrated protection from Source df F P
herbivory by cyanobacterial epiphytes: (A) change in percent-
A) Effects of herbivory and
age of cover; (B) change in thallus height of cleaned plots and epiphytes on algal growth
plots with epiphytes on Uva Island reef in March 2003; (C)
change in thallus height for an experiment varying epiphyte and Caged vs. uncaged 1 8.648 0.0058
Epiphytized vs. cleaned 1 2.782 0.1043
herbivore presence in March 2004 on Uva Island reef. Error
Interaction 1 19.139 0.0001
bars represent 6SE. Residual 32
B ) Effect of nutrients on algal growth
across all reef zones tested (Fig. 3B, Table 2C). In some
N addition 1 1.626 0.2205
zones, growth more than doubled in the presence of slow- P addition 1 14.634 0.0015
release fertilizer, approaching 30% over four days Interaction 1 4.163 0.0582
(maximum, 7.5% per day). Herbivores selectively grazed Residual 16
algal tissue enriched in nutrients (Fig. 3C, Table 2D) C) Effects of nutrients and reef zone
on algal growth
across all locations on the reef slope. Herbivory was up
Reef zone 3 1.409 0.2581
to four times greater on enriched algae. Although Nutrient treatment 1 17.517 0.0002
herbivory rates appeared spatially variable, differences Interaction 3 0.246 0.8634
among locations were not significant. Residual 32
D) Effects of nutrients and location
DISCUSSION on reef slope on algal growth
Algal dominance of some tropical reef communities of Location 3 2.55 0.0727
Nutrients 1 18.450 0.0002
the Eastern Pacific that was initiated by ENSO dis- Interaction 3 0.354 0.7868
turbance of coral was facilitated by the protection from Residual 32
herbivory afforded by epiphytic cyanobacteria. Numer- Notes: Panels (C) and (D) are mixed models with reef zone
ous studies have established that many tropical macro- and location, respectively, being random factors. Levels of
algae are expert in chemical warfare, accumulating factors are described in Methods.
May 2006 CYANOBACTERIA MAINTAINS SHIFT TO ALGAE 1167
functionally decrease herbivory pressure even with
constant numbers of herbivores (Williams et al. 2001);
this would be a phase shift in response to changing
conditions. However, unless herbivore populations are
regulated by factors other than food resources, one
would expect compensatory increases in population size
to eliminate the algal bloom over time, and no such
compensatory increases occurred. While it was clear that
herbivory was an important force on the study reefs as
unepiphytized macroalgal biomass was diminished or
canopy heights reduced by herbivory across all reef
zones, herbivory did not eliminate algal dominance in
these reef systems over the six years of study.
We believe it was unlikely that a change in nutrient
supply alone such as occurs during La Nina conditions
˜
could sustain macroalgal dominance in this system.
Clearly nutrient supply was important to primary
productivity. Our bioassay experiments showed that
macroalgae from the bloom that dominated all zones of
Cavada Island reef was primarily limited by P and may
be secondarily limited by N. Furthermore, elevated
nutrient supply to algae transplanted across all reefs
zones on Uva Island reef increased growth. These results
suggested that La Nina conditions may have facilitated
˜
the initiation of the macroalgal blooms. However,
nutrient-stimulated growth rates (maximum of 7% per
day) never exceeded consumption rates of unprotected
thalli of A. spicifera (minimum of 18% per day). In
addition, thalli with nutrient enriched tissue were
preferentially consumed, presumably negating the pos-
itive effects of nutrients on growth of opportunistic, fast-
growing species that can store nutrients.
Our results suggested that this tropical reef system may
be an ideal area to test experimentally whether this large-
FIG. 3. Experiments investigating the role of nutrients. (A) scale community shift was a phase shift as is commonly
A microcosm ‘‘bioassay’’ experiment tested for nitrogen vs. thought to occur on other reef systems, or represents
phosphorus limitation in May 2000 with algae from Cavada alternative stable states. To be a phase shift, there must be
Island; treatments are þN, nitrogen enrichment; þP, phospho- an underlying change in an environmental parameter,
rus enrichment; þNþP, nitrogen and phosphorus enrichment;
and C, ambient seawater (control). (B) An in situ experiment
such as herbivory pressure, that triggered a threshold
tested the effects of enriched vs. ambient nutrient supplies on response in the community. It is possible that widespread
the growth of algae in different zones on Uva reef in March coral mortality changed herbivory pressure without
2002. (C) An in situ experiment quantified herbivory on changing herbivore numbers by diluting the force of
enriched and ambient algal tissue outplanted to Uva reef in
herbivory over a larger grazing area; thus, to test for a
May 2000. Error bars represent 6SE.
phase shift it will be necessary to study the relationship
between herbivore numbers, areal grazing rates, and
herbivores were higher. However, in most places where benthic community structure in this system. To be
herbivore populations have declined historically, the rate alternate stable states, the shift must meet several criteria
of the decline has accelerated during the recent, recorded (reviewed Petraitis and Dudgeon 2004b), including that
past (Pandolfi et al. 2005). This pattern did not occur in each state is self-sustaining over many generations. While
the area of the Eastern Tropical Pacific with benthic our observations certainly exceeded many generations of
community shifts (Glynn 1990, 2004), although decreases the macroalga, they did not cover a single coral
in abundance and size of herbivores have been recorded generation; therefore the jury will remain out on stability
in more populated areas of the region (Smith 2005). The for many years to come. However, several of the criteria
second assumption is that the number of herbivores is a for shifts between alternate stable states were met. First,
good estimate of herbivory rate in this system. While this the shift was in response to a large-scale pulse disturb-
may be true for many coral reefs, large disturbances such ance, the 1997–1998 ENSO that caused widespread coral
as ENSO open large portions of the substrate to algal mortality (Glynn et al. 2001). Second, this shift was most
colonization that may increase resource availability and probably initiated by stochastic rather than deterministic
1168 PEGGY FONG ET AL. Ecology, Vol. 87, No. 5
events as establishment of the algal community must have Glynn, P. W. 1990. Coral mortality and disturbances to coral reefs
been initiated through co-occurrence of a sufficient in the tropical eastern Pacific. Pages 55–126 in P. W. Glynn,
editor. Global ecological consequences of the 1982–83 El Nino–
˜
supply of larval or asexual recruits of both the macroalgal Southern Oscillation. Elsevier, Amsterdam, The Netherlands.
host and the cyanobacterial epiphyte. These were most Glynn, P. W. 2004. High complexity food webs in low-diversity
likely stochastic processes as they occurred across many eastern Pacific reef-coral communities. Ecosystems 7:358–367.
reef zones at least on Cavada Island reef, suggesting ´
Glynn, P. W., J. L. Mate, A. C. Baker, and M. O. Calderon. ´
´
2001. Coral bleaching and mortality in Panama and Ecuador
occurrence across many environmental gradients. Third,
during the 1997–98 El Nino–Southern Oscillation event:
˜
once established, algal dominance was maintained by spatial/temporal patterns and comparisons with the 1982–83
strong positive associations within the algal community event. Bulletin of Marine Science 69:79–109.
that reduced herbivory. Bertness et al. (2004b) suggested Hatcher, B. G., and A.W. D. Larkum. 1983. An experimental
that stochastic events may be more important in areas analysis of factors controlling the standing crop of the
epilithic algal community on a coral reef. Journal of
where strong determinism associated with consumer Experimental Marine Biology and Ecology 69:61–84.
control is relaxed. In our systems, we found that the Hay, M. E. 1996. Marine chemical ecology: what’s known and
positive interaction between macroalga and cyanobacte- what’s next? Journal of Experimental Marine Biology and
rial epiphytes uncoupled consumer control of community Ecology 200:103–134.
Hughes, T. P., et al. 2003. Climate change, human impacts, and
structure. Thus, large-scale and long-term experiments the resilience of coral reefs. Science 301:929–933.
investigating alternate stable states should be a focus of Jackson, J. B. C., et al. 2001. Historical overfishing and the
future study in this tropical reef community. recent collapse of coastal ecosystems. Science 293:629–638.
Knowlton, N. 1992. Threshold and multiple stable states in coral
ACKNOWLEDGMENTS reef community dynamics. American Zoologist 32:674–682.
Funding from this study was provided by the Biological Lapointe, B. 1997. Nutrient thresholds for eutrophication and
Oceanography Program, U.S. NSF grant OCE-0002317 to P. macroalgal overgrowth of coral reefs in Jamaica and south-
W. Glynn and P. Fong and a Smithsonian Tropical Research east Florida. Limnology and Oceanography 42:1119–1131.
Institute Short-Term Predoctoral fellowship to T. B. Smith. We Lewis, S. M. 1986. The role of herbivorous fishes in the
thank Peter Glynn for his advice and encouragement on all organization of a Caribbean reef community. Ecological
aspects of this research; he is a mentor and cherished friend to Monographs 56:183–200.
us all. R. Cohen, C. Hueerkamp, A. Romanski, D. Manzillo, McCook, L. J. 1999. Macroalgae, nutrients and phase shifts on
and the captain and crew of the R.V. Urraca all provided coral reefs: scientific issues and management consequences
enthusiastic help in the field. for the Great Barrier Reef. Coral Reefs 18:357–367.
McGlathery, K. J. 1995. Nutrient and grazing influences on a
LITERATURE CITED subtropical seagrass community. Marine Ecology Progress
Series 122:239–252.
Aronson, R. B., J. F. Bruno, W. F. Precht, P. W. Glynn, C. D. Nagle, D. G., and V. J. Paul. 1998. Chemical defense of a
Harvell, L. Kaufman, C. S. Rogers, E. A. Shinn, and J. F. marine cyanobacterial bloom. Journal of Phycology 34:43.
Valentine. 2003. Causes of coral reef degradation. Science Nagle, D. G., and V. J. Paul. 1999. Production of secondary
302:1502–1504. metabolites by filamentous tropical marine cyanobacteria:
Aronson, R. B., I. G. Macintyre, C. M. Wapnick, and M. W. ecological functions of the compounds. Journal of Phycology
O’Neill. 2004. Phase shifts, alternative states, and the 35:1412–1421.
unprecedented convergence of two reef systems. Ecology Pandolfi, J. M., et al. 2003. Global trajectories of the long-term
85:1876–1891. decline of coral reef ecosystems. Science 301:955–958.
Beisner, B. E., D. T. Haydon, and K. Cuddington. 2003. Pandolfi, J. M., J. B. C. Jackson, N. Baron, R. H. Bradbury, H.
Alternative stable states in ecology. Frontiers in Ecology and M. Guzman, T. P. Hughes, C. V. Kappel, F. Micheli, J. C.
Environmental Science 1:376–382. Odgen, H. P. Possingham, and E. Sala. 2005. Are U. S. coral
Bertness, M. D., G. C. Trussell, P. J. Ewanchuk, and B. R. reefs on the slippery slope to slime? Science 307:1725–1726.
Silliman. 2004a. Do alternate stable community states exist in Peterson, C. H. 1984. Does a rigorous criterion for environ-
the Gulf of Maine rocky intertidal zone? Reply. Ecology 85: mental identity preclude the existence of multiple stable
1165–1167. points? American Naturalist 124:127–133.
Bertness, M. D., G. C. Trussell, P. J. Ewanchuk, B. R. Silliman, Petraitis, P. S., and S. R. Dudgeon. 2004a. Do alternate stable
and C. M. Crain. 2004b. Consumer-controlled community community states exist in the Gulf of Maine rocky intertidal
states on Gulf of Maine rocky shores. Ecology 85:1321–1331. zone? Comment. Ecology 85:1160–1165.
Boyer, K. E., P. Fong, A. R. Armitage, and R. A. Cohen. 2004. Petraitis, P. S., and S. R. Dudgeon. 2004b. Detection of
Elevated nutrient content of macroalgae increases rates of alternative stable states in marine communities. Journal of
herbivory in coral, seagrass, and mangrove habitats. Coral Experimental Marine Biology and Ecology 300:343–371.
Reefs 23:530–538. Schroder, A., L. Persson, and A. M. De Roos. 2005. Direct
¨
Connell, J. H., and W. P. Sousa. 1983. On the evidence needed experimental evidence for alternative stable states: a review.
to judge ecological stability or persistence. American Oikos 110:3–19.
Naturalist 121:789–824. Smith, T. B. 2005. The dynamics of coral reef algae in an
Estes, J. A., and J. F. Palmisano. 1974. Sea otters: their role in upwelling system. Dissertation. University of Miami, Miami,
structuring nearshore communities. Science 185:1058–1060. Florida, USA.
Fong, P., K. E. Boyer, K. Kamer, and K. A. Boyle. 2003. Sutherland, J. P. 1974. Multiple stable points in natural
Influence of initial tissue nutrient status of tropical marine communities. American Naturalist 108:859–873.
algae on response to nitrogen and phosphorus additions. Szmant, A. M. 2002. Nutrient enrichment on coral reefs: is it a
Marine Ecology Progress Series 262:111–123. major cause of coral reef decline? Estuaries 25:743–766.
Fox, H. E., P. J. Mous, J. S. Pet, A. H. Muljadi, and R. L. Williams, I. D., N. V. C. Polunin, and V. J. Hendrick. 2001.
Caldwell. 2005. Experimental assessment of coral reef Limits to grazing by herbivorous fishes and the impact of low
rehabilitation following blast fishing. Conservation Biology coral cover on macroalgal abundance on a coral reef in
19:98–107. Belize. Marine Ecology Progress Series 222:187–196.