Supplemental Data                                                               S1


Phase Shifts, Herbivory,
and the Resilience of Coral Reefs
to Climate Change
Terence P. Hughes, Maria J. Rodrigues,
David R. Bellwood, Daniela Ceccarelli,
Ove Hoegh-Guldberg, Laurence McCook,
Natalie Moltschaniwskyj, Morgan S. Pratchett,
Robert S. Steneck, and Bette Willis

Supplemental Experimental Procedures                   macroalgae had thinner tissues (1.45 6 0.11 and 1.15 6 0.10 mm,
                                     respectively), a reduction of 22% and 36% (F = 4.55, p < 0.05). Sim-
Sublethal Impacts on Corals                        ilarly, the reproductive output of experimental fragments of Monti-
Tissue thickness, an indicator of coral biomass and energetic re-     pora digitata transplanted into cages and placed beneath algal can-
serves [S1], did not differ among colonies of Porites cylindrica inside  opies declined by half compared to adjacent, unshaded controls
or outside cages provided that they were not in competition with     inside and outside of cages. Egg size, the number of eggs per polyp,
algae (1.80 6 0.06 [S.E.] and 1.82 6 0.12 mm, respectively). However,   and number of reproductive polyps per cm2 of tissue all declined in
colonies in cages that were shaded or partially overgrown by       coral fragments placed beneath algal canopies (by an average of




                                                Figure S1. Biomass of Herbivorous Fishes
                                                Plotted against Body Length for Three Exper-
                                                imental Treatments
                                                (A) Cages that excluded medium- and large-
                                                sized fishes.
                                                (B) Partial cages that remained accessible to
                                                fishes of all size.
                                                (C) Open plots.
                                                The fish-exclusion cages worked very effec-
                                                tively at keeping out medium and large fishes,
                                                reducing the biomass of herbivorous fishes
                                                by close to an order of magnitude lower
                                                than the partial cages or open plots (Figure 3,
                                                F = 7.79, p < 0.001). Fishes greater than 15 cm
                                                standard length were virtually absent from
                                                the cages, yet they comprised greater than
                                                80% of the biomass in the two other experi-
                                                mental treatments. Although they made
                                                a negligible contribution to biomass, juvenile
                                                parrotfishes recruited readily into the cages
                                                and were four times more abundant there at
                                                the end of the experiment than elsewhere.
                                                The biggest fish (>30 cm), primarily roving
                                                scarids and acanthurids, accounted for ap-
                                                proximately half of the total herbivore bio-
                                                mass in partial cages and open plots. Their
                                                mobility, schooling behavior, and relatively
                                                low numbers (compared to copious numbers
                                                of smaller recruits) account for the large stan-
                                                dard errors around the average biomass. The
                                                similarity between the partial cages and open
                                                plots is striking, indicating that the former al-
                                                lows unhindered access to fishes. Error bars
                                                are SE.
S2




Figure S2. Canonical Discriminant Analysis Showing the Successional Changes in Macroalgal Composition
Each red circle (i.e., centroids with 95% confidence limits) indicates the taxonomic composition inside fish-exclusion cages, at 13 times over the
course of the experiment (units are in months). The blue circle encompasses all centroids for algal assemblages inside the partial cages and open
plots at each of the sampling periods, indicating that these two treatments are very similar and did not change over time. The first two axes in the
CDA explain 73% of the variation among treatments and times, confirming the rapid successional changes in macroalgal assemblages within the
cages. An initial bloom in the first 6–12 months was dominated by fast-growing Padina (ANOVA, F = 5.74, p < 0.001), with smaller amounts of
Hydroclathrus (F = 5.26, p < 0.0001) and a diverse range of other taxa. Subsequently, massive stands of Sargassum became dominant in cages
until the end of the experiment (F = 5.58, p < 0.001; Figure 1B). Beneath the Sargassum canopies, a speciose understory flora developed, includ-
ing Padina, Labophora, Hypnea, Turbinaria, and Peysonnelia crusts. On the Great Barrier Reef, fleshy algae such as Padina and Sargassum are
generally abundant only on near-shore reef flats, widely recognized as a spatial refuge from herbivorous fishes [S2, S3], and absent from heavily
grazed reef crests (the location of our experiment) where epilithic algal turfs (<10 mm tall) and coralline crusts predominate. Once the mesh was
removed from cages (after the 29th month, indicated by the dotted line), the Sargassum-dominated algal stands disappeared rapidly over a
period of 3–4 weeks as the flora converged once more toward the heavily grazed partial cages and open plots.



9.5%, 10.2%, and 25.8%, respectively) and thus yielded an overall
decline in reproductive output of 45% compared to controls (F =
5.04, p < 0.01).

Supplemental References

S1. Barnes, D.J., and Lough, J.M. (1992). Systematic variations in
  the depth of skeleton occupied by coral tissue in massive colo-
  nies of Porites from the Great Barrier Reef. J. Exp. Mar. Biol.
  Ecol. 159, 113–128.
S2. Hay, M.E. (1981). Herbivory, algal distribution, and the mainte-
  nance of between-habitat diversity on a tropical fringing reef.
  Am. Nat. 118, 520–540.
S3. McCook, L.J. (1999). Macroalgae, nutrients and phase shifts on
  coral reefs: Scientific issues and management consequences
  for the Great Barrier Reef. Coral Reefs 18, 357–367.
S4. Steneck, R.S. (1986). The ecology of coralline algal crusts:
  Convergent patterns and adaptive strategies. Annu. Rev. Ecol.
  Syst. 17, 273–303.
                      S3




Figure S3. Cover of Porolithon onkodes, in
the Three Experimental Treatments, after 26
Months
Total coralline cover fell by 19% in the cages
compared to a smaller loss of 5% in the par-
tial cages and a gain of 9% in the open plots.
By the end of the experiment, the dominant
herbivore-resistant Indo-Pacific species,
Porolithon onkodes, was 40% less abundant
inside the cages compared to elsewhere
(ANOVA, F = 7.2, p < 0.05), whereas a suite
of cryptic, understory taxa became increas-
ingly prevalent beneath the growing canopy
of fleshy seaweeds. Foliose Mesophyllum
purpurescens (Figure 1C) accounted for
24% of coralline cover below the macroalgal
canopy in the cages but was virtually absent
elsewhere. The decline and taxonomic shift
of corallines, which mimics similar trends on
degraded and overfished reefs, are signifi-
cant because of their role in promoting reef
accretion and coral settlement [S4]. Error
bars are SE.