Detection and Learning of Floral Electric Fields by Bumblebees

Abstract

Flowers and Bees Have “Sparks”
Plants and their pollinators have very intimate interactions, as emphasized by the many classic cases of coevolution among species of each. Such close relationships require signaling between plant and pollinator. Coordination between plant signals and pollinator perception has been shown to exist in flower color, shape, and odor. Clarke et al. (p. 66 , published online 21 February) report the potential for a distinct mode of plant–pollinator communication, electric fields. Natural floral electric fields, which are impacted by visits from naturally charged bees, were easily discriminated by bees, based on their level, pattern, and structure, and improved the rate at which bees remembered the location of a nectar reward.

Full text

mortality or breeding phenology, including fre-
quency of winter reproduction, should be part of
future research, as should quantifying vegetation
quality that could reflect, for example, continent-
scale long-term variation in climate or nutrient
deposition.
Irrespective of the proximate process (or pro-
cesses) affecting winter population growth rate, the
coherence of the changes coinciding with a period
of ongoing global environmental change suggests
increasingly frequent prolonged periods of low
amplitude, although high-amplitude vole peaks—
as seen in 2011 in northern Fennoscandia—may
occasionally reappear. The loss of years of super-
abundant voles could reduce zoonotic disease risk
and crop damage (27). Continent-scale collapses
in population cycles are likely to be deleterious
for vole predators because for most, reproduc-
tion is modulated by vole density in spring, which
is when the strongest and most consistent dam-
pening occurs. Large impacts on vegetation (6)
and predator populations (1,28) could see cas-
cading effects on other compartments of the food
webs (3,29) in ecosystems as diverse as farm-
land, forest, and arctic tundra.
References and Notes
1. B. Hörnfeldt, T. Hipkiss, U. Eklund, Proc. Biol. Sci. 272,
2045 (2005).
2. K. B. Strann, N. G. Yoccoz, R. A. Ims, Ecography 25,
81 (2002).
3. R. A. Ims, E. V. A. Fuglei, Bioscience 55, 311 (2005).
4. R. A. Ims, J.-A. Henden, S. T. Killengreen, Trends Ecol. Evol.
23, 79 (2008).
5. J. Bêty, G. Gauthier, E. Korpimäki, J.-F. Giroux, J. Anim.
Ecol. 71, 88 (2002).
6. J. Olofsson, H. Tommervik, T. V. Callaghan, Nat. Clim.
Change, published online 22 May 2012 (10.1038/
nclimate1537).
7. F. Ecke, P. Christensen, P. Sandström, B. Hörnfeldt,
Landscape Ecol. 21, 485 (2006).
8. T. Saitoh, B. Cazelles, J. O. Vik, H. Viljugrein, N. C. Stenseth,
Clim. Res. 32, 109 (2006).
9. S. M. Bierman et al., Am. Nat. 167, 583 (2006).
10. J. E. Brommer et al., Glob. Change Biol. 16, 577 (2010).
11. N. C. Stenseth, Oikos 87, 427 (1999).
12. K. L. Kausrud et al., Nature 456, 93 (2008).
13. T. F. Hansen, N. C. Stenseth, H. Henttonen, Am. Nat.
154, 129 (1999).
14. Materials and methods are available as supplementary
materials on Science Online.
15. R. A. Fleming, H. J. Barclay, J.-N. Candau, Ecol. Modell.
149, 127 (2002).
16. T. F. Hansen, N. C. Stenseth, H. Henttonen, J. Tast,
Proc. Natl. Acad. Sci. U.S.A. 96, 986 (1999).
17. L. Hansson, Oikos 40, 258 (1983).
18. I. Hanski, H. Henttonen, J. Anim. Ecol. 65, 220 (1996).
19. R. A. Ims, N. G. Yoccoz, S. T. Killengreen, Proc. Natl.
Acad. Sci. U.S.A. 108, 1970 (2011).
20. P. Turchin, L. Oksanen, P. Ekerholm, T. Oksanen,
H. Henttonen, Nature 405, 562 (2000).
21. B. Hörnfeldt, Oikos 107, 376 (2004).
22. D. M. Johnson et al., Proc. Natl. Acad. Sci. U.S.A. 107,
20576 (2010).
23. J. R. Bell et al., Ecol. Lett. 15, 310 (2012).
24. L. Korslund, H. Steen, J. Anim. Ecol. 75, 156 (2006).
25. N. C. Stenseth et al., Proc. Natl. Acad. Sci. U.S.A. 100,
11478 (2003).
26. M. J. Smith, A. White, X. Lambin, J. A. Sherratt, M. Begon,
Am. Nat. 167, 695 (2006).
27. N. C. Stenseth et al., Front. Ecol. Environ 1, 367
(2003).
28. N. M. Schmidt et al., Proc. R Soc. B 279, 4417
(2012).
29. R. W. Summers, L. G. Underhill, E. E. Syroechkovski,
Ecography 21, 573 (1998).
30. T. Royama, Analytical Population Dynamics, M. B. Usher,
Ed. (Chapman & Hall, London, 1992), pp. 55–60.
Acknowledgments: This research was funded by the national
funders Natural Environment Research Council, Research
Council of Norway, and Agence Nationale de la Recherche,
part of the 2008 ERA-Net BiodivERsA call for research
proposals. D.A.E. was funded by the Scottish Government.
We thank R. B. O’Hara and F. Barraquand for useful
methodological inputs and all the contributors to vole
sampling over decades. General correspondence should be
addressed to X.L. and specific requests to T.C. (cornulier@
abdn.ac.uk). Raw data are available in supplementary text
section IX). Authors declare no conflicts of interest. A.B., B.H.,
B.J., C.I., E.F., E.T., F.E., H.H., H.P., J.E.B., J.J., K.Z., O.H., S.J.P.,
V.B., and X.L. led the data collection; T.C., N.G.Y., R.A.I.,
A.M., and X.L. conceived the ideas for the paper and its
structure; T.C., N.G.Y., D.A.E., and X.L. designed the analyses;
T.C. and X.L. wrote the manuscript; and all authors discussed
the results and commented on the manuscript.
Supplementary Materials
www.sciencemag.org/cgi/content/full/340/6128/63/DC1
Materials and Methods
Supplementary Text
Figs. S1 to S6
BUGS Code
Data
References (31–47)
17 August 2012; accepted 12 February 2013
10.1126/science.1228992
Detection and Learning of Floral
Electric Fields by Bumblebees
Dominic Clarke,* Heather Whitney,* Gregory Sutton, Daniel Robert†
Insects use several senses to forage, detecting floral cues such as color, shape, pattern, and
volatiles. We report a formerly unappreciated sensory modality in bumblebees (Bombus terrestris),
detection of floral electric fields. These fields act as floral cues, which are affected by the visit of
naturally charged bees. Like visual cues, floral electric fields exhibit variations in pattern and
structure, which can be discriminated by bumblebees. We also show that such electric field
information contributes to the complex array of floral cues that together improve a pollinator’s
memory of floral rewards. Because floral electric fields can change within seconds, this sensory
modality may facilitate rapid and dynamic communication between flowers and their pollinators.
Flowers produce a diverse range of cues
and attractants to pollinators and in doing
so act as sensory billboards (1). The di-
versity of floral cues encompasses intricate color
hues and patterns, petal texture, fragrant volatiles,
local air humidity, and echolocation fingerprints
(1–4). The impact of floral cues on pollinator be-
havior has been observed since Aristotle (5), yet
new floral cues are still being discovered (3,4).
Multimodal floral cues have been found to en-
hance both pollinator foraging efficiency and
pollination (6), and thus facilitate increased seed
and fruit set.
Flying insects, including pollinators like honey-
bees, usually possess a positive electric poten-
tial (7–10). Conversely, flowers often exhibit a
negative potential (7,11). Electric fields arising
as a result of this potential difference between
flowers and insects promote pollen transfer and
adhesion over short distances (7,8,12,13). Fur-
thermore, these fields differ according to the
pollination status of the flower, as the deposi-
tion of pollen and resulting pollination changes
flower electric potential (14,15). However, the
use of electric fields by pollinators as informa-
tive cues has not been investigated. In the com-
plex world of plant-pollinator interactions, any
cue that increases pollination and foraging effi-
ciency should be mutually beneficial. Here, we
report that bumblebees can detect and learn to
use floral electric fields, and their structural var-
iation, to assess floral reward and discriminate
among flowers.
The electrical interactions between the bee
and the flower arise from the charge carried by
the bee and the potential of the flower in rela-
tion to the atmospheric electric field. To quan-
tify bee charge, individual B. terrestris workers
were trained to fly into a Faraday pail that con-
tained a sucrose reward. The net charge qcar-
ried by the bee was measured from the induced
voltage on a calibrated capacitor (methodology
described in supplementary materials). Measur ed
on 51 individuals, 94% of bees were positively
charged and 6% negatively charged (q
mean
=32T
5 pC, SD = 35pC) (Fig. 1A). These results cor-
roborate previous measurements on the honeybee
Apis melifera (9) and establish that the majority
of bees flying in the arena carry a positive charge
susceptible to transfer.
Electrical interaction between bee and flower
was further explored by placing Petunia integrifolia
flowers in an arena with free-flying foraging
bees. The electric potential in Petunia stems was
recorded to assess the electrical signature produced
by the approach and landing of an individual
charged bee. Charge transfer to the flower re-
sulted in a positive change in electric potential
recorded in the stem. The landing of 50 indi-
School of Biological Sciences, University of Bristol, Woodland
Road, Bristol BS8 1UG, UK.
*These authors contributed equally to this work.
†Corresponding author. E-mail: d.robert@bristol.ac.uk
5 APRIL 2013 VOL 340 SCIENCE www.sciencemag.org66
REPORTS

viduals resulted in a mean potential change last-
ing ~100 s, which peaked at ~25 T3mV(SD=
24, n= 50) (Fig. 1B). Such change exceeds
natural fluctuations in the absence of bees (Fig.
1B) and outlasts the presence of the bee on the
flower. This change in potential is often initiated
before contact with the bee (movie S1), sug-
gesting that this is not simply a hydraulic wound-
response variation potential as in (16) but involves
direct electrostatic induction between the charged
bee and the grounded flower as hypothesized
in (7,8).
Because the floral electric potential is di-
rectly affected by pollination (14,15)andbee
visitation (Fig. 1B), it potentially carries infor-
mation for other visiting pollinators regarding
floral resources. Visiting pollinators affect floral
cues directly, by leaving scent marks on the pe-
tal surface, or by initiating changes in floral cues,
such as color, shape, and humidity (4,17–19).
Such changes typically occur in the time frame
of minutes to hours. The variation potential pro-
duced by bee visitation occurs within a time
frame of seconds (Fig. 1B).
For a floral electric field to act as a cue, it
must be possible for pollinators to detect and
discriminate it from the background. We used
differential conditioning (3) to test the ability
of bumblebees to discriminate between artifi-
cial flowers (E-flowers) with differing electric
fields. E-flowers consisted of a 35-mm-diameter
by 1.5-mm-thick steel base disk decorated with
a purple epoxy top disk. Half the E-flowers
were held at a biologically relevant 30-V dc bias
voltage. This voltage was chosen as a proxy for
the electric field of an isolated flower standing
30-cm tall in a typical 100 V m
−1
atmospheric
electric field (20). Charged E-flowers offered a
sucrose reward, while identical E-flowers were
held at ground (0 V) and provided a bitter qui-
nine hemisulfate solution (3). E-flowers were
indistinguishable in every other respect. During
the course of 50 bee visits, there was an increase
in the relative number of visits to rewarding
charged flowers (Fig. 2A). To measure bee learn-
ing, we compared the mean accuracy of the final
10 visits (visit 41 to 50) to a random choice
model. In their final 10 visits to 30-V charged
E-flowers, bees (n= 11) achieved 81 T3%
accuracy (T
1-sample
= 10.8, P=7.4×10
−7
). Both
flower types were then grounded and the choice
test continued. Without the electric cue, the same
set of trained bees could no longer discriminate
between the rewarding and unrewarding E-flowers,
also demonstrating the absence of systematic
experimental bias. Accuracy after the electric cue
is removed was 54 T4%, which does not differ
significantly from random choice (T
1-sample
=1,
P= 0.35) (Fig. 1B). Using a 10-V bias failed to
elicit significant learning (n= 10, mean accu-
racy = 56 T4%, T
1-sample
=1.4,P= 0.19) (Fig.
2, A and B).
Floral cues are diverse and address the mul-
timodal perception of pollinators. Working in
concert, floral cues enhance foraging efficien-
cy (6) and constitute a complex informational
ecology of competing flower advertisement. Col-
or cues rely both on hue and on contrast between
hues and their geometrical patterns. Nectar
guides constitute such patterns, providing infor-
mation attractive to pollinators and facilitating
foraging efforts (21,22). By analogy, the geom-
etry of floral electric fields may carry additional
information important for pollinators. The diver-
sity of floral electric field geometry can be ex-
perimentally visualized by coating flowers with
positively charged colored particles released as
an aerosol close to the corolla. The heterogeneous
Fig. 1. Electric charge carried
by bumblebees and its trans-
fer to flowers. (A)Histogramof
electric charge of flying bumble-
bees. Boxplot shows median, SD,
interquartile range, and outliers.
(B) Mean variation potential in
the Petunia stem resulting from
bee landings (red, n=51),shown
with T1 SEM (gray). Distribution
of the natural variation of stem
potential (measured along 35 sam-
ples of 30 s) in absence of bees,
truncated at 2 SD (blue).
Count
Charge (pC)
A
0
5
10
15
20
25
0
Stem Potential (mV)
Time (s)
B
0
2
4
6
8
10
12
14
-110
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
80
90
Mean = 32pC
Median = 29pC
SD = 35pC
SEM = 5pC
n=51
100
8060
40
20
40
45
50
55
60
65
70
75
80
85
90
10 50
Visit Number
30V (n=11)
10V (n=10)
ON OFF
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
30V E-Flower 10V E-Flower
ON
56% 54%
54%
OFFON
81%
OFF
% Correct Choices
AB
% Correct Choices
30 402010
504030
20
Fig . 2. Bumblebees learn the presence of an electric field. (A) Learning curves of foraging bees, trained to 30-V (red diamonds) or 10-V (blue circles)
E-flowers. Dashed line shows switching off electric field. (B) Mean correct choices to 30-V (left) and 10-V (right) E-flowers over visits 41 to 50 in (A) during
training (voltage on) and control (voltage off). Error bars show SEM.
www.sciencemag.org SCIENCE VOL 340 5 APRIL 2013 67
REPORTS

pattern of color deposition reveals the struc-
ture of the electric field at the flower’s surface
(Fig. 3A).
Electric field structure was also visualized
using finite element (FE) modeling of an ide-
alized 30-cm-tall flower in a physically realis-
tic, 100 V m
−1
atmospheric electric field (20)
(Fig. 3B, left). Plants are conductively linked
to ground via their stems and roots, a connec-
tion that maintains them close to ground poten-
tial (7). Hence, a grounded 30-cm-tall plant in
such an atmospheric electric field exhibits a
30-V potential difference between its inflores-
cent structures and the surrounding air, exhib-
iting a patterned electric field (Fig. 3B). This
experimental and modeling evidence reveals
that flower morphology determines electric field
geometry.
To test the bee’s ability to discriminate E-field
geometries, differential conditioning was used
with two types of E-flowers, providing similar
voltage but different local patterns (Fig. 3C). Re-
warding E-flowers presented a bull’s eye pat-
tern, with the outer ring held at +20 V and the
center ring at –10 V. Aversive E-flowers presented
a homogeneous voltage at +20 V (Fig. 3C). Bees
(n= 10) learned to discriminate between these two
patterns, reaching 70 T3% accuracy over their
final 10 visits, performing significantly better than
random choice (T
1-sample
=6.7,P=8×10
−5
)
(Fig. 3E). After this task, a subset of the bees
(n= 4) was allowed to complete 50 additional
visits to rewarding and aversive E-flowers with
identical homogeneous +20 V fields. These bees
failed to discriminate between E-flowers (Fig.
3E). Altogether, these tests show that bumble-
bees can discriminate charged from uncharged
flowers and can distinguish between flowers that
differ in the geometry of their electric field. As
such, E-fields could be used by flowers to pro-
vide information to their pollinators.
Floral cues can work individually or com-
plementarily (1,6). When presented together,
multimodal cues enhance the certainty of sen-
sory information used by honeybees. Specifical-
ly, the association of color with olfactory floral
cues reduces the bees’perceptual uncertainty
relatedtoanindividualfloralcueandincreases
their ability to distinguish between rewarded
Fig. 4. Multimodal fa-
cilitation. Colors (A)and
voltage configurations (B)
associated with rewarding
and aversive E-flowers. (C)
Mean number of visits taken
by bees in each group to
reach 80% correct choices.
Hue + E-Field Hue Only
0
5
10
15
20
25
30
35
40
24
35
Mean # of visits
Hue Stimuli
120°
Sucrose
140°
Quinine
-
+
++
E-Field Stimuli
Sucrose
Quinine
A
B
C
Fig. 3. Geometry of floral
electric field and discrim-
ination task. (A)Flowers
before (left half) and after
(right) spraying with electro-
static colored powder; (a)
Gerbera hybrida,(b)Digital-
is purpurea,(c)Geranium
magnificum,(d)Calibrachoa
hybrida,(e)Petunia hybrida,
(f) Clematis armandii.Den-
sity of powder deposition
reflects the variation in elec-
tric field strength at the flow-
ers’surface. (B) FE model of
an idealized 30-cm-tall flow-
er, equipotential with ground,
in an atmospheric field of
100 V/m. Left: scalar electric
potential. Right: electric field
magnitude. (C)FEmodels
of electric field produced by
E-flowers. (D) Color scale for
(B) and (C). (E) Pattern dis-
crimination as mean per-
centage of correct choices
over the last 10 visits for
patterns on and off. Error
bars show SEM.
5 APRIL 2013 VOL 340 SCIENCE www.sciencemag.org
68
REPORTS

and aversive stimuli (23). The hypothesis can
be formulated that the floral electric field re-
inforces the effectiveness of other floral cues.
If true, an electric cue paired with a color cue
should produce an enhanced learning outcome
equivalent to that obtained with the test using
color and scent. Differential conditioning was
used to test this hypothesis. The same two green
target hues were used as in (23), but olfactory
cues were replaced with a patterned electric field
(Fig. 3C). Bees were trained to discriminate be-
tween E-flowers of hue 120°HSB (hue, satura-
tion, brightness) which offered a sucrose reward,
and E-flowers of hue 140°HSB, which provided
an aversive quinine solution (Fig. 4A). Bees
learned to discriminate between the rewarding
and aversive chargeless E-flowers either using
color information alone (n= 16) or in combi-
nation with the patterned E-field (n= 18) (Fig.
4A). When learning color on its own, discrim-
ination to 80% success (i.e., 8 out of the last 10
choices correct) took 35 T3 visits. When com-
bined with the E-field pattern, the number of
visits required was significantly reduced to 24 T
3(T
2-sample; unequal
= 2.86, P= 0.008) (Fig. 4A).
This demonstrates that the combination of two
cues, E-field and hue, enhances the bee’s ability
to discriminate.
Our results show that electric field consti-
tutes a floral cue. Contributing to a varied floral
display aimed at pollinator senses, electric fields
act to improve both speed and accuracy with
which bees learn and discriminate rewarding re-
sources. As such, electric field sensing consti-
tutes a potentially important sensory modality,
which should be considered alongside vision
and olfaction. The ubiquity of electric fields in
nature and their integration into the bees’sen-
sory ecology suggest that E-fields play a thus
far unappreciated role in plant-insect interac-
tions. The present study raises the possibility of
reciprocal information transfer between plants
and pollinators at time scales of milliseconds
to seconds, much faster than previously de-
scribed alterations in floral scent, color, or hu-
midity (4,18,19). The remarkably accurate
discrimination and learning of color patterns
by bees was revealed by both laboratory and
field training experiments (19,21–23). Sim-
ilarly, the present laboratory study reveals that
floral electric fields occur in patterns and that
they can be perceived. Hence, our study pro-
vides a framework for exploring the function
and adaptive value of the perception of weak
electric fields by bees in nature.
References and Notes
1. R. A. Raguso, Curr. Opin. Plant Biol. 7, 434 (2004).
2. R. Simon, M. W. Holderied, C. U. Koch, O. von Helversen,
Science 333, 631 (2011).
3. H. M. Whitney et al., Science 323, 130 (2009).
4. M. von Arx, J. Goyret, G. Davidowitz, R. A. Raguso,
Proc. Natl. Acad. Sci. U.S.A. 109, 9471 (2012).
5. Aristotle, Historia Animalium (Harvard Univ. Press,
Cambridge, MA, 1970).
6. A. S. Leonard, A. Dornhaus, D. R. Papaj, Curr. Zoology
57, 215 (2011).
7. S. A. Corbet, J. Beament, D. Eisikowitch, Plant Cell
Environ. 5, 125 (1982).
8. Y.Vaknin,S.Gan-Mor,A.Bechar,B.Ronen,D.Eisikowitch,
Plant Syst. Evol. 222, 133 (2000).
9. M. E. Colin, D. Richard, S. Chauzy, J. Bioelectric. 10,17
(1991).
10. Y. K. Yes’kov, A. M. Sapozhnikov, Biophysics (Oxf.) 21,
1124 (1976).
11. G. E. Bowker, H. C. Crenshaw, Atmos. Environ. 41,
1587 (2007).
12. E. H. Erikson, S. L. Buchmann, in Handbook of
Pollination Biology, C. E. Jones, R. J. Little, Eds.
(Van Nostrand Reinhold, New York, 1983).
13. S. Gan-Mor, Y. Schwartz, A. Bechar, D. Eisikowitch,
G. Manor, Can. Agric. Eng. 37, 189 (1995).
14. W. N. We¸dzony, M. Filek, Acta Physiol. Plant. 20, 291
(1998).
15. J. Fromm, M. Hajirezaei, I. Wilke, Plant Physiol. 109,
375 (1995).
16. B. Stanković, T. Zawadzki, E. Davies, Plant Physiol. 115,
1083 (1997).
17. J. C. Stout, D. Goulson, Anim. Behav. 62, 183 (2001).
18. M. R. Weiss, Am. J. Bot. 82, 167 (1995).
19. P. Willmer, D. A. Stanley, K. Steijven, I. M. Matthews,
C. V. Nuttman, Curr. Biol. 19, 919 (2009).
20. M. J. Rycroft, S. Israelsson, C. Price, J. Atmos. Sol. Terr. Phys.
62, 1563 (2000).
21. N. M. Waser, M. V. Price, Nature 302, 422 (1983).
22. A. S. Leonard, D. R. Papaj, Funct. Ecol. 25, 1293 (2011).
23. A. S. Leonard, A. Dornhaus, D. R. Papaj, J. Exp. Biol. 214,
113 (2011).
Acknowledgments: This work was sponsored by a grant from
the Leverhulme Trust (RPG 173). H.W. is supported by the
European Research Council and Association for the Study of
Animal Behaviour. D.R. is supported by the Royal Society of
London. The authors declare no conflict of interest. All data
are available in the supplementary materials. We thank
K. Strickland for help with data collection and C. Evans for
illustrative work. We thank A. Radford, J. Matthews, and
S. Rands for reading the manuscript and helpful feedback.
Supplementary Materials
www.sciencemag.org/cgi/content/full/science.1230883/DC1
Materials and Methods
References (24–26)
Movie S1
Data File S1
1 October 2012; accepted 5 February 2013
Published online 21 February 2013;
10.1126/science.1230883
Recovery of an Isolated Coral Reef
System Following Severe Disturbance
James P. Gilmour,
1
*Luke D. Smith,
1
†Andrew J. Heyward,
1
Andrew H. Baird,
2
Morgan S. Pratchett
2
Coral reef recovery from major disturbance is hypothesized to depend on the arrival of propagules
from nearby undisturbed reefs. Therefore, reefs isolated by distance or current patterns are
thought to be highly vulnerable to catastrophic disturbance. We found that on an isolated reef
system in north Western Australia, coral cover increased from 9% to 44% within 12 years of a
coral bleaching event, despite a 94% reduction in larval supply for 6 years after the bleaching.
The initial increase in coral cover was the result of high rates of growth and survival of remnant
colonies, followed by a rapid increase in juvenile recruitment as colonies matured. We
show that isolated reefs can recover from major disturbance, and that the benefits of their isolation
from chronic anthropogenic pressures can outweigh the costs of limited connectivity.
Coral reefs are dynamic ecosystems pe-
riodically subjected to severe disturbances,
such as cyclones, from which they typ-
ically recover at scales of one to two decades
(1,2). Today, this recovery is undermined by
increasing anthropogenic pressures leading to
global declines in coral cover (3,4) and diver-
sity (5,6). Understanding the global degradation
of coral reef ecosystems requires long-term data
on population and community dynamics, espe-
cially demographic processes (7–9). However,
the rarity of such data has precluded a thorough
assessment of the future of coral reef ecosystems
in the IPCC report on climate change (10,11),
and current knowledge is mostly derived from
studies of reef degradation (9,12)ratherthanreef
recovery. Here, we document the recovery of coral
assemblages at Australia’s largest oceanic reef
system, where changes in assemblage structure
and key demographic parameters were quantified
for 16 years, through a regime of disturbances
beginning with a catastrophic mass bleaching
event in 1998.
The Scott system of reefs is surrounded by
oceanic waters on the edge of Western Australia’s
continental shelf. It is more than 250 km from
the mainland and other reefs in the region, and
more than 1000 km from a major center of ur-
banization (fig. S1). There is little fishing pres-
sure at the reefs, apart from the harvesting of
sea cucumber, trochus, and shark fin by In-
donesian islanders using traditional fishing meth-
ods for more than 300 years (13,14). Such oceanic
reef systems may provide a critical refuge for
coral reef assemblages because they are far re-
moved from most direct anthropogenic pres-
sures. Conversely, isolation and a consequent
lack of connectivity may make such systems
1
Australian Institute of Marine Science (AIMS), University of
Western Australia Oceans Institute, Perth, WA 6009, Australia.
2
ARC Centre of Excellence for Coral Reef Studies, James Cook
University, Townsville, QLD 4811, Australia.
*Corresponding author. E-mail: j.gilmour@aims.gov.au
†Present address: Woodside Energy Limited, Perth, WA 6000,
Australia.
www.sciencemag.org SCIENCE VOL 340 5 APRIL 2013 69
REPORTS