Ecological Monographs, 77(4), 2007, pp. 503­525
Ó 2007 by the Ecological Society of America
EFFECTS OF CLIMATE AND SEAWATER TEMPERATURE VARIATION
ON CORAL BLEACHING AND MORTALITY
TIMOTHY R. MCCLANAHAN,1,2,5 MEBRAHTU ATEWEBERHAN,2 CHRISTOPHER A. MUHANDO,3 JOSEPH MAINA,2
4
AND MOHAMMED S. MOHAMMED
1Wildlife Conservation Society, Marine Programs, Bronx, New York 10460 USA
2Coral Reef Conservation Project, Wildlife Conservation Society, P.O. Box 99470, Mombasa, Kenya
3Institute of Marine Sciences, University of Dar es Salaam, P.O. Box 668, Zanzibar, Tanzania
4Department of Science, State University of Zanzibar, P.O. Box 146, Zanzibar, Tanzania
Abstract.
Coral bleaching due to thermal and environmental stress threatens coral reefs
and possibly people who rely on their resources. Here we explore patterns of coral bleaching
and mortality in East Africa in 1998 and 2005 in a region where the equatorial current and the
island effect of Madagascar interact to create different thermal and physicochemical
environments. A variety of temperature statistics were calculated, and their relationships
with the degree-heating months (DHM), a good predictor of coral bleaching, determined.
Changes in coral cover were analyzed from 29 sites that span .1000 km of coastline from
Kenya to the Comoros Islands. Temperature patterns are influenced by the island effect, and
there are three main temperature environments based on the rise in temperature over 52 years,
measures of temperature variation, and DHM. Offshore sites north of Madagascar that
included the Comoros had low temperature rises, low DHM, high standard deviations (SD),
and the lowest relative coral mortality. Coastal sites in Kenya had moderate temperature rises,
the lowest temperature SD, high DHM, and the highest relative coral mortality. Coastal sites
in the south had the highest temperature rises, moderate SD and DHM, and low relative coral
mortality. Consequently, the rate of temperature rise was less important than background
variation, as reflected by SD and kurtosis measures of sea surface water temperature (SST), in
predicting coral survival across 1998. Coral bleaching responses to a warm-water anomaly in
2005 were also negatively related to temperature variation, but positively correlated with the
speed of water flow. Separating these effects is difficult; however, both factors will be
associated with current environments on the opposite sides of reefs and islands. Reefs in
current shadows may represent refugia where corals acclimate and adapt to environmental
variation, which better prepares them for rising temperature and anomalies, even though these
sites are likely to experience the fastest rates of temperature rise. We suggest that these sites are
a conservation priority and should be targeted for management and further ecological
research in order to understand acclimation, adaptation, and resilience to climate change.
Key words:
acclimation/adaptation; climate change; coral bleaching; coral cover; degree-heating
weeks/months (DHW/DHM); East Africa; Indian Ocean; island effects; sea surface water temperature
(SST); temperature history; temperature variation; water flow.
INTRODUCTION
management activities (West and Salm 2003, Wool-
Coral reefs are increasingly threatened by a variety of
dridge and Done 2004, Wooldridge et al. 2005).
factors, including high and destructive fishing, sedimen-
Implementation requires an evaluation of large-scale
tation, pollution, and climate change (McClanahan
patterns of environmental variation, changes over time,
2002, Hughes et al. 2003). Climate change presents a
projections into the future, and consequences of
unique challenge as the effects are broad scale and not
environmental change for the acclimation/adaptation
easily alleviated by local action or management (Hughes
and persistence of the studied organisms and ecosys-
et al. 2005). Therefore, a key concern of modern marine
tems. Additionally, to be confident about predictions, an
ecology and associated management is to identify sites
improved understanding of the causation and the time
or regions that are likely to persist as the climate changes
scale of environmental variation, acclimation/adapta-
and to develop management that will improve the
tion, and persistence are needed. The patterns of
chances for persistence. There is growing awareness of
causation are complex and interactive, sometimes
this need and to increase associated research and
counterintuitive, and an area for growth in scientific
understanding (McClanahan et al. 2005a).
Manuscript received 14 July 2006; revised 8 February 2007;
Coral reefs are among the ecosystems most threatened
accepted 26 March 2007; final version received 4 May 2007.
by climate change as corals live near their upper thermal
Corresponding Editor (ad hoc): T. E. Essington.
5
limits and are sensitive to modest increases in back-
Present address: P.O. Box 99470, Mombasa, Kenya.
E-mail: tmcclanahan@wcs.org
ground seasonal seawater temperatures (Kleypas et al.
503

---

504
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
2001, Coles and Brown 2003, Guinotte et al. 2003,
Clanahan et al. 2003) are all known to mediate the
Berkelmans et al. 2004, Jokiel 2004). Seawater temper-
bleaching effect and influence the elasticity of the
atures are predicted to rise further and cause coral
threshold. Therefore, environments with attenuating
mortality in the next few decades (Hoegh-Guldberg
properties are likely to improve the chances for
1999, Hughes et al. 2003, Sheppard 2003). Elevated
persistence of corals (Glynn 2000, Riegl and Piller
temperature interacting with other physicochemical,
2003, West and Salm 2003, Barton and Casey 2005).
biological, and anthropogenic factors is generally
Better understanding of these factors will improve
considered to be the primary stress causing coral
predictions, and conservation and management priorities
bleaching (Coles and Brown 2003). The understanding
can be established, but only once the effects of these
of water temperature regimes and organism survival is
environmental factors have been tested with field studies
based on descriptive and often site-specific field studies
at an appropriate scale (McClanahan et al. 2005a).
(McField 1999, Loya et al. 2001, McClanahan et al.
This paper advances the understanding by examining
2001, Mumby et al. 2001a, Aronson et al. 2002,
the spatial patterns of temperature variation and the
Berkelmans 2002, Berkelmans et al. 2004) and labora-
recent temperature rise in the East African Coastal
tory experiments, usually with limited species, factors,
Current System from 1951 to 2002 and its possible
acclimation and evolutionary histories, and time (Jokiel
relationship with the observed spatial variation in coral
and Coles 1977, Berkelmans and Willis 1999, D'Croz et
mortality that followed the 1998 El Nin~o Southern
al. 2001, Jones and Hoegh-Guldberg 2001, Nakamura
Oscillation (ENSO) event. The primary focus was to
and van Woesik 2001, Nakamura et al. 2003, Warner et
examine the interaction between the equatorial and
al. 2002, Lesser and Farrell 2005). In addition, compar-
coastal current system and the island of Madagascar and
ing coral bleaching and mortality in sites with different
how their interaction influences the oceanographic and
environmental backgrounds has been insightful as they
thermal environments and the potential for these
indicate the importance of the interaction between the
environments to create acclimation/adaptation or refuge
environment and corals on a scale large enough to be
from climate change. Because this island­current inter-
relevant for field predictions and management priorities
action is widespread, our findings should have equally
(Sheppard 1999, Glynn et al. 2001, Podesta and Glynn
widespread application. Additionally, we examined the
2001, Berkelmans 2002, McClanahan and Maina 2003,
influence of a variety of environmental factors on a
Riegl 2003, Riegl and Piller 2003, McClanahan et al.
milder bleaching event in 2005 in an effort to distinguish
2005b, 2007a, b). These larger and synthetic studies lend
the dominant attributes of temperature and other
insight into the species, sites, and factors that are
environmental variations on the intensity of bleaching.
expected to lead to persistence of corals and the coral
The combination of these studies increases the chances
reef ecosystem (Glynn 2000, McClanahan 2002).
for determining causation and also factors that are good
The scientific and management challenge is to scale the
proxies for this causation. The ultimate purpose is to
above studies and important findings up to larger regions
identify the spatial and temporal structure of the
(Sheppard 2003, Wooldridge and Done 2004, Sheppard
temperature rise and coral mortality to assist in making
and Rioja-Nieto 2005), to test regional environmental
predictions concerning the expected temperature in-
models with field data, and to improve the understanding
crease on reef organisms and to set associated regional
of causation between environmental histories, acclima-
priorities for research and conservation. The other
tion/adaptation, and survival in order to evaluate and
principal objective of this study is to determine if useful
predict the future of climate on corals (Wooldridge et al.
indicator variables can be distinguished and also to
2005). Currently, we know that high SSTs above some
identify areas that will either not experience high
site-specific threshold will lead to coral bleaching and
environmental stress or that have high resistance to
mortality but that this threshold is sensitive to the taxa,
stress and quick recovery, both of which are paramount
regions, and associated environmental backgrounds
to our understanding of coral reef resilience in the face
(Coles et al. 1976, Coles and Brown 2003, Jokiel 2004,
of global climate change. Because SST is believed to be
McClanahan 2004, McClanahan et al. 2004). There is
the dominant factor causing coral bleaching and
considerable progress, but it is still less clear how the
mortality and because there are data sets of moderate
background temperature and environmental history
spatial and temporal resolution, we explore variation in
influences acclimation, adaptation, and persistence with-
SST and coral mortality due to bleaching in East Africa
in regions and different physicochemical environments
on the scale of ;1000 km and ;50 years. We also
(Riegl and Piller 2003). Past bleaching (Baker et al. 2004,
combine these studies with more environmental vari-
Berkelmans and van Oppen 2006), water flow (Naka-
ables to improve the chances for understanding the
mura and van Woesik 2001, McClanahan et al. 2005b),
difference between causations and proxies of causation.
temperature variation (Podesta and Glynn 2002, Mc-
METHODS
Clanahan and Maina 2003), persistence of cool water
(Glynn et al. 2001, Riegl and Piller 2003), light (Iglesias-
Study area and background
Prieto et al. 1992, Mumby et al. 2001b, Lesser and Farrell
East Africa is well known for its dynamic oceano-
2004, Gill et al. 2006), and inorganic nutrients (Mc-
graphic setting, covering several latitudes and different

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
505
FIG. 1.
Maps of (A) sea surface topography, (B) distribution of the cells used in the HadISST analysis, (C) photosynthetically
active radiation (PAR), and (D) chlorophyll a in East Africa. Patterns roughly represent the hydrodynamics in the East African
Coastal Current System and are more or less consistent among years and seasons. Sea surface topography data for 20 April 2005
are available at hhttp//sealevel.jpl.nasa.gov/science/data.htmli. Data for chl a and PAR are available at hhttp://oceancolor.gsfc.
nasa.gov/cgi/level3.pl?DAY¼&PER¼&TYP¼swchl&RRW¼16i.
marine habitats influenced by the Indian Ocean mon-
northeast monsoon from November to March, but this
soon seasonality, tropical currents, and different topog-
only affects the most northern part of the study site.
raphies (Benny 2002; Fig. 1). The Equatorial Current in
Wind, wave height, and current data suggest that the
East Africa is displaced south of the equator such that it
highest physical energy is found at the northern tip of
passes over northern Madagascar and swings north
Madagascar and then in the northern part of this region.
along the East African coastline where it is known as the
Madagascar and to some extent the islands of Tanzania
East African Coastal Current (Benny 2002). There is a
(Songo Songo, Mafia, Zanzibar, and Pemba) act as
temporary reversal of the surface current during the
barriers to the larger Indian Ocean physical energy and

---

506
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
TABLE 1.
Geographic positions, country names, and study sites in East Africa; coral cover before and after the 1998 bleaching
event; and changes in cover relative to cover before the 1998 bleaching.
Country
Cover
Cover after
Absolute
and location
Reef/site
Latiude
Longitude
before 1998 (%)
1998 (%)
change (%)
Comoros
Comoros (CM)
Mitsamuli
À12.21
43.52
55
40
15
Comoros (CM)
Itsamia
À12.21
43.52
54
36
18
Kenya
Kenya (KN)
Diani
À4.2117
39.3469
21.21
8.34
12.87
Kenya (KN)
Ras Iwatine
À4.0117
39.4357
17.78
5.26
12.52
Kenya (KN)
Mombasa
À3.593
39.4502
41.98
18.75
23.23
Kenya (KN)
Kanamai
À3.5524
39.4721
22.03
21.51
0.52
Kenya (KN)
Vipingo
À3.4796
39.5017
19.81
7.86
11.95
Kenya (KN)
Watamu
À3.2259
39.5933
37.67
8.99
28.68
Kenya (KN)
Malindi
À3.1534
40.0844
44.22
11.09
33.13
Kenya (KN)
Kisite
À4.709
39.3703
22.23
9.98
12.25
Tanzania
Mafia (MA)
Tutia
À8.06552
39.38546
80
13
67
Mafia (MA)
Mange
À8.03063
39.35301
70
12
58
Mafia (MA)
Juani
À7.58518
39.48518
70
11
59
Mnazi Bay (MB)
Chumbo
À10.19162
40.21349
60
20
40
Mnazi Bay (MB)
Cha Kati
À10.18537
40.2208
60
20
40
Pemba (PM)
Misali3
À5.15127
39.35066
30
5
25
Pemba (PM)
Misali1
À5.14137
39.35396
74
17
57
Pemba (PM)
Misali2
À5.14012
39.35453
57
7
50
Songo songo (SS)
Amana
À8.42187
39.2622
35
30
5
Songo songo (SS)
Jewe
À8.38059
39.26151
30
30
0
Tanga (TN)
Upangu
À5.1857
39.04444
50
42
8
Tanga (TN)
Chanjale
À5.18308
39.03529
45
30
15
Tanga (TN)
Taa
À5.17357
39.05461
67
16
51
Tanga (TN)
Kitanga
À5.1713
39.04304
50
45
5
Unguja (ZN)
Kwale
À6.22519
39.16569
30
15
15
Unguja (ZN)
Chumbe
À6.16291
39.1025
52
42
10
Unguja (ZN)
Bawe
À6.08326
39.08051
53
45
8
Unguja (ZN)
Chapwani
À6.07174
39.11351
44
25
19
Unguja (ZN)
Changuu
À6.06522
39.10053
50
33
17
Notes: Absolute change is the difference between cover before 1998 and after the 1998 bleaching mortality. Relative change is the
absolute cover over cover before 1998. Abbreviations are country and region names: KN is Kenya; MA is Mafia, southern-central
Tanzania; MB is Mnazi Bay, southern Tanzania; PM is Pemba, northern Tanzania; SS is Songo Songo, southern-central Tanzania;
TN is Tanga, northern Tanzania; ZN is Zanzibar, northern Tanzania; CM is Comoros.
create a low-energy shadow that extends from Mozam-
readings from NOAA satellite images and calculated
bique to northern Tanzania (Fig. 1). This region is
from the HadISST and JCOMMSST (satellite images
influenced by the El Nin~o Southern Oscillation (ENSO;
available online).6 The HadISST is a set of quality-
Cole et al. 2000), and the Indian Ocean also contains its
controlled global long-term in situ SST data collected
own endogenous cycle known as the Indian Ocean
from ships and buoys (Rayner et al. 2003) and provided
Dipole (IOD), which is a pressure gradient between
as monthly means for 1 3 1 degree longitude­latitude
Sumatra and the Arabian Peninsula and has greatest
squares (data available online).7 JCOMM is a shorter
impacts in the eastern and western margins of the Indian
time series that includes SST data from satellites that are
Ocean (Saji et al. 1999, Webster et al. 1999, Kayanne et
corrected using those recorded from in situ measure-
al. 2006). The natural variability of the western Indian
ments and are presented as weekly and monthly means in
Ocean SST dipole and ENSO has been detected from
1 3 1 degree latitude­longitude squares (Reynolds et al.
coral skeleton isotope records of the past 300 years
2005; data available online).8 We undertook analysis of
(Charles et al. 1997, Cole et al. 2000, Zinke et al. 2004,
the HadISST data from the East African coast from 168 S
to 5
2005, Damassa et al. 2006, Kayanne et al. 2006).
8 N and from 388 E to 458 E from 1951 to 2002. In situ
temperatures were recorded using Hobo temperature
DATA SOURCE AND ANALYSES
loggers (Onset Corporation, Pocasset, Massachusetts,
USA). Loggers were placed at eight Kenyan reef
Temperature data
locations that covered ;200 km in a north­south
Four sources of temperature data were used in
direction along the coastline. Except the southern-most
determining spatial variability in East Africa: direct
7 hhttp://www.cru.uea.ac.uki
6 hhttp://www.osdpd.noaa.gov/PSB/EPS/SSTi
8 hhttp://iri.Columbia.edu/climate/monitoring/ipbi

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
507
TABLE 1.
Extended.
or studied before the 1998 bleaching event (McClanahan
et al. 1999, McClanahan et al. 2001, Muhando and
Mohammed 2002; Table 1). A mixture of haphazard and
permanent transects were used at each site. Permanent
Relative change
Reference
transects were recorded using 10­20 m Line Intercept
Transects (LIT). Haphazard transects were sampled
0.27
Quod and Bigot (2000)
using the Line Point Intercept (LPI) method by dividing
0.33
Quod and Bigot (2000)
a 50-m transect into 200 points at intervals of 25 cm.
Most sites were surveyed using 2­3 transects with the
0.61
McClanahan et al. (2001)
0.70
McClanahan et al. (2001)
exception of the sites on Mafia Island (9­10 transects).
0.55
McClanahan et al. (2001)
Published and unpublished data collected by McClana-
0.02
McClanahan et al. (2001)
han and colleagues (1997 and 1999; McClanahan et al.
0.60
McClanahan et al. (2001)
0.76
McClanahan et al. (2001)
2001, McClanahan and Maina 2003) were used for most
0.75
McClanahan et al. (2001)
Kenyan sites. For Kisite Island, there were no 1997 and
0.55
T. R. McClanahan, unpublished data
1999 benthic cover data, but data from 1996 and 2001
were available and used. Coral cover in Kenya was
0.84
Muhando and Mohammed (2002)
surveyed using 10-m LIT (n ¼ 9­12 transects at each
0.83
Muhando and Mohammed (2002)
site/reef). Information from Quod and Bigot (2000) was
0.84
Muhando and Mohammed (2002)
0.67
Muhando and Mohammed (2002)
used for the two Comoros sites where cover data were
0.67
Muhando and Mohammed (2002)
gathered according to the Global Coral Reef Monitor-
0.83
Muhando and Mohammed (2002)
ing Network (GCRMN) rapid LIT assessment. Nine-
0.77
Muhando and Mohammed (2002)
0.88
Muhando and Mohammed (2002)
teen of the Kenyan and Tanzanian sites had data on
0.14
Muhando and Mohammed (2002)
community structure and analysis of taxa-specific
0.00
Muhando and Mohammed (2002)
bleaching mortality was possible.
0.16
Muhando and Mohammed (2002)
0.33
Muhando and Mohammed (2002)
Bleaching response in 2005
0.76
Muhando and Mohammed (2002)
0.10
Muhando and Mohammed (2002)
Following the appearance of a hot spot in the
0.50
Muhando and Mohammed (2002)
southern Indian Ocean in January 2005, we surveyed
0.19
Muhando and Mohammed (2002)
0.15
Muhando and Mohammed (2002)
coral bleaching in eight reefs along the Kenyan coast in
0.43
Muhando and Mohammed (2002)
April 2005, a few weeks after peak water temperatures.
0.34
Muhando and Mohammed (2002)
Bleaching was defined as a loss of color and estimated by
evaluating the color intensity of haphazardly selected
corals within a radius of ;2 m (McClanahan et al. 2001,
site of Kisite, the reefs are lagoonal and isolated from the
McClanahan 2004). This process was repeated several
open ocean during low tide (tidal range of 4 m at spring
times, and between 486 and 992 coral colonies were
low tide) and were not expected to differ significantly in
sampled at each site and assigned to one of six
their physicochemical properties. This isolation causes
categories: (1) unbleached (normal coloration), (2) pale
the temperature in the lagoons to rise above that of the
(lighter color than usual for the time of the year), (3) 0­
open water during the warm season. The reefs differ in
20% of the surface bleached (white coral skeletal
their distance from shore (0.1­1.0 km), the height of the
coloration), (4) 20­50% bleached, (5) 50­80% bleached,
reef above the datum, their proximity to channels or reef
and (6) 80­100% bleached. The assessment of bleaching
depressions that connect to open sea (McClanahan and
from coloration, especially in colonies of the same
Maina 2003). The physical differences, particularly the
species and the presence of genetically different symbi-
height and isolation of the reef from the open ocean, were
onts with different environmental tolerances and light
expected to result in different SST environments,
absorption capacities, poses a number of difficulties
especially during the warm season (from February to
(Knowlton et al. 1992, Jones 1997, Enriquez et al. 2005).
April; McClanahan et al. 2001, McClanahan and Maina
The field method was originally developed by Gleason
2003). In situ temperature data are presented for the
(1993), modified by McClanahan et al. (2001), and tested
2003­2005 time period when data were available for all
between regions and observers (McClanahan et al.
sites. Means of the in situ data recordings from
2004); two related field and laboratory studies that
Mombasa, one of the sites most exposed to the open
found good correlations between color ranks and
ocean, are highly correlated to those from NOAA SST
pigment concentrations and cell densities (Edmunds et
(R2 ¼ 0.86; McClanahan et al. 2001).
al. 2003, Siebeck et al. 2006). Enriques et al. (2005)
suggest that light absorption did not increase greatly for
Coral cover before and after the 1998 bleaching episode
pigment concentrations above 20 mg chl a/m2 in Porites
Coral cover data before and after the bleaching event
banneri and light-absorption-based estimates of bleach-
in 1998 was obtained from the literature and field survey
ing will not be accurate above this pigment concentra-
data. We present data from reef areas that were surveyed
tion. Their data (Enriques et al. 2005: Fig. 3A) does,

---

508
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
however, suggest a nearly linear relationship for
correlation analysis. Single model regression analysis
absorption and pigment concentrations for samples
was used to predict the influence of the mean, SD,
below 50 mg/m2 of chl a and very low samples sizes
minimum, maximum, median, kurtosis, and skewness of
above this concentration. Other tests of color rank and
SST on the 1998 degree-heating months (DHM98). We
pigment concentrations below this level have found
present the best-fit model for either the linear or second
good and nearly linear relationships between color ranks
degree polynomial selecting the model with the largest
and light-absorption measures (Edmunds et al. 2003,
R2. The properties of the distribution of the temperature
Siebeck et al. 2006), and taking the three studies together
data are expected to have a significant effect on short-
would suggest a strong relationship with the range of
term acclimation and longer term adaptation, and
pigment concentrations that are commonly found in
therefore, characterization of the data included skewness
field situations during warm seasons and bleaching
and kurtosis. These two measures have been widely used
events. This method has been shown to produce
by physical scientists in describing distributions where
comparable results that will allow for scaling bleaching
skewness measures the distance of outliers and kurtosis
intensity and make comparisons between site, times, and
the flatness or peakiness of the data distribution. The
regional and global comparison (McClanahan et al.
most important SST variables were identified with a
2004, 2007a). Bleaching response was calculated as a
stepwise multiple regression models--the SD was
scaled percentage from the observations in each color
excluded due to the high correlation they have with
category and a Bleaching Index (BI) generated as
the other SST statistics, and SD is a component of
follows: BI (%) ¼ (0c1 þ 1c2 þ 2c3 þ 3c4 þ 4c5 þ 5c6)/5,
skewness and kurtosis.
where c is the percentage of observations in each of the
Degree heating refers to the time that a temperature is
six bleaching categories (McClanahan et al. 2005b).
above the mean maximum temperature for a specific site
Most studies of pigment and cell density concentrations
and is used to describe and predict the heating stress and
during bleaching events have found at least an order of
bleaching in corals (Liu et al. 2005, McClanahan et al.
magnitude reduction (Edmunds et al. 2003, Enriquez et
2007a). Often it is presented at NOAA web sites in
al. 2005, Siebeck et al. 2006). Therefore, the spread in
weeks (degree-heating weeks, DHW; data available
the BI response implied by this equation may be
online).9 For the HadISST data set it is, however,
somewhat lower than found for pigment and cell
calculated in months (degree-heating months, DHM), as
concentrations.
the data are compiled and presented as monthly means.
In order to test for a relationship between water flow
Currently NOAA's DHW is calculated as the cumula-
and bleaching, we estimated water flow speed using the
tive positive anomaly from the mean SST climatology of
dissolution of plaster of Paris (calcium sulfate) clod
the climatologically warmest month at a location (Liu
cards (Doty 1971) as described in McClanahan et al.
et al. 2005). It is presented as the accumulation of
(2005b). Water flow was sampled at each site, multiple
anomalies (also called hot spots) at a grid of 50 3 50 km
times, between 2003 and 2005, and each sample was
over a rolling 12-week time period, and only anomalies
based on the deployment of 4­6 clods that were fastened
of 18 and above are considered. Values ,18C are
to the reef surface such that the total sample number per
disregarded as insufficient to cause visible stress in
reef was between nine and 27. Tests of the clod
corals (Liu et al. 2005). Degree-heating weeks or months
dissolution method to predict water-flow speed in
for the HadISST and JCOMMSST data were calculated
experimental flumes have found strong relationships
for each grid as the cumulative positive anomalies above
(R2 . 0.96; Jokiel and Morrissey 1993, Anzai 2001).
the long-term mean summer maximum for the three
warmest months: February, March, and April (Barton
Data analysis
and Casey 2005). Anomalies were calculated from the
Temperature data.--To determine temporal trends in
long-term baseline: 1950­1997 for the HadISST data,
East Africa, we undertook regression analysis of the
1981­1997 for the JCOMMSST data, and direct
monthly means of the averages, minima, and maxima of
readings from satellite images for the NOAA data
the HadISST data with time and also by separating out
(1998) from February to April 1998. DHW were not
ENSO (El Nin~o­Southern Oscillation) and IOD (Indian
calculated for the in situ seawater temperature data from
Ocean Dipole) years. We used the ENSO and IOD years
Kenya, as the measurement time is not long enough to
provided by Saji et al. (1999), which do not include weak
provide a reliable climatology for all the sites.
events (ENSO years were 1957­1958, 1965­1966, 1972­
Change in coral cover.--Change in absolute coral
1973, 1982­1983, 1987­1988, 1991­1992, and 1997­
cover was determined by calculating the difference
1998; IOD years were 1961, 1963, 1967, 1972, 1977,
between the pre- and post-1998 bleaching coral cover
1982, 1991, 1994, and 1997; ENSO­IOD years where
for each site or reef area. The relative cover was
both events coincide were 1972, 1982, and 1997). Within
obtained as a percentage proportion of the absolute
each geographic cell in the HadISST database, a variety
cover over the prebleaching cover. A single model
of statistics of the SST data were calculated for the
regression analysis was conducted on the absolute cover
whole time series, and relationships among the different
temperature statistics were investigated with Spearman's
9 hhttp://www.osdpd.noaa.gov/IPD/IPD.htmli

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
509
as a function of the cover before 1998. In addition, the
relationship between change in cover and cover of
Acropora and non-Acropora before 1998 was analyzed in
order to determine whether the change in cover was due
to a higher differential mortality of Acropora, one of the
most dominant taxa in the Indo­Pacific and highly
susceptible to bleaching (see Plate 1). Temperature
variations from in situ recordings in Kenya, expressed
in coefficients of variation (CV) and standard deviations
(SD), were used to describe between-site and reef
variation. The effect of coral cover before 1998 on the
absolute change in coral cover was compared between
Kenya and Tanzania with a factorial ANOVA by taking
country as a fixed variable and cover before 1998 as a
continuous one. The effect of temperature variation (SD
and CV) from in situ recordings on the absolute and
relative change in coral cover was tested using a single
regression model for Kenyan sites.
Bleaching response in 2005.--The site specific Bleach-
ing Index (%BI) was calculated by pooling the BI of all
taxa in each site in Kenya where in situ temperature data
were available. The relationship between temperature
variation (SD) and water flow was determined with
single and multiple regression analysis. A stepwise
multiple regression model with BI as a response and
temperature SD and flow speed and their interaction as
predictor variables was used. All statistical analyses were
performed using JMP 5.1 for Mac (SAS Institute, Cary,
PLATE 1.
Branching corals such as this species of Acropora
North Carolina, USA) and SPSS 10.0 for Windows
are a complex but regular structure created by the coral animal,
(SPSS, Chicago, Illinois, USA), and spatial analysis and
the algal symbiont, and the calcium carbonate skeleton. Photo
credit: T. R. McClanahan.
mapping using ArcView GIS 3.2 (ESRI, Redlands,
California, USA).
R
due to the different patterns of these years. The mean
ESULTS
SSTs increased significantly during both non-ENSO and
Temporal variation in SST
ENSO years, respectively 0.00758C/yr and 0.0158C/yr (P
The mean and maximum SSTs in East African coastal
, 0.05), and the increases were higher during ENSO
waters increased at a rate of 0.018C/yr over the last ;50
than non-ENSO years (Factorial ANOVA: non-EN-
years (Fig. 2; R2 ¼ 0.37; P , 0.05). The minimum
SO/ENSO as fixed and year as continuous variable; P ¼
increased at a rate of 0.0078C/yr. This amounts to a total
0.038). The increases in maximum SSTs were also higher
rise of 0.58C of the mean and maxima and 0.358C of the
during ENSO (0.0218C/yr) than non-ENSO years
minima in the last half a century. Analysis of individual
(0.00658C/yr; P ¼ 0.006). Increase in the minimum SST
confidence interval (95%) identified that mean SSTs of
was significant only during non-ENSO years
1983, 1987, and 1998 were higher from the trend of SST
(0.0068C/yr; P ¼ 0.04). The y-intercept was highest
increase over time (P , 0.05). Similarly, maxima of 1983
during IOD/ENSO years for mean and minima; ENSO
and 1998 and minima of 1972 and 1983 were notably
years had higher values than non-ENSO years. IOD
higher. Minima of 1964 and 1996 were unusually lower
years had highest y-intercept for maxima; ENSO-IOD
(P , 0.05). Mean (27.888C) and maximum (30.008C) of
years were higher than ENSO and non-ENSO years.
1998 were the highest and minimum of 1964 (24.448C)
The three ENSO events of 1982­1983, 1987­1988, and
the lowest. Generally second degree polynomials had
1997­1998 had the highest recorded mean SSTs in East
good fit for all three SST parameters, but differences
Africa (Figs. 2 and 3). Comparisons of the three indicate
from the linear fit were small, respectively 0.7%, 4.2%,
that both mean and maximum of 1997­1998 were the
and 1.6% for the mean, maximum, and minimum.
highest. Minimum temperatures were highest in 1987­
Analysis of the individual spatial cells indicated consid-
1988; all three statistics were lower in 1982­1983 than in
erable spatial variation ranging from nonsignificant
1987­1988 and 1997­1998. The 1982­1983 ENSO did
changes, to linear, to weak but also included decreasing
not warm as early or as quickly as in 1987 or 1997 and
second degree polynomials and weak exponential rises.
the 1997­1998 event persisted longer than the other two
Non-ENSO, ENSO, IOD, and combination of IOD
events. The 1982­1983 SSTs were higher only in the end
and ENSO (IOD-ENSO) years were analyzed separately
of the warm season in June and July. The 1987­1988

---

510
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
trends (Figs. 4A­D and 5A) with values increasing to the
south from southern Somalia to the Comoros. Values
also decreased from the east off Madagascar to the west
toward Tanzania and Mozambique, but this decrease was
of a lower magnitude compared to the latitudinal trend.
Modal SSTs generally had a weak but decreasing trend in
a north­south direction (Fig. 5B). The few squares that
had lower values were scattered mainly in offshore
locations without a clear gradient in distribution.
The properties of the distribution of the temperature
data were tested and, according to a test of significance
for kurtosis and skewness (Tabachnick and Fidell 1996),
all 81 cells in East Africa had significant negative
kurtosis or a flat distribution; 45 had negative skewness
and five had positive skewness. The remaining 31 cells
FIG. 2.
Time series of annual mean, maximum, and
minimum HadISST increases in East Africa. Inner dotted lines
indicate 95% CI of the regression line, and outer lines indicate
95% CI of the individual points. Outliers are indicated. Data are
pooled means of all cells.
ENSO event had an earlier and stronger build up in the
cold season. During the hot season, January to April,
the 1997­1998 event was the strongest followed by the
1987­1988 for the mean, maximum, and minimum
SSTs.
Spatial variations in SST
The overall statistical properties and spatial distribu-
tions of the temperature data in the East African region
FIG. 3.
Monthly mean, maximum, and minimum HadISST
sea surface temperatures in East Africa during the 1982­1983,
show both latitudinal and longitudinal variation (Figs. 4­
1987­1988, and 1997­1998 ENSO events. Data are pooled
7). The mean, SD, and median SSTs showed similar
means of all cells.

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
511
FIG. 4.
Spatial distributions in (A, B) mean and (C, D) standard deviation of HadISST sea surface temperatures in East Africa
for the 1951­2002 period. Non-ENSO and ENSO years are compared. See Methods section for classification of ENSO and non-
ENSO years.
did not have significant skewness in SST distributions.
Temperature rises and degree-heating months
Both SST skewness and kurtosis increased in a northerly
The rates of temperature rise during ENSO years were
direction (Fig. 5C, D). Patterns in maximum SSTs
higher than non-ENSO years for most of the squares
generally increased from inshore to offshore locations
(Fig. 7A, B). Nearshore and southern sites had the
but had high variations (Fig. 6A).
fastest rises, and the slowest rises were north of
Minimum SSTs increased in a south­north direction
Madagascar and included the Comoros. Between-site
(Fig. 6B). Highest minima occurred in inshore and
variability was higher during non-ENSO years. Differ-
offshore Kenya and Somalia, and lowest values in
ence in mean annual temperatures between ENSO and
inshore locations of Tanzania and Mozambique. Off-
non-ENSO years increased southward and reached
shore sites had higher minimum SSTs than inshore
locations in the southern sector. Mean SST values were
highest in southern Tanzania­northern Mozambique
higher in ENSO than during normal non-ENSO years
while offshore sites of Kenya had the lowest values (Fig.
for all sites (Fig. 4A, B), although the patterns in spatial
7C). Degree-heating months (DHM) in 1998 increased
distribution were similar between the two periods.
in a northerly direction with southern Somalia­northern
Differences in temperature variations (SD) among the
Kenya having the highest values (Fig. 7D). Lowest
different sites between non-ENSO and ENSO years
values were distributed mainly offshore including the
(Fig. 4C, D) were not as marked as the mean values.
Comoros. Southern Tanzanian sites near Mafia, north-
Many of the SST variables had significant correlations
ern Mozambique, and south off the Comoros also
with each other (Table 2). The kurtosis and skewness
experienced low DHM.
were strongly correlated with the mean, median, and
The relationship between the DHM and the different
SD. The last three had significant positive correlations
temperature statistics are worth noting as the degree-
with each other.
heating weeks (DHW) is often used for predicting coral

---

512
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
FIG. 5.
Spatial distributions in (A) median, (B) mode, (C) kurtosis, and (D) skewness of HadISST sea surface temperatures in
the grids for the 1951­2002 period in East Africa.
bleaching (data available online).10 From simple regres-
gradient in the relationship between longitude and
sion models, mean (R2 ¼ 0.84), standard deviation (R2 ¼
DHM or longitude and SD (P . 0.05).
0.80) and kurtosis (R2 ¼ 0.86) gave very high fit with
Change in coral cover following the 1998 bleaching event
heating (Table 3). From stepwise multiple regressions,
the mean, median, kurtosis, and maximum SST ex-
Coral cover in East Africa appears to be affected by
plained most of the variation in DHM (R2 ¼ 0.89; F
latitude and, in Kenya, management of fishing. All sites
79,1 ¼
167.3; P , 0.0001). Mean SST had a negative
in Mafia, one site on Pemba, and two sites in Comoros
relationship with DHM, while the remaining variables
had the highest coral cover. Unfished Kenyan reefs
all had positive relationships. DHM was mostly
except Kisite had higher cover than fished Kenyan reefs.
influenced by the variability in kurtosis SST (81.9%;
Most reefs in Tanga, on Zanzibar and Pemba had
Table 4). The remaining ;18% was explained by the
intermediate but higher cover than unfished Kenyan
median (8.9%), mean (6.3%), and maximum (2.9%).
reefs. Except one site in Tanzania (Songo Songo), all
DHM98 linearly decreased with increase in SST SD
reefs surveyed were affected by the bleaching as
(Fig. 8). It showed a very high latitudinal gradient in its
indicated by the significant decreases in cover (Fig.
distribution, the largest gradient occurs north of 108 S
10). All sites in Songo Songo, Chumbe and Bawe in
between Tanzania and Kenya (Fig. 9A). Related to the
Zanzibar, Upangu and Kitanga in Tanga, and the two
strong negative relationship between DHMs and SD, the
sites in Comoros (Itsamia, Mitsamuli) had maintained
latter has a very strong increase trend toward the south
relatively higher coral cover (30­45%) after bleaching in
(R2 ¼ 0.82; P , 0.0001; Fig. 9B). There was no clear
1998. Tutia (Mafia Island) suffered the highest loss
(67%), and only 13% remained in 1999. Other sites that
10 hhttp://www.osdpd.noaa.gov/PSB/EPS/SSTi
suffered high losses of .50% were Mange (Mafia),

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
513
FIG. 6.
Spatial distributions in (A) maximum and (B) minimum HadISST sea surface temperatures in the grids for the 1951­
2002 period in East Africa.
Misali (Pemba), and Taa (Tanga). Kanamai was the
change in cover: fished reefs 0.18 6 0.16; reefs closed to
least affected reef from Kenyan sites (0.53%), and the
fishing 0.38 6 0.21; F76,4 ¼ 8.43; P ¼ 0.008). However,
northern parks at Malindi (33%) and Watamu (29%)
the Kisite Park on the Kenyan­Tanzanian boundary
suffered most. Generally, reefs that were heavily fished
had lower coral cover before 1998 (22.3%) and was less
were least affected in both Kenya and Tanzania (relative
affected than the northern parks (12.3% loss), despite a
change in cover: fished reefs 0.43 6 0.29; reefs closed to
comparable level of management protection to that of
fishing 0.66 6 0.23; F76,4 ¼ 4.87; P ¼ 0.037; absolute
Malindi and Watamu parks.
FIG. 7.
Spatial distributions in mean sea surface temperature (SST) rises during (A) ENSO and (B) non-ENSO years, (C) mean
differences between ENSO and non-ENSO events, and (D) degree-heating months in 1998 (DHM98) in the grids for the 1951­2002
period in East Africa. See the Methods section for classification of ENSO and non-ENSO years.

---

514
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
TABLE 2.
Spearman's correlations among the different SST statistics.
Mean
Median
SD of SST
CV of SST
Skewness
Kurtosis
Minimum
Median
0.97 (,0.0001)
SD
0.87 (,0.0001)
0.85 (,0.0001)
CV
0.87 (,0.0001)
0.84 (,0.0001)
1.0 (,0.0001)
Skewness
À0.73 (,0.0001) À0.82 (,0.0001) À0.58 (,0.0001) À0.58 (,0.0001)
Kurtosis
À0.75 (,0.0001) À0.73 (,0.0001) À0.96 (,0.0001) À0.95 (,0.0001)
0.58 (,0.0001)
Minimum
0.090 (0.40)
À0.25 (0.06)
À0.26 (0.05)
À0.48 (0.0001)
0.50 (,0.0001) À0.33 (0.01)
Maximum
0.60 (,0.0001)
0.48 (0.0001)
0.48 (0.0001)
0.26 (0.05)
À0.38 (0.003)
0.23 (0.08) 0.03 (0.85)
Note: P values are in parentheses.
Coral mortality following the 1998 bleaching event
model regression analysis on absolute change showed
showed a strong exponential relationship with coral
significant effects of cover before 1998 and in situ
cover (Fig. 10), with reefs that had high initial coral
temperature variation (R2 ¼ 0.97; F ¼ 37.9; P , 0.0001)
cover suffering the most. Change in cover was also
with no significant interaction (P . 0.05). The different
significantly related to cover of Acropora (R2 ¼ 0.40; P ¼
SST statistics and DHM from the NOAA and HadISST
0.006) and non-Acropora (R2 ¼ 0.28; P ¼ 0.025). The
data set did not give significant correlations with coral
change across 1998 could not be largely attributable to
cover, mainly due to the low spatial resolution of the
Acropora dominance at sites before 1998 as the change
temperature data.
was nearly as strongly predicted by non-Acropora (R2 ¼
Bleaching response in 2005
0.62; P , 0.0001) as Acropora cover (R2 ¼ 0.76; P ,
0.0001). Sites that had high initial coral cover of ;60%
The bleaching response in 2005 was highest in Malindi
and were mostly from Tanzania. Below this level of
2 (17.74%), Watamu (18.12%), Mombasa (Coral Gar-
cover, the mortality­cover relationship was highly
den, 13.59%), and lowest in Kanamai (0.85%; Fig. 12).
variable. Generally, Kenyan reefs had higher change in
Mean temperature did not show a significant effect on
cover or mortality than Tanzanian and Comorian reefs
bleaching in 2005 (P . 0.05), but the relationship
(F
between bleaching and temperature variation (SD; Fig.
cntry ¼ 9.4; P ¼ 0.005). The effects of cover before 1998
did not differ between Tanzania and Kenya. Tanzanian
12A) was negative, while the relationship with water
reefs of Songo Songo, Tanga, and Zanzibar, the shallow
flow was positive (Fig. 12B). Temperature variation and
fished reef at Kanamai, Kenya, and the two reefs from
water flow were strongly correlated (r ¼ À0.85; P ¼
Comoros suffered the least (P , 0.05). The unfished
0.008). Stepwise multiple regression analysis showed
reefs of Kenya and Tanzanian reefs at Misali had higher
significant effects only for water flow; temperature
relative coral mortality (P , 0.05). Unfished Kenyan
variation, mean temperature, and their interactions with
sites suffered higher mortalities than their Zanzibar
water flow were removed from the model.
counterparts despite the lower cover they had before the
DISCUSSION
bleaching episode.
Both absolute and relative changes in coral cover
Temperature rise and ENSO events in East Africa
showed negative relationship with temperature variation
The level of rise in sea surface temperature in East
(SD and CV) obtained from in situ measurements on
Africa (Fig. 2) is in general agreement with other
Kenyan reefs (P , 0.01; Fig. 11). Results of mixed
findings but indicates considerable spatial variation in
TABLE 3.
Summary of simple regression analysis on the effects of the different SST statistics on degree-heating months (DHM98).
Predictor
R2
F79,1
P(F)
Term
t
Prob . jtj
Equation
Average
0.47
70.21
,0.0001
intercept
8.76
,0.0001
y ¼ 101.50 À 3.56x
x
À8.38
,0.0001
SD
0.80
322.74
,0.0001
intercept
30.82
,0.0001
y ¼ 10.47 À 4.44x
x
À8.38
,0.0001
Minimum
0.07
5.94
0.02
intercept
À2.08
0.04
y ¼ À25.89 þ 1.24x
x
2.44
0.02
Maximum
0.08
4.38
0.016
intercept
À0.07
0.94
y ¼ À1.18 À 0.19x À 7.39x2
x
À0.35
0.73
x2
À2.82
0.006
Median
0.56
99.78
,0.0001
intercept
10.80
,0.0001
y ¼ 12.89 À 0.30x
x
À9.99
,0.0001
Kurtosis
0.86
488.04
,0.0001
intercept
60.08
,0.0001
y ¼ 6.82 þ 2.18x
x
22.09
,0.0001
Skewness
0.60
117.51
,0.0001
intercept
72.68
,0.0001
y ¼ 4.88 þ 3.31x
x
10.84
,0.0001
Notes: The best-fit model between a linear and second-degree polynomial is presented for each variable. Direction of the
relationship is indicated by the t ratio.

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
515
TABLE 4.
Summary results of mixed model stepwise regression
(Enfield 2001, Glynn et al. 2001), can not be used as
analysis on degree-heating months in 1998 (DHM98).
evidence for a higher SST because corals undergo some
acclimatization and adaptation to bleaching (Coles and
Predictor
Estimate
t
F76,4
P
Brown 2003, Baker et al. 2004, Berkelmans and van
Intercept
24.44
2.31
Oppen 2006, McClanahan et al. 2007a). In some studied
Mean SST
À4.01
À4.09
16.76
,0.0001
Eastern Pacific reefs, coral bleaching was higher in
Median SST
2.58
4.87
23.68
,0.0001
Kurtosis SST
2.83
14.79
218.83
,0.0001
1982­1983 even though SST anomalies were lower or the
Maximum SST
0.70
2.80
7.82
,0.007
same as in 1997­1998 (Glynn et al. 2001). Evidence for
Notes: Only significant values are presented. Direction of the
significant bleaching and mortality in the western Indian
relationship is indicated by the t ratio. R2 ¼ 0.89; F ¼ 167.25,
Ocean in 1982­1983 is scarce but may be due to a lack of
df ¼ 76, 4; P , 0.0001.
investigation. The 1997­1998 event had both ENSO and
IOD components (Saji et al. 1999), and from this and
the responses. Global mean SSTs have increased by 0.4­
other analyses (Sheppard 2003), was the most geograph-
0.88C over the last ;100 years (Pittock 1999, McCarthy
ically widespread and most severe anomaly in the Indian
et al. 2001). Long-term temperature records for the
Ocean in recorded history.
Indian Ocean from coral cores taken in the Seychelles,
Kenya, Madagascar, and Western Australia indicate a
Change in coral cover
0.8­1.48C rise over the past ;200 years, with much of
The effect of the 1998 bleaching event is significantly
that rise reported since the mid 1970s (Charles et al.
related to coral cover, sites of high coral cover suffering
1997, Kuhnert et al. 1999, Cole et al. 2000, Zinke et al.
the most. These are sites that mostly had high cover of
2004, 2005). The Seychelles, Kenyan, and Tanzanian
branching and encrusting forms, which have thin tissue,
coral core records indicate cycles that average ;5.5
high growths rates, and are highly susceptible to
years but vary from 2­8 years and are contained in a
bleaching (Gates and Edmunds 1999, Marshall and
longer cycle of about 8­14 years (Charles et al. 1997,
Baird 2000, Loya et al. 2001, McClanahan 2004,
Cole et al. 2000, Damassa et al. 2006).
McClanahan et al. 2007b). In many places during warm
Most of the outliers in the SST rise belong to ENSO,
years, seasonally high water temperatures cause bleach-
IOD, and ENSO-IOD events (Fig. 2). Some ENSO
ing of the susceptible taxa, but recovery follows in a few
events, such as the ones in 1957­1958 and 1991­1992
months. In 1997­1998, even the more resistant forms
(Goreau and Hayes 1994, Hoegh-Guldberg 1999, Saji et
were bleached, but mortality was higher in the fast-
al. 1999, Edwards et al. 2001), were weak and not
growing forms (McClanahan 2004). For example,
distinguished from background variability by our linear
Kenyan reefs in parks had the highest coral cover, the
regression analysis. The 1996 minimum outlier from
most susceptible branching and encrusting taxa, and the
these data was not recorded as an ENSO, IOD, or
highest mortality. Coral cover on Kenyan reefs was
ENSO­IOD event. SST rises in East Africa are higher
reduced to 10­15% cover of mainly bleaching-resistant
during ENSO than non-ENSO years, and ENSO years
massive/sub-massive growth forms in both protected
have higher baselines ( y-intercepts) indicating a higher
fishery closures and unprotected reefs (McClanahan et
SST buildup during those events.
al. 2001). A direct physiological trade-off has been
The three ENSO events of 1982­1983, 1987­1988, and
suggested between fast growth and resistance to
1997­1998 have the highest recorded mean SSTs in East
Africa (Figs. 2 and 3). Based on the response of the
atmospheric climate, it has been suggested that the
magnitude of the 1982­1983 event was higher than or
equal to the 1997­1998 event (Wolter and Timlin 1998).
The fact that South African locations suffered less severe
drought and that the failure of the Asian Monsoon was
less drastic in 1997­1998 than 1983­1984 was used as a
correlate in comparing the two events. Our results do
not support the contention that the 1982­1983 event was
comparable in its strength to 1997­1998. Mean and
maxima SSTs were higher in 1997­1998. The 1997­1998
event persisted longer; the 1982­1983 event was higher
only in the end of the warm season of 1983 when SSTs
had already started dropping below the mean annual
maximum (Fig. 3; Spencer et al. 2000). During the
bleaching season, 1997­1998 was highest followed by the
1987­1988 mean SSTs.
FIG. 8.
The relationship between sea surface temperature
Differences in the intensity of coral bleaching, such as
standard deviation (SST SD) and degree-heating months in
higher bleaching responses in 1982­1983 in some regions
1998 (DHM98) in East Africa.

---

516
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
coral cover and dominant taxa in determining absolute
change across bleaching events. It is necessary to remove
cover effects before examining for the effects of other
environmental variables in analyses that involve spatial
variations and between-site comparisons in change in
cover (Co^te´ et al. 2005).
Temperature variations, degree-heating months,
and coral mortality
More relevant to the scope of this paper is the
considerable spatial variation seen in the distribution of
the different SST statistics in East Africa and how it may
influence ecological communities. The strong negative
relationship between SD and DHM may seem counter-
intuitive, but high SST SDs make it less likely that
anomalies are rare events, and herein may lie the
mechanism that creates the potential for acclimation/
adaptation and persistence of corals through rare events.
Northern sites, which have high DHM and coral
mortality, have slightly lower mean and high skewness
and kurtosis in SSTs. Relative differences in mean
temperatures are considerably smaller compared to
those of SDs and kurtosis (Fig. 4). The stronger effects
of SD and kurtosis on DHM tend to override other
temperature indices and are probably the key measures
for associating with coral communities and the potential
for acclimation/adaptation or susceptibility to bleaching
and mortality (Table 4). Although all East African sites
have negative kurtosis, low DHM sites have generally
lower kurtosis or have platykurtic SST distributions that
FIG. 9.
Latitudinal variation in HadISST standard devia-
also tend to be concave and bimodal (Fig. 5D). High
tion (SST SD) and degree-heating months in 1998 (DHM98) in
DHM sites have their temperatures more narrowly
East Africa.
distributed around the mean, a tendency toward
peakiness, and no evidence for bimodality.
Despite the strong relationship between change in
bleaching-induced mortality (Berkelmans and Wills
1999, Gates and Edmunds 1999, Loya et al. 2001). The
hard coral cover or mortality and original cover,
change in coral cover in East Africa was significantly
Kenyan reefs suffered higher mortality than Tanzanian
correlated with the cover of both Acropora and non-
or Comoros reefs relative to the initial coral cover.
Acropora, indicating that variables other than the
Tanzanian and Comoros sites had higher long-term SD
dominant taxa played a significant role either directly
and lower skewness and kurtosis in SSTs than Kenyan
or by influencing community structure. On many
sites, whereas the differences in mean SSTs were not
Tanzanian reefs, where the overall change in coral cover
strong (respectively 27.43 and 27.53 for Kenya and
was lower, mortality of Acropora was low and the genus
Tanzania for the warmest month, April 1998). Stronger
is a main component of the community structure,
evidence for the influence of temperature variation
especially on reefs protected from fishing (McClanahan
comes from the in situ temperature data on Kenyan
et al. 2007b).
reefs, where the scale of sampling is more accurate and
Despite the moderate fit, the relationship between
appropriate than the large-scale 36 3 36 km NOAA or
absolute change in cover and temperature variation
1 3 18 Hadley cells. The relationship between bleaching
from in situ measurements (SD and CV) was not
and mortality with temperature variation (CV and SD)
significant for Kenyan reefs, while the relative cover
becomes more robust and significant after controlling
showed that mortality was higher in reefs with low
for cover before 1998. Kanamai, the shallow-most site
temperature variance (Fig. 11). There are not a large
with the highest temperature variation, experienced the
number of reefs with long-term in situ data, so these
least bleaching in 2005 and reduction in coral cover in
relationships need confirmation from more study sites,
1998. Kanamai was an outlier when using large-scale
but both plots indicate that reduced mortality did not
data because of the shallowness and height of the reef
occur until the SD was above 1.8 or a CV of .7%. The
and this produced higher temperature variation and low
study underlines the importance of the initial level of
water flow and is a good example of how local- can

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
517
FIG. 10.
The relationship between absolute change in coral cover and cover before 1998. Dashed lines indicate 95% CI of the
regression equation. Absolute change is the difference between cover before 1998 and after the 1998 bleaching mortality. Relative
change is the absolute cover divided by cover before 1998. Note that Cha Kati and Chumbo (MB) are represented by the same
point. The location name is followed by a code that indicates country and region names: KN, Kenya; MA, Mafia, southern-central
Tanzania; MB, Mnazi Bay, southern Tanzania; PM, Pemba, northern Tanzania; SS, Songo Songo, southern-central Tanzania; TN,
Tanga, northern Tanzania; ZN, Zanzibar, northern Tanzania; CM, Comoros.
sometimes override large-scale factors (McClanahan
the taxonomic composition of the corals at a site
and Maina 2003).
(McClanahan et al. 2007a).
Despite the devastating effects of the 1998 bleaching,
Spatial resolution and variability
the patterns in bleaching and mortality within the Indian
Temperature data based on larger areas, such as the
Ocean were variable and contained many exceptions
HadISST and JCOMMSST data sets, and to some
(Goreau et al. 2000, McClanahan et al. 2007b).
extent the NOAA data, are useful for determining global
Exploring these exceptions through analyzing the site-
or regional trends but in many cases are too crude in
specific temperature history and its interactions with
their spatial resolution to correctly predict coral
other local environmental parameters and the suscepti-
bleaching at individual reefs as indicated by the low-fit
bility of the coral community aids understanding on
and nonsignificant relationship between change in cover
how corals cope with rare events (McClanahan et al.
and the HadISST data. NOAA satellite data have a
2005a, b). Analysis of the Western Indian Ocean
spatial resolution that is considerably finer than the
JCOMM­SST data indicates that DHM is significantly
HadISST data but are still not sufficiently resolved to be
predicted by SD/kurtosis, maximum, median, minimum,
useful for some shallow nearshore and lagoonal reefs
and skewness (M. Ateweberhan and T. R. McClanahan,
(Wellington et al. 2001). Temperature anomalies (Wool-
unpublished data). For example, in spite of their higher
dridge and Done 2004) and bleaching (Andrefouet et al.
mean SST of 28.938C, Maldivian reefs experienced the
2002, McClanahan and Maina 2003, McClanahan et al.
highest bleaching mortalities in 1998, with reefs having
2005b) are highly variable at small spatial scales. A
coral cover of ,5% soon after 1998 (McClanahan 2000,
large-scale study in the Indian Ocean indicated that as
Edwards et al. 2001). In contrast, mean JCOMMSST for
much as half of the response to bleaching may be due to
Tanzania is 27.548C. The oceanic Maldives has more

---

518
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
FIG. 11.
Relationships between (A, C) absolute and (B, D) relative change in coral cover and in situ temperature variations
(SD, CV) on Kenyan reefs. Absolute change is the difference between cover before 1998 and after the 1998 bleaching mortality;
relative change is the absolute change divided by cover before 1998.
uniform and more rare warm temperatures (SD SST ¼
Mauritius, sites on Mafia and Misali at Pemba that
0.60, skewness ¼ 0.50, kurtosis ¼ À0.17) than Tanzania
receive direct oceanic waters suffered among the highest
(SD ¼ 1.41, skewness ¼À0.19, kurtosis ¼À0.17). The SD
mortalities (Fig. 10; Mohammed et al. 2000). Glynn et
in East Africa is, in fact, 49­123% higher than the
al. (2001) also reported the highest mortality of corals
highest variation present in the entire Maldives
across 1998 in the offshore islands of the eastern Pacific.
(HadISST data). The more positive skewness of
Lower DHWs/DHMs in Tanzania have been ex-
Maldivian temperatures reflects the occasional high
plained in terms of cooling effects of oceanic waters,
DHM. In comparison, only a few squares (n ¼ 9
high cloud cover and rainfall, and turbidity from river
squares; 17%) had significant positive skewness in East
runoff (Muhando and Mohammed 2002, Obura 2005).
Africa and all of the East African cells with positive
Although other factors may have interacted with
skewness had the highest DHM (5.4­6.0).
temperature, we hypothesize that the low DHM/DHW
In contrast, reefs of Mauritius have some of the
in Mauritius and East Africa resulted mainly from the
highest SD SSTs in the southern tropical Indian Ocean
effect of the high temperature variation. It is possible
(1.70­1.82) and were among reefs least affected by the
that some features, related to local hydrography and
1998 bleaching (McClanahan et al. 2005b). The temper-
monsoon-related seasonality, localized upwelling, and
ature history of the sites has been overlooked and hardly
protection could have interacted with temperature
reported, and the low bleaching in 1998 has been
(Glynn 2000, Riegl and Piller 2003, Manzello et al.
attributed to the cooling effect of Cyclone Anecelle, high
2007). Nonetheless, because temperature is the principal
cloud cover, and rainfall (Turner et al. 2000, Obura
factor affecting coral bleaching and mortality (Berkel-
2005). Windward and offshore sites are influenced most
mans and Willis 1999), anecdotal explanations of other
by oceanic effects, but they were reported as the most
effects and interactions should be considered carefully or
bleached (Turner et al. 2000). A study of the 2004
investigated after controlling for the effects of temper-
bleaching event in Mauritius indicated site-specific
ature. Furthermore, prioritizing and planning for
difference in bleaching that correlated well with tem-
management requires permanent features such as island
perature variation and water flow (McClanahan et al.
effects as opposed to what may largely be stochastic
2005b). Bleaching correlated negatively with tempera-
factors such as spatial and temporal changes in weather
ture variation (SD) but positively with water flow, and
or light (Gill et al. 2006).
increased from the leeward to the windward side of the
Given the small spatial scales at which temperature
island as might be expected from temperature and water
anomalies and bleaching operate, measurements of
flow histories around islands. Similar to that in
other environmental data that are expected to alleviate

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
519
anomalous temperatures effects are best matched by
similar spatial details as opposed to meteorological
recordings in one or few stations or that average
conditions over large areas. Similarly, there needs to
be a matching of the temporal scale of environmental
variables with the temporal scale of bleaching. The 1998
ENSO has been well identified for its steady warming
over many months and bleaching is often a result of
persistent warm water where DHW usually exceeds four
weeks (Liu et al. 2005). It is less likely that short-term
climatic events that pass in a few hours or days will have
a significant effect on warming and bleaching. Cloud
cover, although with exceptions such as the permanence
of monsoons or other persistent weather events, is
usually an impermanent and less reliable factor of
attenuation. Similarly, aerosols have been shown to
reduce bleaching effects but are often controlled by
volcanic eruptions or other stochastic weather patterns
(Gill et al. 2006). Shading by permanent objects, such as
high islands, emergent rocks, and coral heads (West and
Salm 2003), position on a reef in relation to depth (Riegl
and Piller 2003), and coral orientation (Dunne and
Brown 2001) are recognized as permanent and reliable
attenuation factors from high radiation. Additionally,
our study identifies the area from northern Madagascar
to Comoros is an area of localized and weak upwelling,
associated with the departure of the equatorial current
from the island of Madagascar, that has both a low
temperature rise and degree-heating profile that is likely
to survive relatively well with a warming climate. This
FIG. 12.
Relationships between bleaching index and (A)
water flow speed and (B) variation (SD) of in situ seawater
area is one of a number of similar small-scale upwelling
temperature in 2005 on Kenyan reefs.
areas reported to have low bleaching mortality and
expected persistence (Glynn et al. 2001, Riegl and Piller
2003).
(Sheppard 1999, McClanahan and Maina 2003). Con-
sequently, using coral taxa that originate in different
Water flow and coral bleaching
flow­temperature environments and including the loss
The effects of water flow and temperature variation
of flow in temperature anomaly experiments are needed.
on bleaching appear essentially the same but differ in
Bleaching predictions and models
their direction (McClanahan et al. 2005b) and are
difficult to separate (van Woesik et al. 2005). This and
Predictions for the future of coral reefs under climate
a study in Mauritius indicate that water flow may be the
change scenarios are based on the current and predicted
dominant influence, but because both studies are based
rate of SST, thresholds for bleaching, and acclimation
on correlations, this conclusion provokes the need for
rates (Hoegh-Guldberg 1999, Hughes et al. 2003,
experimental studies that can tease apart these factors in
Sheppard 2003). Projections in the Indian Ocean, based
the context of historical and genetic influences of the
on pooling temperature data into large regions such as
experimental taxa. A number of studies have shown that
East Africa and the IPCC model of temperature
water flow has a positive effect on coral survival and
projections, indicate that reefs will become susceptible
recovery during warm water events in laboratory
to 1998-like conditions every five years by ;2030 and
experimental studies (Nakamura and van Woesik 2001,
sooner in many latitudes (Sheppard 2003). Most models
Nakamura et al. 2003, Nakamura and Yamasaki 2005),
of temperature rise and bleaching predict a linear or even
but discrepancies with field observation indicate a need
accelerating rise of 2­48C in temperature to the year 2100
to reevaluate the mechanisms of bleaching and survival
(ECHAM4/OPYC3 [Hoegh-Guldberg 1999]; IPCC mod-
(McClanahan et al. 2005a). Warm water anomalies are
els [McCarthy et al. 2001]; HADCM3 and IS92a
often associated with reduced water flow and it may be
[Sheppard 2003]). Saturation of temperatures is not
that corals that originate from high water flow
predicted for most models until after 2100, where this
environments are less acclimated/adapted to the loss of
decelerating rise is expected from an oceanic thermostat
flow that can occur during warm-water anomalies, and
feedback mechanism, such as increased ocean mixing,
therefore more susceptible to bleaching and mortality
clouds, or evaporative cooling (Li et al. 2000, Loschnigg

---

520
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
and Webster 2000). If we use the data presented here and
expect is due to a combination of acclimation and local
assume a linear increase for East Africa, at the current
community or genetic adaptation. Understanding the
rise of mean SST in the warmest month (April; rise of
importance of each of these factors will require further
0.018C/yr), the 1998 mean temperature of 308C would be
study. Temperature rise and variation are likely to
reached near 2077. This study suggests, however, that
increase together in coastal and leeward island sites, as
there is significant spatial heterogeneity and nonlinearity
the retention of water on coasts and behind islands is the
in temperature rise, thresholds, acclimation, and adap-
likely cause of this positive covariance. Models that
tation, and that pooling data and predictions for large
assume higher baselines and rises will certainly increase
areas will miss much of the spatial dynamics that may
the predicted probability of recurrences of the anoma-
result in considerable deviation from predictions based
lous events, but mortality will depend on the acclimation
on pooling.
and adaptation rates relative to rates of rise (Hughes et
Global patterns in warming are heterogeneous in
al. 2003). Acclimation rates are likely to be slowest in
space and time (Casey and Cornillon 2001, McCarthy et
low SST SD areas (Sheppard 1999, McClanahan and
al. 2001, Barton and Casey 2005). Detailed analysis of
Maina 2003), and these are areas with the lowest
each Hadley Cell in East Africa reveals significant
temperature rise, probably associated with equatorial
between-site variations that roughly fall into three
current downwelling on continental margins, but the
categories (Fig. 13). First, sites of highest DHM are
highest mortality.
well explained by a decreasing second-order polynomial
An alternative model is that coral acclimation only
function for SST with time (Fig. 13, cluster 1). For the
occurs on short time scales of less than a month (Coles
span of the HadISST data set used, the highest rise in
and Brown 2003), upper thresholds to coral bleaching
SST took place between the 1950s and mid 1970s,
and mortality have little elasticity between sites and
stability or decline started in the early 1980s, rose again
habitats (Berkelmans and Willis 1999, D'Croz et al.
from 1993­2003, and there was a notably large decline
2001), and this leads to little acclimation/adaptations or
between 2003 and 2005 (Lyman et al. 2006). The small,
tolerance to environmental variation at the scale of
improved difference in fit between the linear and second
months to years. If this second model were true, one
degree polynomial fit is mainly caused by the earlier
would expect persistence of corals only in areas that
marked rise relative to the later decrease. Similarly,
maintain temperatures below these rigid mortality
recent analyses of satellite SST and in situ seawater
thresholds, such as 318C (Coles and Brown 2003).
temperatures indicate that the overall trend in SSTs in
Determining the correct model of coral acclimation/
many tropical areas has been stable or has fallen during
adaptation is critical to predictions and establishing
the last two decades (Barton and Casey 2005). Second,
management policies that prioritize areas based on a
sites of lowest DHM do not show a strong rise in SST
triage approach. Both models do, however, predict that
(Fig. 13, cluster 2). Third, as SST rise increases above
areas that maintain stable cool water without rare
0.018C/year in the southern part of this region, the
anomalies will survive climate change and are a
relationship with DHM becomes flat (Fig. 13, cluster 3).
conservation priority (Glynn 2000, Riegl and Piller
Sites in the northern region, which have moderate
2003, West and Salm 2003).
temperature rise and high but variable DHM, are those
From this spatial analysis, we demonstrated that
sites most likely to experience the dual effect of the
temperature variation (SD and kurtosis) is the best
predicted temperature rise and anomalous temperatures,
predictor of warming and the associated bleaching
which is predicted to lead to mortality and possibly
related mortality in East Africa. Water temperature
extinction of corals. Pooling data into regions and
history of a reef and the time corals spend at these
assuming mortality is proportional to the rise or when a
temperatures determine their threshold for bleaching
threshold is reached will fail to detect the uniqueness of
and survival (Coles and Brown 2003, McClanahan and
the categories displayed here, and this is also expected
Maina 2003, West and Salm 2003). The effect of
for larger scale global analyses.
temperature variability on the rise of DHW/DHM,
Temperature patterns in East Africa are highly
and thus bleaching, is insufficiently addressed in coral
variable, depending on the location and temperature
research. Most studies focus on the effects of the mean
history of the sites, and predictions generally do not
and maximum threshold temperatures, and the variance
concur and may even contradict current models. The
is too often ignored. This has significant implications to
relationship between ocean warming and DHM appears
our capacity to correctly predict responses of biological
to be strongly spatially dependent and therefore
communities. The mean and maximum temperatures
nonlinear such that baseline temperatures and rises
may be important in larger spatial comparisons but have
cannot be directly translated into thresholds and DHM,
low or no significant predictive power at site-specific or
the most likely cause of coral mortality. Assumptions of
local scales (Table 4).
linear or exponential SST increases and coral mortality
are likely to produce poor predictions, as our data
Diversity and acclimation/adaptation to bleaching
suggest that the areas with the highest rise also have the
With the increasing threat to coral reefs from rising
highest SST SD and low relative mortality, which we
temperatures and other interacting factors, there has

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
521
FIG. 13.
Summary of the relationship between sea surface temperature (SST) rise and degree-heating months (DHM) in East
Africa. Three main clusters are indicated; colors match those in the map (inset). Line widths and arrow directions represent relative
strengths in model fit.
been a growing interest in identifying reefs that maintain
publications is now 226 species, and Tanzanian reefs
high coral cover, biodiversity, and ecological function-
have relatively higher cover and species diversity (211
ing. This concept of resilience, addressing the capacity of
species) than their Kenyan counterparts (163 species;
ecosystems to recover and regenerate following major
Hamilton and Brakel 1984, Lemmens 1993, Johnstone et
ecological disturbances, is increasingly becoming a main
al. 1998). Tanzanian sites, especially those in Songo
focus in ecological and resource management research
Songo near the southern end of the geographic range of
(Hooper et al. 2005, Hughes et al. 2005). Although
this study, have escaped the main effect of the 1998
sources of resilience are expected to be complex and
bleaching and still maintain substantial cover of
operate at multiple scales, biomass/cover and local and
Acropora and other branching and encrusting taxa that
regional diversity have widely been recognized as the two
are thought to be less tolerant of warm water and have
main measurable components of resilience in coral reef
become rare in Indian Ocean reefs affected by bleaching
ecology (McClanahan et al. 2002, Salm and West 2003,
(Obura 2005; McClanahan et al. 2007b).
Bellwood et al. 2004, Hughes et al. 2005, Obura 2005).
The environmental gradient and the resulting differ-
This regional study and the bleaching survey in 2005
ential bleaching in East Africa is most likely created by
indicate that many Tanzanian reefs show high resistance
the local hydrography linked to the position of the
to bleaching, fast recovery after bleaching, and maintain
Island of Madagascar and, to a lesser extent, the other
high coral cover and diversity. The bleaching study in
smaller islands along the Tanzanian coast, most notably
2005 also allowed analysis of community structure, and
Songo Songo, Mafia, Zanzibar, and Pemba. These
Tanzanian reefs were found to be some of the most
islands reduce the influence of the Equatorial and East
diverse in the Indian Ocean, with a comparable generic
African Currents and result in higher water retention
diversity to that of the Maldives (McClanahan et al.
time and high temperature variation (Figs. 1 and 7).
2007b). This observation is also in agreement with
Coral and other taxa could be influenced in two
species diversity patterns for East Africa (McClanahan
interacting ways. First, the low to intermediate DHW
1990). The hard coral species record from three
reduce severe impacts of extreme conditions that might

---

522
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
otherwise far exceed normal variation. Secondly, higher
reefs are not at their most productive or diverse can
warming resulting from semi-isolation could create a
provide important information on coral reef resistance,
selective pressure for taxa tolerant of warm water.
but more important to management of coral reef
Community change and genetically based adaptations
diversity is how resilience evolves in reefs that maintain
might evolve as spatial variations in abiotic conditions
high coral cover, high diversity, and ecological func-
impose divergent selective mechanisms and promote
tioning. Tanzanian and Comorian (and probably
genetically based differentiation of zooxanthellae, host
northern Madagascar) reefs appear to provide these
corals, or their symbiosis (Baker et al. 2004, Buddemeier
conditions as they have maintained high cover and
et al. 2004, Berkelmans and van Oppen 2006). A thermal
species diversity over a period of extreme disturbance,
adaptation that operates at a regional scale of hundreds
and are, therefore, a high priority for future research
of kilometers as indicated for the Great Barrier Reef
and management.
(Berkelmans 2002) could be inferred from the patterns
ACKNOWLEDGMENTS
observed in East Africa. The east African region,
therefore, provides a unique environmental gradient to
The Wildlife Conservation Society, through grants from
Western Indian Ocean Marine Science Association (WIOM-
examine the relationships between environmental vari-
SA), Marine Science for Management Program (MASMA),
ation, climate change, and adaptation, and its conse-
Leverhulme Trust, Eppley and Tiffany Foundations, and
quences for biodiversity and ecological functions.
World Bank Targeted Research Group on Coral Bleaching
supported this research. The Institute of Marine Sciences of the
CONCLUSIONS
University of Dar es Salaam and CORDIO East Africa
supported the work in Tanzania. We thank the U.K. Met
This investigation of the East Africa coastline reveals
Office Hadley Centre for use of their global sea ice and SST
distinct regional variation in SST rise, variation and
(HadISST1.1) data set and are grateful to A. Edwards for
DHM, and response to the 1997­1998 bleaching event.
assistance with extracting the data for the area of interest.
This regional variation largely falls into three major
LITERATURE CITED
categories which are expected to respond differentially
Andrefouet, S., R. Berkelmans, L. Odriozola, T. Done, J.
to climate change and provide a basis for more
Oliver, and F. Muller-Karger. 2002. Choosing the appropri-
comprehensive analyses of climate change predictions
ate spatial resolution for monitoring coral bleaching events
in other regions and on a global basis. Our data
using remote sensing. Coral Reefs 21:147­154.
underline the need to evaluate the appropriate scale at
Anzai, R. 2001. The effects of coral morphology and water flow
rates on coral growth and passive diffusion. Thesis.
which temperature operates in creating different accli-
University of Ryukyus, Okinawa, Japan.
mation/adaptation environments. Temperature base-
Aronson, R. B., W. F. Precht, M. A. Toscano, and K. H.
lines, rises, and thresholds are unlikely to predict the
Koltes. 2002. The 1998 bleaching event and its aftermath on a
future of coral reefs unless they are based on more
coral reef in Belize. Marine Biology 144:435­447.
information about the temperature histories and accli-
Baker, A. C., C. J. Starger, T. R. McClanahan, and P. W.
Glynn. 2004. Corals' adaptive response to climate change.
mation/adaptation potential of corals within various
Nature 430:741.
thermal environments. The response of marine organ-
Barton, A. D., and K. S. Casey. 2005. Climatological context
isms to temperature variation can be complex and
for large-scale bleaching. Coral Reefs 24:536­554.
greatly depend on their history of exposure and ability
Bellwood, D. R., T. P. Hughes, C. Folke, and M. Nystrom.
2004. Confronting the coral reef crisis. Nature 429:827­832.
to acclimate and adapt to changes (Berkelmans 2002,
Benny, P. N. 2002. Variability of western Indian Ocean
Coles and Brown 2003, Baker et al. 2004, Buddemeier et
currents. Western Indian Ocean Journal of Marine Science
al. 2004). A view of coral reef ecosystems that em-
1:81­90.
phasizes regional and historical variability and acclima-
Berkelmans, R. 2002. Time-integrated thermal bleaching
tion/adaptation to various environments (Lough 1994,
thresholds of reefs and their variation on the Great Barrier
Reef. Marine Ecology Progress Series 299:73­82.
Brown 1997) is likely to be more accurate than one that
Berkelmans, R., G. De'ath, S. Kininmonth, and W. Skirving.
sees them as characterized by stable and benign
2004. A comparison of the 1998 and 2002 coral bleaching
temperature regimes close to their upper thresholds.
events on the Great Barrier Reef: spatial correlation and
This change in focus is likely to hold the key to viewing
predictions. Coral Reefs 23:74­83.
Berkelmans, R., and M. J. H. van Oppen. 2006. The role of
the persistence of reefs under future climate change
zooxanthellae in the thermal tolerance of corals: a ``nugget of
predictions and increase the optimism needed to value
hope'' for coral reefs in an era of climate change. Proceedings
further research and management efforts on coral reefs.
of the Royal Society B 273:2305­2312.
Corals acclimatized and adapted to thermal stress
Berkelmans, R., and B. L. Willis. 1999. Seasonal and local
show more tolerance to elevated temperatures and
spatial patterns in the upper thermal limits of corals on the
inshore Central Great Barrier Reef. Coral Reefs 18:219­
radiation and bleach less (Goreau et al. 1992, Brown
228.
et al. 2000, 2002, West and Salm 2003) but often at the
Brown, B. E. 1997. Coral bleaching: causes and consequences.
expense of losing taxa with low tolerance to environ-
Coral Reefs 16:S129­S138.
mental fluctuations (McClanahan and Maina 2003).
Brown, B. E., R. P. Dunne, M. S. Goodson, and A. E. Douglas.
2000. Bleaching patterns in reef corals. Nature 404:142­143.
With the need to understand coral reef resilience, recent
Brown, B. E., R. P. Dunne, M. S. Goodson, and A. E. Douglas.
research has focused on these stressed or marginal reefs
2002. Experience shapes the susceptibility of a reef coral to
(Perry and Larcombe 2003). Stressed locations where
bleaching. Coral Reefs 21:119­126.

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
523
Buddemeier, R. W., A. C. Baker, D. G. Fautin, and J. R.
during the 1997­98 El Nino­Southern Oscillation event:
Jacobs. 2004. The adaptive bleaching hypothesis. Pages 427­
spatial/temporal patterns and comparisons with the 1982­
444 in E. Rosenberg and Y. Loya, editors. Coral health and
1983 event. Bulletin of Marine Science 69:79­109.
disease. Springer, Berlin, Germany.
Goreau, T. J., and R. L. Hayes. 1994. Coral bleaching and
Casey, K. S., and P. Cornillon. 2001. Global and regional
ocean ``hot spots.'' Ambio 23:176­180.
seasurface temperature trends. Journal of Climatology 14:
Goreau, T., T. McClanahan, R. Hayes, and A. Strong. 2000.
3801­3818.
Conservation of coral reefs after the 1998 global bleaching
Charles, C. D., D. E. Hunter, and R. D. Fairbanks. 1997.
event. Conservation Biology 14:5­15.
Interaction between the ENSO and the Asian Monsoon in a
Guinotte, J. M., R. W. Buddemeier, and J. A. Kleypas. 2003.
coral record of tropical climate. Science 277:925­928.
Future coral reef habitat marginality: temporal and spatial
Cole, J., R. Dunbar, T. McClanahan, and N. Muthiga. 2000.
effects of climate change in the Pacific basin. Coral Reefs 22:
Tropical Pacific forcing of decadal variability in SST in the
551­558.
western Indian Ocean. Science 287:617­619.
Hamilton, H. G. H., and W. H. Brakel. 1984. Structure and
Coles, S. L., and B. E. Brown. 2003. Coral bleaching--capacity
coral fauna of East African reefs. Bulletin of Marine Science
for acclimatization and adaptation. Advances in Marine
34:248­266.
Biology 46:183­223.
Hoegh-Guldberg, O. 1999. Climate change, coral bleaching and
Coles, S. L., P. L. Jokiel, and C. R. Lewis. 1976. Thermal
the future of the world's coral reefs. Marine and Freshwater
tolerance in tropical versus subtropical reef corals. Pacific
Research 50:839­866.
Science 30:159­166.
Hooper, D. U., et al. 2005. Effects of biodiversity on ecosystem
Co^te´, I. M., J. A. Gill, T. A. Gardner, and A. R. Watkinson.
functioning: a consensus of current knowledge. Ecological
2005. Measuring coral reef decline through meta-analysis.
Monographs 75:3­35.
Philosophical Transactions of the Royal Society of Britain,
Hughes, T. P., et al. 2003. Climate change, human impacts, and
London 360:385­395.
the resilience of coral reefs. Science 301:929­933.
Damassa, T. D., J. E. Cole, H. R. Barnett, T. R. Ault, and T. R.
Hughes, T. P., D. R. Bellwood, C. Folke, R. S. Steneck, and J.
McClanahan. 2006. Enhanced multidecadal variability in
Wilson. 2005. New paradigms for supporting the resilience of
the 17th century from coral isotope records in the western
marine ecosystems. Trends in Ecology and Evolution 20:380­
Indian Ocean. Paleoceanography 21:PA20162010.1029/
386.
2005PA001217.
Iglesias-Prieto, R., J. L. Matta, W. A. Robins, and R. K.
D'Croz, L. D., J. L. Mate, and J. E. Oke. 2001. Responses to
Trench. 1992. Photosynthetic response to elevated tempera-
elevated sea water temperature and UV radiation in the coral
ture in the symbiotic dinoflagellate Symbiodinium micro-
Porites lobata from upwelling and non-upwelling environ-
adriaticum in culture. Proceedings of the National Academy
ments on the Pacific Coast of Panama. Bulletin of Marine
of Sciences (USA) 89:10302­10305.
Science 69:203­214.
Johnstone, R. W., C. A. Muhando, and J. Francis. 1998. The
Doty, M. S. 1971. Measurement of water movement in
status of the coral reefs of Zanzibar: one example of a
reference to benthic algal growth. Botanica Marina 14:32­35.
regional predicament. Ambio 27:700­707.
Dunne, R. P., and B. E. Brown. 2001. The influence of solar
Jokiel, P. L. 2004. Temperature stress and coral bleaching.
radiation on bleaching of shallow water reef corals in the
Pages 401­425 in E. Rosenberg and Y. Loya, editors. Coral
Andaman Sea, 1993­1998. Coral Reefs 20:201­210.
health and disease. Springer, Berlin, Germany.
Edmunds, P. J., R. D. Gates, and D. F. Gleason. 2003. The
Jokiel, P. L., and S. L. Coles. 1977. Effects of temperature on
tissue composition of Montastraea franksi during a natural
the mortality and growth of Hawaiian reef corals. Marine
bleaching event in the Florida Keys. Coral Reef 22:54­62.
Biology 43:201­208.
Edwards, A. J., S. Clark, H. Zahir, A. Rajasuriya, A. Naseer,
Jokiel, P. L., and J. I. Morrissey. 1993. Water motion on coral
and J. Rubens. 2001. Coral bleaching and mortality on
reefs: evaluation of the ``clod card'' technique. Marine
artificial and natural reefs in Maldives in 1998, sea surface
Ecology Progress Series 93:175­181.
temperature anomalies and initial recovery. Marine Pollution
Jones, R. J. 1997. Changes in zooxanthellae densities and
Bulletin 42:7­15.
chlorophyll concentrations in corals during and after a
Enfield, D. B. 2001. Evaluation and historical perspective of the
bleaching event. Marine Ecology Progress Series 158:51­59.
1997­1998 El Nino­Southern Oscillation event. Bulletin of
Jones, R. J., and O. Hoegh-Guldberg. 2001. Diurnal changes in
Marine Science 69:7­25.
the photochemical efficiency of the symbiotic dinoflagellates
Enriquez, S., E. R. Mendez, and R. Iglesias-Prieto. 2005.
(Dinophyceae) of corals: photoprotection, photoinactivation
Multiple scattering on coral skeletons enhances light
and the relationship to coral bleaching. Plant, Cell and
absorption by symbiotic algae. Limnology and Oceanogra-
Environment 24:89­99.
phy 50:1025­1032.
Kayanne, H., H. Iijima, N. Nakamura, T. R. McClanahan, S.
Gates, R. D., and P. J. Edmunds. 1999. The physiological
Behera, and Y. Yamagata. 2006. Indian Ocean Dipole index
mechanisms of acclimatization in tropical reef corals.
recorded in Kenyan coral annual density bands. Geophysical
American Zoological 39:30­43.
Research Letters 33:10­14 [doi: 10.1029/2006GL027168].
Gill, J. A., A. R. Watkinson, J. P. McWilliams, and I. M. Cote.
Kleypas, J. A., R. W. Buddemeier, and J. Guttuso. 2001. The
2006. Opposing forces of aerosol cooling and El Nin~o drive
future of coral reefs in an age of global change. International
coral bleaching on Caribbean reefs. Proceedings of the
Journal of Earth Sciences 90:426­437.
National Academic Sciences (USA) 103:18870­18873.
Knowlton, N., E. Weil, L. A. Weight, and H. M. Guzman.
Gleason, M. G. 1993. Effects of disturbance on coral
1992. Sibling species in Montastraea annularis, coral
communities: bleaching in Moorea, French Polynesia. Coral
bleaching, and the coral climate record. Science 255:330­
Reefs 12:193­201.
334.
Glynn, P. W. 2000. El Nino­Southern Oscillation mass
Kuhnert, H., J. Patzold, B. G. Hatcher, K.-H. Wyrwoll, A.
mortalities of reef corals: a model of high temperature
Eisenhauer, L. B. Collins, Z. R. Zhu, and G. Wefer. 1999. A
marine extinctions? Pages 117­133 in E. Insalaco, P. W.
200-year coral stable oxygen isotope record from a high-
Skelton, and T. J. Palmer, editors. Carbonate platform
latitude reef off Western Australia. Coral Reefs 18:1­12.
systems: components and interactions. Geological Society of
Lemmens, J. W. T. J. 1993. Reef-building corals (Cnidaria:
London London, UK.
Scleractinia) from the Watamu marine national reserve,
Glynn, P. W., J. L. Mate, A. C. Baker, and M. O. Calderon.
Kenya; an annotated species list. Zoologische Mededelingen
2001. Coral bleaching and mortality in Panama and Ecuador
67:453­465.

---

524
TIMOTHY R. M
Ecological Monographs
CCLANAHAN ET AL.
Vol. 77, No. 4
Lesser, M. P., and J. H. Farrell. 2004. Exposure to solar
McClanahan, T. R., J. Maina, R. Moothien Pillay, and A. C.
radiation increases damage to both host tissues and algal
Baker. 2005b. Effects of geography, taxa, water flow, and
symbionts of corals during thermal stress. Coral Reefs 23:
temperature variation on coral bleaching intensity in
367­377.
Mauritius. Marine Ecology Progress Series 298:131­142.
Li, T., T. F. Hogan, and C.-P. Chang. 2000. Dynamic and
McClanahan, T. R., N. A. Muthiga, A. T. Kamukuru, H.
thermodynamic regulation of ocean warming. Journal of
Machano, and R. W. Kiambo. 1999. The effects of marine
Atmospheric Sciences 57:3353­3365.
parks and fishing on coral reefs of northern Tanzania.
Liu, G., A. E. Strong, W. Skirving, and L. F. Arzayus. 2005.
Biological Conservation 89:161­182.
Overview of NOAA coral reef watch program's near-real
McClanahan, T. R., N. A. Muthiga, and S. Mangi. 2001. Coral
time satellite global coral bleaching monitoring activities.
and algal changes after the 1998 coral bleaching: interaction
Pages 1783­1793 in roceedings of the Ninth International
with reef management and herbivores on Kenyan reefs. Coral
Coral Reef Symposium, Okinawa, Japan.
Reefs 19:380­391.
Loschnigg, J. P., and P. Webster. 2000. A coupled ocean-
McClanahan, T. R., N. V. C. Polunin, and T. Done. 2002.
atmospheric system of SST modulation for the Indian Ocean.
Ecological states and the resilience of coral reefs. Conserva-
Journal of Climate 13.
tion Ecology 6:18.
Lough, J. M. 1994. Climate variation and El Nino-Southern
McClanahan, T. R., E. Sala, P. Stickels, B. A. Cokos, A. Baker,
Oscillation events on the Great Barrier Reef: 1958 to 1987.
C. J. Starger, and S. Jones. 2003. Interaction between
Coral Reefs 13:181­195.
nutrients and herbivory in controlling algal communities
Loya, Y., K. Sakai, K. Yamazato, Y. Nakano, H. Sambali, and
and coral condition on Glover's Reef, Belize. Marine
R. van Woesik. 2001. Coral bleaching: the winners and
Ecology Progress Series 261:135­147.
losers. Ecology Letters 4:122­131.
McField, M. D. 1999. Coral response during and after mass
Lyman, J. M., J. K. Willis, and G. C. Johnson. 2006. Recent
bleaching in Belize. Bulletin of Marine Science 64:155­172.
cooling of the upper ocean. Geophysical Research Letters 33:
Mohammed, M. S., C. Muhando, and H. Machano. 2000.
L18604 [doi: 10.1029/2006GL027033].
Assessment of coral reef degradation in Tanzania: results of
Manzello, D. P., M. Brandt, T. B. Smith, D. Lirman, J. C.
coral reef monitoring 1999. Pages 35­42 in D. Souter, D.
Hendee, and R. S. Nemeth. 2007. Hurricanes benefit
Obura, and O. Linden, editors. Coral reef degradation in the
bleached corals. Proceedings of the National Academic
Indian Ocean. CORDIO, Vasteras, Sweden.
Sciences. hwww.pnas.org/cgi/doi/10.1073/pnas.0701194104i
Muhando, C. A., and M. S. Mohammed. 2002. Coral reef
Marshall, P. A., and A. H. Baird. 2000. Bleaching of corals on
benthos and fisheries in Tanzania before and after the 1998
the Great Barrier Reef: differential susceptibilities among
bleaching and mortality event. Western Indian Ocean
taxa. Coral Reefs 19:155­163.
Journal of Marine Science 1:43­52.
McCarthy, J. J., O. F. Canziani, N. A. Leary, D. J. Dokken,
Mumby, P. J., J. R. M. Chisholm, A. J. Edwards, and S.
and K. S. White. 2001. Climate Change 2001: impacts,
Andrefouet. 2001b. Cloudy weather may have saved Society
adaptation, and vulnerability. Contribution of Working
Island reef corals during the 1998 ENSO event. Marine
Group II to the third assessment report of the Intergovern-
Ecological Progress Series 222:209­216.
mental Panel on Climate Change (IPCC). Cambridge
Mumby, P. J., J. R. M. Chisholm, A. J. Edwards, C. D. Clark,
University Press, Cambridge, UK.
E. B. Roark, S. Andrefouet, and J. Jaubert. 2001a.
McClanahan, T. R. 1990. Kenyan coral reef-associated
Unprecedented bleaching-induced mortality in Porites spp.
gastropod assemblages: distribution and diversity patterns.
at Rangiroa Atoll, French Polynesia. Marine Biology 139:
Coral Reefs 9:63­74.
183­189.
McClanahan, T. R. 2000. Bleaching damage and recovery
Nakamura, T., and R. van Woesik. 2001. Differential survival
potential of Maldivian coral reefs. Marine Pollution Bulletin
of corals during the 1998 bleaching event is partially
40:587­597.
explained by water-flow rates and passive diffusion. Marine
McClanahan, T. R. 2002. The near future of coral reefs.
Ecology Progress Series 212:301­304.
Environmental Conservation 29:460­483.
Nakamura, T., and H. Yamasaki. 2005. Requirement of water-
McClanahan, T. R. 2004. The relationship between bleaching
flow for sustainable growth of Pocilloporid corals during
and mortality of common corals. Marine Biology 144:1239­
high temperature periods. Marine Pollution Bulletin 50:
1245.
1115­1120.
McClanahan, T. R., M. Ateweberhan, C. Ruiz Sebastia´n,
Nakamura, T., H. Yamasaki, and R. van Woesik. 2003. Water
N. A. J. Graham, S. K. Wilson, J. H. Bruggemann, and
flow facilitates recovery from bleaching in the coral
M. M. M. Guillaume. 2007a. Predictability of coral bleaching
Stylophora pistillata. Marine Ecology Progress Series 256:
from synoptic satellite and in situ temperature observations.
287­291.
Coral Reefs 10.1007/s00338-006-0193-7.
Obura, D. O. 2005. Resilience and climate change: lessons from
McClanahan, T. R., M. Ateweberhan, C. Ruiz Sebastia´n,
coral reefs and bleaching in Western Indian Ocean. Estuarine
N. A. J. Graham, S. K. Wilson, M. M. M. Guillaume, and
Coastal and Shelf Science 63:353­372.
J. H. Bruggemann. 2007b. Western Indian Ocean coral
Perry, C. T., and P. Larcombe. 2003. Marginal and non-reef
communities: bleaching responses, and susceptibility to
building coral reef environments. Coral Reefs 22:427­432.
extinction. Marine Ecology Progress Series 337:1­13.
Pittock, A. B. 1999. Coral reefs and environmental change:
McClanahan, T. R., A. H. Baird, P. A. Marshall, and M. A.
adaptation to what? American Zoology 39:10­29.
Toscano. 2004. Comparing bleaching and mortality respons-
Podesta, G. P., and P. W. Glynn. 2001. The 1997­98 El Nin~o
es of hard corals between southern Kenya and the Great
event in Panama and Galapagos: an update of thermal stress
Barrier Reef, Australia. Marine Pollution Bulletin 48:327­
indices relative to coral bleaching. Bulletin of Marine Science
335.
69:43­59.
McClanahan, T. R., A. C. Baker, M. Ateweberhan, J. Maina,
Quod, J., and J. Bigot. 2000. Coral bleaching in the Indian
and R. Moothien Pillay. 2005a. Refining coral bleaching
Ocean islands: ecological consequences and recovery in
experiments and models through reiterative field studies.
Madagascar, Comoros, Mayotte and Reunion. Pages 108­
Marine Ecology Progress Series 305:301­303.
113 in D. Souter, D. Obura, and O. Linden, editors. Coral
McClanahan, T. R., and J. Maina. 2003. Response of coral
reef degradation in the Indian Ocean. CORDIO, Vasteras,
assemblages to the interaction between natural temperature
Sweden.
variation and rare warm-water events. Ecosystems 6:551­
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V.
563.
Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan. 2003.

---

November 2007
CLIMATE CHANGE AND CORAL BLEACHING
525
Global analyses of SST, sea ice and night marine air
Van Woesik, R., T. Nakamura, H. Yamasaki, and C. R. C.
temperature since the late nineteenth century. Journal of
Sheppard. 2005. Comment on ``Effects of geography, taxa,
Geophysical Research 108:4407.
water flow, and temperature variation on coral bleaching
Reynolds, R. W., H.-M. Zhang, T. M. Smith, C. L.
intensity in Mauritius.'' Marine Ecology Progress Series 305:
Gentemann, and F. Wentz. 2005. Impact of in situ and
207­209.
additional satellite data on the accuracy of a sea-surface
Warner, M. E., G. C. Chilcoat, F. K. McFarland, and W. K.
temperature analysis for climate. International Journal of
Fitt. 2002. Seasonal fluctuations in the photosynthetic
Climatology 25:857­864.
capacity of photosystem II in symbiotic dinoflagellates in
Riegl, B. 2003. Climate change and coral reefs: different effects
the Caribbean reef-building coral Montastrea. Marine
in two high-latitude areas (Arabian Gulf, South Africa).
Biology 141:31­38.
Coral Reefs 22:433­446.
Webster, P. J., A. M. Moore, J. P. Loschnigg, and R. R. Leben.
Riegl, B., and W. E. Piller. 2003. Possible refugia for reefs in
1999. Coupled ocean­atmosphere dynamics in the Indian
times of environmental stress. International Journal of Earth
Ocean during 1997­98. Nature 401:356­359.
Science 92:520­531.
Wellington, G. M., A. E. Strong, and G. Merlen. 2001. Sea
Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T.
surface temperature variation in the Galapagos Archipelago:
Yamagata. 1999. A dipole mode in the tropical Indian Ocean.
a comparison between AVHRR nighttime satellite data and
Nature 401:360­363.
in situ instrumentation (1982­1998). Bulletin of Marine
Sheppard, C., and R. Rioja-Nieto. 2005. Sea surface temper-
Science 69:27­42.
ature 1887­2099 in 38 cells in the Caribbean region. Marine
West, J. M., and R. V. Salm. 2003. Resistance and resilience to
Environmental Research 60:389­396.
coral bleaching: implications for coral reef conservation and
Sheppard, C. R. C. 1999. Coral decline and weather patterns
management. Conservation Biology 17:956­967.
over 20 years in the Chagos Archipelago, Central Indian
Wolter, K., and M. S. Timlin. 1998. Measuring the strength of
Ocean. Ambio 28:472­478.
ENSO events: How does 1997/98 rank? Weather 53:315­324.
Sheppard, C. R. C. 2003. Predicted recurrences of mass coral
mortality in the Indian Ocean. Nature 425:294­297.
Wooldridge, S., and T. Done. 2004. Learning to predict large-
Siebeck, U. E., N. J. Marshall, A. Kluter, and O. Hoegh
scale coral bleaching from past events: a Bayesian approach
Guldberg. 2006. Monitoring coral bleaching using a colour
using remotely sensed data, in-situ data, and environmental
reference card. Coral Reefs 25:453­460.
proxies. Coral Reefs 23:96­108.
Spencer, T., K. Teleki, C. Bradshaw, and M. C. Spalding. 2000.
Wooldridge, S., T. Done, R. Berkelmans, R. Jones, and P.
Coral bleaching in the southern Seychelles during the 1997­
Marshall. 2005. Precursors for resilience in coral communi-
1998 Indian Ocean warm event. Marine Pollution Bulletin 40:
ties in a warming climate: a belief network approach. Marine
569­586.
Ecology Progress Series 295:157­169.
Tabachnick, B. G., and L. S. Fidell. 1996. Using multivariate
Zinke, J., W. C. Dullo, G. A. Heiss, and A. Eisenhauer. 2004.
statistics. Third edition. Harper Collins, New York, New
ENSO and Indian Ocean subtropical dipole variability is
York, USA.
recorded in a coral record off southwest Madagascar for the
Turner, J., E. Hardman, R. Klaus, I. Fagoone, D. Daby, R.
period 1659 to 1995. Earth and Planetary Science Letters 228:
Baghooli, and S. Persands. 2000. The reefs of Mauritius.
177­194.
Pages 94­107 in D. Souter, D. Obura, and O. Linden,
Zinke, J., M. Pfeigger, O. Timm, W. C. Dullo, and G. R.
editors. Coral reef degradation in the Indian Ocean. Status
Davies. 2005. Atmosphere­ocean dynamics in the Western
Report 2000. CORDIO/SAREC Marine Science, Stockholm,
Indian Ocean recorded in corals. Philosophical Transactions
Sweden.
of the Royal Society A 363:121­142.

---