Coral Reefs

DOI 10.1007/s00338-014-1162-1

REPORT

Forecasting decadal changes in sea surface temperatures

and coral bleaching within a Caribbean coral reef
Angang Li • Matthew A. Reidenbach

Received: 10 August 2013 / Accepted: 29 April 2014

Ó Springer-Verlag Berlin Heidelberg 2014

Abstract Elevated sea surface temperature (SST) caused Keywords Climate change  Coral reefs  Sea surface
by global warming is one of the major threats to coral reefs. temperature  Coral bleaching
While increased SST has been shown to negatively affect
the health of coral reefs by increasing rates of coral
bleaching, how changes to atmospheric heating impact SST Introduction
distributions, modified by local flow environments, has
been less understood. This study aimed to simulate future A primary direct consequence of global warming is
water flow patterns and water surface heating in response increasing sea surface temperature (SST; Bindoff et al.
to increased air temperature within a coral reef system in 2007). From the 1950s to 2009, the mean global SST has
Bocas del Toro, Panama, located within the Caribbean Sea. increased by approximately 0.4 °C (Levitus et al. 2009),
Water flow and SST were modeled using the Delft3D- which has had profound and diverse impacts on marine
FLOWÓ computer simulation package. Locally measured ecosystems (Doney et al. 2012), including coral reefs. Coral
physical parameters, including bathymetry, astronomic reefs provide crucial ecosystem services, including shoreline
tidal forcing, and coral habitat distribution were input into protection (Wilkinson 1996; Bradley et al. 2009), tourism
the model and water flow, and SST was simulated over a benefits (Reaser et al. 2000; Rehr 2012), and shelter for one-
four-month period under present day, as well as projected quarter of all known marine species (Pratchett et al. 2009;
warming scenarios in 2020s, 2050s, and 2080s. Changes in Thur 2010). Projected increases in SST may cause coral reefs
SST, and hence the thermal stress to corals, were quantified to disappear entirely within 20–50 yrs if corals cannot adjust
by degree heating weeks. Results showed that present-day to tolerate this increased heating (Hoegh-Guldberg et al.
reported bleaching sites were consistent with localized 2007). Consequently, it is important to determine how
regions of continuous high SST. Regions with highest SST atmospheric and ocean heating affects the spatial distribu-
were located within shallow coastal sites adjacent to the tion of SST surrounding reef areas. Such patterns in SST
mainland or within the interior of the bay, and character- highlight potential perturbations to coral reef ecosystems,
ized by low currents with high water retention times. Under reveal possible shifts in species richness and abundance, and
projected increases in SSTs, shallow reef areas in low flow suggest vulnerable areas that might be prioritized for marine
regions were found to be hot spots for future bleaching. conservation efforts.
Studies have shown that SST anomalies are the major
cause of mass coral mortality, which is initiated through
coral bleaching (Hughes et al. 2003; Donner 2009). When
SST exceeds the historical mean of local summer maxima
Communicated by Biology Editor Prof. Brian Helmuth by 1–2 °C for 3–4 weeks (Hoegh-Guldberg 1999), coral
bleaching occurs due to the loss of endosymbiotic dino-
A. Li  M. A. Reidenbach (&)
flagellates (Symbiodinium spp.) or their photosynthetic
Department of Environmental Sciences, University of Virginia,
Charlottesville, VA 22904, USA pigments (Glynn 1993). These symbiotic algae fuel 95 %
e-mail: [email protected] of the metabolic activity of their coral hosts (Hoegh-

123
 Coral Reefs

Guldberg et al. 2007; Donner 2009). Even though the 50 km with a lead time of 3 months and has effectively
actual temperature threshold of coral bleaching depends on predicted severe bleaching events (Reynolds and Smith
reef location and coral species (Hughes et al. 2003), 1994; Reynolds et al. 2005). However, these SST predic-
thresholds at most locations range from 29 to 32 °C (Baird tions can fall short in real practice, not only due to their
et al. 2009), and SST anomalies of \1 °C may exceed relatively short period forecasted, but also due to consid-
physiological tolerances (Walther et al. 2002) and cause erable variations in local SST distribution across spatial
large-scale bleaching. While coral bleaching does not scales of \50 km.
necessarily result in coral death, thermal stress typically The main objectives of this study were (1) to examine
leads to reduced growth rate and fecundity, and partial the local water flow patterns and SST under present-day
bleaching tends to make corals more vulnerable to disease scenarios utilizing both in situ observations and a numer-
(Harvell et al. 2002; Bruno et al. 2007; Muller et al. 2008). ical model (Delft3D-FLOWÓ), (2) to simulate heating
Studies have observed that the spatial distribution of SST scenarios in the 2020s, 2050s, and 2080s under predicted
and occurrence of coral bleaching are patchy, both within a global warming, and (3) to quantify local thermal stress on
reef and among reefs (Berkelmans and Oliver 1999; Jokiel coral reefs in comparison with existing coral bleaching
and Brown 2004; Ateweberhan and McClanahan 2010). Such records. We hypothesize that, under projected atmospheric
variations originate from different environmental factors in warming, more shallow regions than offshore areas will
each local region (Glynn 1993). Solar radiation, in the form of reach SST that exceed coral maximum threshold temper-
photosynthetically active radiation (PAR, 400–700 nm) and atures, and it is those shallow regions that have a tendency
ultraviolet radiation (UVR, 290–400 nm; Lesser and Farrell to induce intensive coral bleaching. Our study site is the
2004), can inhibit photosynthesis within corals by causing coral reef ecosystem fringing the Bocas del Toro archi-
photosynthetic organisms to produce an excess flux of toxic pelago, Panama, located in the southwest Caribbean Sea.
reactive oxygen species (Lesser et al. 1990; Lesser 1996), or This site was chosen due to the existing long-term climate
by damaging photosystems and reducing carbon fixation in and oceanographic records as well as coral bleaching
corals exposed to elevated temperature (Lesser 1996; Ferrier- information within the region (Guzmán et al. 2005).
Pages et al. 2007). Restricted water circulation within bays
and lagoons can result in retention of water and localized
heating above offshore values (Jokiel and Brown 2004). Materials and methods
Within coral reefs of the Red Sea, Davis et al. (2011) found
that the greatest diurnal variation in water temperatures Study site
occurred near the center of larger reef flats and on reefs that are
protected from direct wave forcing, while less diurnal tem- The study site (Fig. 1) is the coral reef ecosystem along the
perature variability occurred on smaller knolls or along the Bocas del Toro archipelago. The Bocas del Toro archi-
edge of reef flats. Doldrums, periods of exceptionally warm pelago forms a shallow bay that is almost entirely sur-
water coupled with calm weather (Hoegh-Guldberg 2011), rounded by land, and, due to the restricted nature of water
amplify the warming of the sea surface by reduced vertical circulation, inshore waters have significantly higher means
mixing due to low wind speed and intense heating (Berkel- and ranges of SST compared to offshore waters (Kauffman
mans et al. 2010). Strong ocean winds, on the contrary, reduce and Thomson 2005). As such, the vast coral reefs, which
SST through enhanced evaporative cooling and by mixing inhabit the shallow water surrounding the archipelago
warm surface water with cool deeper water (Manzello et al. islands and along the mainland coast down to 25 m deep,
2007). Within shallow reef areas, solar radiation penetrating may experience substantially different water temperatures
to the benthos may also be strong enough to heat the seafloor from inshore to offshore. Correspondingly, observations of
and affect the thermal environment of benthic organisms coral bleaching (Fig. 1) have been reported in the year
(Wells et al. 2012). 2002 and 2005 at water depths ranging from 0.5 to 9.6 m
While observations and mechanisms of these environ- (ReefBase 2013), with low to medium severity in general
mental impacts on coral reefs have attracted much atten- within the study site.
tion, the prediction of how local SST responds to local Oceanographic and meteorological data were collected
environmental factors, and thus exerts different impacts on at 15-min intervals from a weather station (9°210 02.9600 N;
corals, has seen less focus. Historically, synoptic forecasts 82°150 28.2700 W) located approximately 100 m offshore of
of elevated SST have been used almost exclusively to the Smithsonian Tropical Research Institute (STRI) station
predict coral bleaching (Aronson et al. 2002; Liu et al. in Bocas del Toro, Panama. The physical variables mea-
2006). Specifically, a suite of satellite products, developed sured at the station include incident solar radiation (Qs), air
by NOAA Coral Reef Watch, has been forecasting global temperature (Ta), seawater temperature (Tw), relative air
SST (Peñaflor et al. 2009) at a resolution of approximately humidity (rhum), cloud fraction (Fc), wind speed at 10 m

123
Coral Reefs

Fig. 1 Site map of the Bocas

del Toro archipelago, with the
location of the study area
marked by a hollowed red
square within the inset image.
Within main image, positions of
hobo loggers are represented by
black circles, points of the IPCC
prediction sites (2003) by blue
circles, observational spots of
coral bleaching (ReefBase
2013) by white circles,
distribution of coral reefs
(UNEP-WCMC 2010) by red
patches, marine protected area
(IUCN and UNEP-WCMC
2010) by polygon shadows, and
the Panama base map (ESRI
2010) by green patches.
Bathymetry contours are of
100-m increments starting from
100 m depth

above the sea surface (U10), wind direction, and water level Table 1 List of the four environmental datasets used in this study
(f). These datasets were chosen to represent the local
Code Location Variables Depth of Tw Time interval
environment of the study site, while additional temperature measured measured (m) (min)
sensors (HOBO data-loggers, Fig. 1) were deployed from
2006 to 2010 by STRI staff to measure hourly water tem- STRI STRI station Tw, Qs, etc. 2.0 15
peratures at various locations throughout the region. These COB Isla Colón Tw 4.6 60
temperature loggers provide 12-bit resolution and ±0.2 °C CAB Isla Cayo Agua Tw 5.2 60
accuracy. Temperature loggers were located in Isla Colón ROB Isla Roldan Tw 4.6 60
(COB) at a depth of 4.6 m, Isla Cayo Agua (CAB) at 5.2 m, Water temperature (Tw) data were measured at different locations and
and Isla Roldan (ROB) at 4.6 m (Table 1). In addition, a depths within the Bocas del Toro Archipelago
vertical temperature profile was also recorded every 15 min
at COB at 4.6, 9.1, and 18.3 m from July to October 2010.
Differences in water temperatures between the depths were rate of 1 Hz. The Aquadopp was secured on a frame and
consistently \0.2 °C, which was within the ±0.2 °C accu- deployed on the seafloor at a mean water depth of 1.5 m, at
racy of the temperature loggers, suggesting vertically well- a site approximately 1 km offshore from the STRI station
mixed conditions across this depth range. Previous studies (9°210 7.6200 N; 82°150 55.9200 W). A Richard Branker
have shown that spatial differences in SST are not appre- Research TWR-2050P submersible tide and wave recorder
ciably affected by rainfall and freshwater runoff, but instead was also deployed adjacent to the Aquadopp to measure
are primarily determined by shoreline geometry, solar wave statistics. The TWR-2050P sampled water pressure at
radiation, and wind speed (Guzmán et al. 2005; Kauffman 4 Hz for 3-min bursts every h.
and Thomson 2005). Therefore, the effects of river runoff
and rainfall on SST were not addressed in this study. Heating scenarios
To validate the numerical flow model, a Nortek Inc.
AquadoppÓ high-resolution velocity profiler was deployed The time frame July–October of 2010 was selected as the
during August 2011 during a 5-d period to obtain velocity ‘present-day’ setting for this study. In addition, the hourly
profiles throughout the water column. The Aquadopp water temperature data collected at the four sites from 2006
measured 3-D velocities within bins spaced at a 0.03 m to 2010 were used to determine the intraannual pattern of
vertical elevation within the water column, at a sampling water temperatures across the study area (Fig. 2). The

123
 Coral Reefs

32 Table 2 IPCC predictions of July weather data (a) after spatial

interpolation across the study site within each scenario: 1961–1990,
31 2010–2039 (2020s), 2040–2069 (2050s), and 2070–2099 (2080);
(b) followed by temporal interpolation for year 2010 and calibration
with the STRI site datasets
Water temperature, °C

30
July of year Qs Ta Tw rhum Fc U10
29 (W m-2) (°C) (°C) (%) (%) (m s-1)

28 (a)
1961–1990 194.1 26.1 26.8 85.0 74.7 1.8
27 2010–2039 194.9 27.0 27.7 84.7 71.5 1.9
2040–2069 204.7 28.2 28.7 81.7 64.2 2.2
26
2070–2099 218.0 29.8 30.3 76.8 57.7 2.3
(b)
25
1 2 3 4 5 6 7 8 9 10 11 12 2010 168.5 27.0 27.3 87.2 65.6 1.6
Month 2024 168.8 27.3 27.7 87.1 64.0 1.6
Fig. 2 Intraannual variability of water temperature from 2006 to 2054 177.3 28.5 28.7 83.9 56.6 1.9
2010, based on hourly water temperature data collected from the four 2084 188.8 30.1 30.3 78.9 52.1 2.0
HOBO sites (Fig. 1). The box diagrams show the medians (horizontal
lines), the 25th and 75th % (lower and upper rectangle edges), and
whiskers extending to the furthest points within each month
at 15-min intervals for future scenarios and applied at each
time step in the hydrodynamic model.
month with the lowest monthly median water temperature
was January (27.5 °C), and the month with the highest Model setup
median water temperature was October (29.8 °C), followed
by September (29.7 °C), June (29.6 °C), May (29.4 °C), The hydrodynamic model implemented to simulate local
and August (29.2 °C). Hence, August–October was selec- water flow and SST was the open-source Delft3D-FLOW,
ted to represent the continuous three-month period in which solves the nonlinear shallow water equations derived
which the study site was the most heated. Additionally, from the three-dimensional Navier–Stokes equations for
July was added to the time frame as the spin-up time incompressible free surface flow (Lesser et al. 2004;
allowing the numerical model to stabilize. Kernkamp et al. 2005). This model was run in a two-
The datasets from the STRI station were used to represent dimensional (2D) mode, and vertically homogeneous
the local weather across the study domain from July to conditions were assumed (Kauffman and Thomson 2005).
October 2010, while the monthly mean weather patterns Therefore, computed water flow and SST estimates are
predicted by the Intergovernmental Panel on Climate depth-averaged values. The same domain of the 2D model
Change (IPCC 2003) were applied for July–October in 2024, was used for all four simulations consisting of July–
2054, and 2084 (Table 2). Two IPCC prediction points most October in 2010, 2024, 2054, and 2084.
adjacent to the study site were selected (Fig. 1) and spatially The computational grid (Fig. 3), encompassing all the
interpolated to determine the parameters Qs, Ta, Tw, rhum, Fc, adjacent bays in addition to the study site, extended offshore
and U10 within each of the future time periods studied. These to a distance of at least 20 km from the study site to minimize
parameters were used within the numerical model to deter- errors occurring at the open boundary. The model domain was
mine heat flux estimates at the water surface. Site 1 and Site 2 partitioned using a Cartesian, curvilinear, orthogonal grid cell
were weighted 1/3 and 2/3, respectively, to allow for a arrangement that contained 514 9 246 nodes to cover
greater Caribbean Sea influence on these variables. To cal- approximately 7,000 km2 of water surface area. The model
ibrate IPCC predictions, the monthly mean weather data for grid had space-varying resolution that allowed for a grid
July–October 2010 were generated (e.g., Table 2a–July) and resolution of B250 m within the study site, to a coarser hor-
were compared with the monthly means from the STRI site izontal resolution of approximately 400 m on the offshore
observations. To adjust for localized changes in the vari- boundary. The available bathymetry data (Table 3), however,
ables, Qs, Ta, rhum, and U10 across the study domain, the had coarser resolution (926 m) than the grid resolution, such
ratios of observed means from the STRI site observations to that multiple small islands were not recognized. To better
predicted means from IPCC in each month of 2010 were represent the shoreline changes, the bathymetry assigned for
computed. This ratio was then used to adjust the mean IPCC each grid cell was determined by a combination of available
predictions in 2024, 2054, and 2084 (e.g., Table 2b–July). topography datasets. The shoreline data were implemented to
Time series of Qs, Ta, Tw, rhum, and U10 were then generated define the actual contour of 0 m in bathymetry. Bounded by

123
Coral Reefs

Fig. 3 Computational grid

(left) and grid resolution (right)
over the study site at Bocas del
Toro. Note that because of the
high spatial resolution of the
grid, for visualization purpose,
the edges of only every 4 9 4
actual cells are shown on the
left. The continuous shade on
the right indicates the space-
varying resolution of the actual
grid

Table 3 List of available topography datasets provided by Delft

DashBoard (DDB) over the Bocas del Toro archipelago Flow boundary conditions

Database Coverage Resolution (m) Year updated

Starting from an initial condition of a uniform water level
GEBCO Ocean 926 2008 of f = 0 m over the entire domain, the flow was primarily
SRTM Land 93 2008 forced by tides at the open boundaries and wind stress at
WVS Shoreline \500 2013 the free surface, while drag was imposed at the sea bed by
bottom roughness. Along the offshore open boundary
The databases were originally collected by DDB from various sour-
ces: GEBCO bathymetry data from the British Oceanographic Data (Fig. 3), water level conditions induced by astronomical
Centre (BODC), SRTM land elevation data from Jarvis et al. (2008), tides were defined using thirteen tidal constituents (M2, S2,
and the WVS shoreline data from the National Geospatial-Intelli- N2, K2, K1, O1, P1, Q1, Mf, Mm, M4, MS4, and MN4)
gence Agency (NGA) extracted from the database TPXO 7.2 Global Inverse Tide
Model (Egbert et al. 1994; Egbert and Erofeeva 2002). At
the shoreline, the land elevation data were projected onto the
the free surface, the effect of wind stress was applied via
model grids using triangular interpolation to identify small
spatially uniform wind speed and wind direction data at
islands at a resolution of 93 m. Similarly bounded, the pro-
15-min intervals for each scenario. While the wind speed
jected General Bathymetric Chart of the Oceans (GEBCO)
data varied from year to year based on IPCC predictions,
bathymetry was added to the empty grids. The remaining grid
the wind direction time series in July–October were
cells were filled by internal diffusion followed by further
assumed to be the same throughout the simulated years. At
smoothing along selected shoreline segments.
the seafloor, a spatially varying bottom roughness was
Based on the grid and bathymetry of the domain, the time
developed using Manning’s n roughness coefficients of
step of the simulation to ensure numerical stability and
0.02 for hardground, 0.035 for sea grass, and 0.05 for coral
accuracy of the model, as indicated by the Courant number
reef locations (Prager 1991; Madsen et al. 2001; Reiden-
(C) for 2D models (Stelling 1984), was determined as:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi bach et al. 2006). Spatial distribution of these benthic
 ffi
1 1 habitats, namely coral reefs (IMaRS-USF and IRD 2005;
C ¼ 2Dt gh þ ð1Þ IMaRS-USF 2005) and sea grasses (Green and Short 2003),
Dx2 Dy2
was retrieved from United Nations Environmental Pro-
where Dt is the time step (s), g is the acceleration of gravity gram-World Conservation Monitoring Center (UNEP-
(m s-2), h is the local water depth (m), and Dx and Dy are WCMC and Short 2005; UNEP-WCMC 2010).
grid lengths in x and y directions (m) within each cell. For
locations with large differences in bottom geometry or Heat flux
coastline, the C should not exceed a value of ten. Con-
strained by this characteristic Courant value, a time step of An initial heat condition was uniformly applied over the
Dt = 6 s was adopted in the simulation to maintain the domain using monthly mean Tw values of July for each
accuracy and stability over the entire domain. scenario (Table 2b). Heat flux at the water surface was

123
 Coral Reefs

where S = 1,368 W m-2 is the solar constant, and c (°) is

the solar elevation angle (Grena 2012).

Model validation and sensitivity analysis

To quantify the ability of the model to predict SST within

the study site, preliminary simulations were conducted for
a period between 1 July and 10 August 2010. Simulated
water temperatures at the end of the time period were
compared with the observed values at the four temperature
logger sites (Fig. 1) using a goodness-of-fit (GoF) function
over time (t), layer (r = 1), and model domain (n, m):
nP P P   o
 
t r n;m Wt;r;n;m  yt;r;n;m  xt;r;n;m
GoF ¼ nP P P  o
t r n;m Wt;r;n;m

ð6Þ
Fig. 4 Heat exchange mechanism at the earth’s water surface, where where W = 1/4 denotes the weight function, y represents
Qsc is the solar insolation (J m-2 s-1), Qs is the solar radiation (J m-2 the model-simulated SST, and x the observed SST. The sets
s-1), Qsr is the reflected solar radiation (J m-2 s-1), Qa is the
atmospheric radiation (J m-2 s-1), Qar is the reflected atmospheric of coefficients tested were (1) the Stanton number (cH) that
radiation (J m-2 s-1), Qbr is the back radiation (J m-2 s-1), Qev is the controls the magnitude of the convective heat flux and the
latent heat flux (J m-2 s-1), and Qco is the sensible heat flux (J m-2 Dalton number (ce) for the evaporative heat flux, (2) the
s-1). The combination of net solar radiation (Qsn = Qs - Qsr) and net horizontal eddy diffusivity (DH) and horizontal eddy vis-
atmospheric radiation (Qan = Qa - Qar) serves as the energy input to
the surface cosity (vH), which were set equal to each other (Boudreau
and Jorgensen 2001) and specify the magnitude of turbu-
lent mixing, and (3) the wind drag coefficient (Cd) that
simulated within the Delft3D-FLOW model and driven by
prescribes the magnitude of the surface wind stress. The
meteorological variables (Fig. 4), and governed by the heat
resulting GoF, allowing for spatial and temporal variations
balance equation (Gill 1982):
in model performance, reflects increasing model quality
Qtot ¼ Qsn þ Qan  Qbr  Qev  Qco ð2Þ with decreasing values.
We investigated the influences of the three-parameter
where Qtot is the total heat flux (W m-2), Qsn is the net
sets by varying each parameter independently over a wide
solar radiation (W m-2), Qan is the net atmospheric radi-
range of physically reasonable values. Values of cH ranged
ation (W m-2), Qbr is the back radiation (W m-2), Qev is
from 7.9 9 10-3 to 2.255 9 10-3, while ce ranged from
the latent heat flux (W m-2), and Qco is the sensible heat
1.2 9 10-3 to 1.885 9 10-3 (Gill 1982). cH and ce were
flux (W m-2). Tw was then simulated by quantifying ver-
both tested using a step size of 0.5 9 10-3 from 1 9 10-3
tical energy exchange at the air–water interface and hori-
to 2 9 10-3. The magnitudes of DH and vH were tested for
zontal advective heat transfer due to fluid motion (Webb
values between 10-1 and 102 m2 s-1 within the possible
and Zhang 1999; Hannah et al. 2004, 2008). The model
range for coastal settings. In addition, while the magnitude
was prescribed by spatially uniform variables rhum, Ta, Qsn,
of Cd showed a strong regional dependency, most of the
and Fc, of which Qsn was derived from:
values fall within the range of 0.5 9 10-3–1.5 9 10-3 as
Qsn ¼ ð1  aÞQs ð3Þ indicated by empirical formulas (Garratt 1977; Heaps
where a = 0.06 is the albedo coefficient defined within the 1965) and therefore were assessed between this range at
model. Fc was calculated through the function (Gill 1982): 0.5 9 10-3 intervals, in addition to the baseline value
0.63 9 10-3. The resulting minimum values of the GoF
Qs were found at cH = 2 9 10-3, ce = 1 9 10-3, DH = -
1:0  0:4Fc  0:38Fc2 ¼ ð4Þ
Qsc mH = 10 m2 s-1, and Cd = 0.5 9 10-3. Due to the mini-
Qsc (W m-2) is the incident solar radiation for clear sky mal change of the GoF values across the range of Cd, the
conditions (Gill 1982), value Cd = 0.63 9 10-3 was chosen, consistent with the
 original heat flux formulation. Overall, the GoF value was
0:76S sinðcÞ sinðcÞ  0 reduced from 0.385 °C in the initial setup to 0.236 °C in
Qsc ¼ ð5Þ
0:0; sinðcÞ\0 the optimized run, showing a 39 % decrease in the

123
Coral Reefs

Table 4 Comparison of the setup of model runs and their GoF values ‘Warning’ is issued when DHW [0 °C weeks, a coral
with respect to the water temperature data observed at the four tem- bleaching ‘Alert Level 1’ when values near a reef reach
perature logger sites (Fig. 1) in August 01–August 10, 2010 DHW C4 °C weeks, and an ‘Alert Level 2’ when reefs
Setup cH ce DH = Cd GoF experience DHW C8 °C weeks. Mass bleaching and the
vH (m2 s-1) (°C) onset of coral mortality are expected after Alert Level 2.
Initial 0.00145 0.0015 1 0.00063 0.385
Optimized 0.002 0.001 10 0.00063 0.236
Results
disagreements between the model results and observations
Model validation
(Table 4).
In addition to the overall model performance evaluated
Based on the optimized model setup, the predicted variations
by GoF, a normalized skill score termed the index of
in water level f (Fig. 5) and water temperature Tw (Fig. 6)
agreement (IAS) was applied to compare model predictions
show good consistency with the observed patterns within the
to in situ observations in the 2010 scenario at each indi-
model domain between August and October 2010. The best
vidual observation site. The IAS is defined as (Willmott
tidal agreement between model and observations occurred at
1982; Warner et al. 2005):
the STRI site with an IAS of 0.937. The Tw prediction at STRI,
P
ðy  xÞ2 however, reproduced less of the high-frequency temperature
IAS ¼ 1  P ð7Þ
ðjy  xj þ jx  xjÞ2 fluctuations observed in situ, but the overall trend was in
agreement. While COB and CAB saw slightly better model
where the over-bar denotes a time average. An IAS value of agreement in predicting Tw than STRI, site ROB had greater
1.0 corresponds to a 100 % agreement between model deviations, and overall lower modeled water temperatures
prediction of SST, y, and observations of SST, x, while compared to observations due to its location within the inner
decreasing values indicate decreasing agreement, or poorer lagoon area, far away from the open boundary tidal forcing.
model performance. The IAS values were shown to be relatively high in general,
ranging from 0.593 to 0.937 (Table 6). The depth-averaged
Degree heating weeks water velocity (mean ± SD) obtained in situ at a site
approximately 1 km offshore from the STRI site was
The cumulative thermal stress to the coral reef ecosystem 1.36 ± 0.69 cm s-1, which compared well to model results
was quantified based upon ‘degree heating weeks’ (DHW), at the same location of 1.47 ± 1.02 cm s-1. The average
in the same way as NOAA Coral Reef Watch program uses significant wave height during this period was 2.8 ± 2.6 cm,
DHW from satellite-derived SST. Specifically, DHW (°C with a maximum wave height of 8.4 cm, suggesting that
week) is calculated by summing the positive deviations waves are typically very small within the protected bays.
(whenever C1 °C) of the weekly average SST over a 12-
week period (Eakin et al. 2009; Strong et al. 2006): Flow and SST
X
DHW ¼ ðSST  MMMÞ ð8Þ
Instantaneous flow magnitude and velocity vectors mod-
where MMM is the maximum monthly mean (°C) calculated eled for 20 August 2010 are shown in Fig. 7 and are rep-
by choosing the mean temperature during the warmest month resentative of predominant spatial flow patterns for all four
of the year. In this study, the MMM value of 28.5 °C was simulated time periods between 2010 and 2084. Maximum
used based on data from the NOAA 50 km local monthly water velocities were approximately 0.4 m s-1 and were
mean satellite-only nighttime SST. The same alert levels as located approximately 2 km offshore. Large reductions in
NOAA (summarized in Table 5) were used, where a flow magnitude were found along a transect from the off-
shore to the inshore, such that interior bays of the Bocas del
Table 5 A summary of the alert levels for coral bleaching modified Toro archipeligo had mean flows of approximately
from NOAA (2009) 0.05 m s-1. In contrast to these weak flows within interior
Status Interpretation Definition bays, water flow at several inlets between adjacent islands
was quite large, generating periodic flows up to 0.4 m s-1.
Warning Thermal stress is accumulating DHW [0 Predicted SST for the years 2010 through 2084 is shown
Alert Level 1 Bleaching expected DHW [4 in Fig. 8. While the flow patterns showed reduced magni-
Alert Level 2 Widespread bleaching and DHW [8 tude from offshore to inshore, an overall reverse gradient of
some mortality expected
SST occurred, with increasing SST magnitude of approx-
The alert levels are based on DHW for each site imately 2–3 °C from the offshore to inshore areas.

123
 Coral Reefs

0.4 Table 6 Index of agreement (IAS), for water level f and water
observed
predicted
temperature Tw between simulations and observations at various
0.3 locations (Fig. 1) in the model domain
Variable Location IAS
0.2
,m

f STRI 0.937
0.1
Tw STRI 0.637
Water level

Tw COB 0.711
0
Tw ROB 0.593
-0.1 Tw CAB 0.773
Values shown are for August–October 2010. Higher IAS values
-0.2
indicate better agreement between the observed and the simulated
variables
-0.3
01-Sep 01-Oct
Date
Temporally, SST increased approximately 1 °C every
Fig. 5 Comparison of the predicted water level f with the observed
values at STRI for September 2010
30 yr at most locations from 2010 to 2084. By the year
2084, virtually all locations inside or near the interior bays
of the Bocas del Toro archipelago reach an SST [31 °C
during August.

Degree heating weeks

Water temperature, °C

33
a observed
32 predicted
31 Derived from the SST predictions, the DHW patterns
30 (Fig. 9) show large variations of the thermally induced
29 stress on coral reef ecosystems at different locations and
28 time periods when compared to the maximum monthly
01-Aug 01-Sep 01-Oct 01-Nov
mean of 28.5 °C (Eq. 8). Within each modeled time period,
there was an increase in DHW values from the open ocean
Water temperature, °C

33 toward the mainland. The largest magnitudes of DHW

32 b
were found within the interior bays and adjacent to islands
31
within shallow waters, which often coincided to regions
30
where coral reefs are located. For increasing time periods
29
from 2010, DHW values increased within the bay, from
28
01-Aug 01-Sep 01-Oct 01-Nov DHW = 0 °C at the open ocean to approximately
DHW = 12–16 °C weeks within the bay by the year 2054.
In addition to the increase in DHW in regions that had
Water temperature, °C

33
c positive DHW under the present-day scenario (2010), the
32
31 contours of DHW also expanded toward the open ocean in
30 future years, covering an increasingly larger area affected
29 by thermal stress. By the year 2084, all regions within or
28 adjacent to the bay reached DHW [8 °C weeks. In addi-
01-Aug 01-Sep 01-Oct 01-Nov
tion, even the open ocean within our study site experienced
significantly higher values of DHW.
Water temperature, °C

33
d To determine spatial variations in DHW across the
32
measurement domain, four locations (shown in Fig. 9d)
31
where coral bleaching events had been reported (ReefBase
30
2013) were chosen. The four bleaching sites started from
29
differing DHW values in 2010, ranging from
28
01-Aug 01-Sep 01-Oct 01-Nov DHW = 2.3 °C weeks at Site 4 to 9.5 °C weeks at Site 3
Date (Fig. 10). In spite of large variations in DHW in 2010, the
DHW values at the four sites converged to a magnitude of
Fig. 6 Comparison of the predicted water temperature Tw with the
observed values at the a STRI, b COB, c ROB, and d CAB from DHW = 18–19 °C weeks in year 2084. While Site 3 had
August to October 2010 the highest DHW in 2010, its value was exceeded by both

123
Coral Reefs

Fig. 7 Predicted magnitude and

vector arrows of the depth-
averaged flow velocities
(m s-1) at 0000 hrs on August
20, 2010, as a representative
flow pattern within the study
site

Fig. 8 Predicted SST (°C) distribution at 0000 hrs on 20 August during a 2010, b 2024, c 2054, and d 2084. Note higher temperatures along
inshore, shallow regions

Site 1 and Site 2 in 2084, by 0.7 and 0.3 °C weeks, 2084. The numerical model predicted that the four sites
respectively. Rapid increases in DHW were especially experienced different thermal stress during different time
evident from 2054 to 2084. In particular, Site 4 increased periods, hence falling into various categories of alert levels
from DHW = 4.8 °C weeks in 2054 to 18.1 °C weeks in (Table 7). Site 3, located inside the bay and close to the

123
 Coral Reefs

Fig. 9 Degree heating weeks DHW (°C week) within the study site depicting reported coral bleaching sites, and black circles in
during August–October of a 2010, b 2024, c 2054, and d 2084, with d showing four representative bleaching sites
hollowed red polygons representing coral reefs, white circles

STRI station, had an Alert Level 2 starting from 2010 and

remained at Alert Level 2 for all future years. In contrast to
this continuously high level of bleaching alert, the other
three bleaching sites, located toward the open ocean, saw 30
#1
lower stress in present day (2010), but had increased alert #2
levels with time. Site 4 had the most drastic change in alert #3
25
#4
level, from a Warning in 2024 to Alert Level 2 in 2084.
20
DHW, °C Week

Water retention times

15
Heating of the water masses, and increased SST, is depen-
dent both upon the water column depth (affecting the surface
10
area to volume ratio of the water) and the magnitude of the
exchange flow with the open ocean. To determine the
5
retention time of water masses within specific regions of the
model domain, the domain was divided into equivalent
0
1 km2 regions, and total water volume (m3) within each 2000 2020 2040 2060 2080 2100
1 km2 region, computed as the surface area multiplied by Year
average water depth, was divided by the average volumetric
Fig. 10 Trend of the DHW values at the four coral bleaching sites
flux of water out of the volume (m3 s-1) over a spring-neap (locations shown on Fig. 9d) throughout time. DHW of all four sites
cycle in August 2010. Here, we define retention time as how showed continuous increases from 2010 to 2084, with a more
long a water parcel, starting from a specified location within centralized distribution of DHW in 2084

123
Coral Reefs

Table 7 List of the alert levels at the four coral bleaching sites Discussion
(Fig. 9) throughout time
Site 2010 2024 2054 2084 Model performance on predicting flow and SST
location
Model validation indicated that the observed f and Tw at
#1 Alert Level Alert Level Alert Level Alert Level
1 2 2 2 different locations within the study site were well repli-
#2 Alert Level Alert Level Alert Level Alert Level cated by the numerical model, with high IAS values. Even
1 1 2 2 though STRI station was the only site where tidal height
#3 Alert Level Alert Level Alert Level Alert Level model performance could be assessed due to the lack of
2 2 2 2 availability of observational data at other locations, accu-
#4 Warning Warning Alert Level Alert Level rate tidal predictions at this interior bay site suggested that
1 2 tidal and flow predictions were reasonably accurate
Threshold of each level was based on DHW and associated criteria throughout the domain. The predictions on Tw (or equiva-
summarized in Table 5 lently, SST), however, saw noticeable spatial variations in
the model agreement with observations, such that COB and
CAB had higher IAS values than STRI and ROB. Both
a waterbody, will remain in the waterbody before exiting COB and CAB were less protected and were more directly
(Monsen et al. 2002). Results show that locations corre- influenced by open ocean conditions than the sites located
sponding to low water velocities had the largest water at STRI and ROB. In addition, trends in observed and
retention times, and locations corresponding to both high modeled Tw, as shown in Fig. 6, depict closer agreement at
water retention times and shallow water depths had the the STRI site than at ROB, likely due to the location of
greatest increases in SST under future warming scenarios ROB being along the interior of the bay, the furthest dis-
(Fig. 11). These regions typically corresponded to near- tance from the outer boundary where tidal forcing was
shore areas where coral reefs are located. Although water imposed. However, much fewer coral reefs are located in
retention times within each 1 km2 area near regions con- the vicinity of ROB than other site locations, and therefore,
taining coral reefs typically ranged between 0 and 20 h, the this lower agreement with observations likely had minor
average time these water parcels resided within the bay were impacts on overall heating results under future warming
typically much greater than this. scenarios.

Fig. 11 Average retention time

of water masses within
equivalent 1 km2 regions,
averaged across a spring-neap
cycle for August 2010

123
 Coral Reefs

Model limitations correlate with locations of highest observed coral bleaching.

This may be due to discontinuities in monitoring or report-
Within the numerical model, the accuracy of the predicted ing of bleaching areas of coral reefs in Bocas del Toro, or
flow and SST relies largely upon the quality of the may be a result of corals that have adapted to high water
boundary and initial conditions, taken predominately from temperatures or highly variable water temperatures. For the
the STRI station. Although the model did an excellent job modeled 2010 scenario, areas of low water velocities and
in predicting weekly to monthly trends in overall thermal high water retention times tended to have both high SST and
heating and cooling, the model, at times, did a relatively high variability in SST, potentially leading to a coral envi-
poor job of predicting diurnal temperature variability. It is ronment of greater thermal resistance (Buddemeier et al.
assumed that data measured at STRI are both accurate and 2004; McClanahan et al. 2007a; Jones et al. 2008).
representative of the entire study site. This assumption may The spatial distribution of the modeled thermal stress, in
poorly reproduce local short-term variability in wind pat- addition to its overall temporal trend, demonstrates the
terns, atmospheric pressure, surface runoff from the land, importance of geographic variations faced by the corals
and precipitation. Furthermore, due to the relatively coarse under predicted climate change scenarios. Even with lim-
resolution of bathymetry data applied in the model, the ited records on specific bleaching events, the bleaching Site
water depth ascribed to near the shorelines, especially those 3, recorded in 2002, was no longer recognized as a coral
adjacent to many of the small islands, is of lower accuracy. reef site after a 2010 assessment (ReefBase 2013). This
These variables can possibly explain the lower model follows trends from predicted DHW and bleaching alerts,
agreement to measured water temperatures for sites at where an Alert Level 1 was estimated to exist in 2010
greater distances from STRI. In addition, the long-term where some bleaching is expected, to Alert Level 2 in
predictions of future global warming scenarios depend 2024, which corresponds to the status at which widespread
largely on the credibility of the IPCC monthly averaged bleaching and coral mortality is expected. DHW estimates,
values. As part of the nature of climate predictions, especially those at the four bleaching sites, showed sig-
approximations and uncertainties always exist. nificant spatial variability from 2010 to 2054, ranging from
The predicted bleaching alert takes into account only the low-level thermal stress to a high mortality threat. By 2084,
thermal stress under the projected global warming. However, almost all coral reefs are susceptible to bleaching-induced
a number of factors influence the susceptibility of corals to mortality, except for a region of relatively lower thermal
bleaching. For example, adaptation by the algal symbionts stress along the outer boundary of the archipelago. By
may increase coral resistance to thermal stress during periods 2084, only corals exposed to open ocean currents are pre-
of elevated SST and assist in coral recovery from bleaching dicted to survive, while the inner bays will be completely
afterward (Baker et al. 2004; Little et al. 2004; Rowan 2004). devoid of coral, if DHW threat warnings prove correct.
Studies have indicated that some coral reefs may shift to The spatial trends in thermal stress within the Panama-
symbiotic algae that are more temperature tolerant as a way nian reefs agree with field and numerical studies conducted
to adapt to coral bleaching (Hoegh-Guldberg 2005). Other along a fringing coral reef-lagoon system in Australia
important factors such as reef morphology, storm events, and where the difference in temperature between the reef and
sea level rise were also not included in this study. surrounding offshore waters was predominantly a function
of both water residence time and the daily mean net heat
Thermal stress, coral bleaching, and implications flux, while diurnal variations in water temperatures were
driven by the diurnal fluctuation in the net heat flux (Zhang
As predicted by the numerical model, Bocas del Toro et al. 2013). Inner lagoon reef temperatures were found to
Archipelago’s SST increased in most locations at about be substantially higher than offshore waters under normal
0.03 °C per year, along with an increase in DHW of more weather conditions, while temperatures across the entire
than 0.1 °C weeks on average. The average DHW increase reef were elevated during extreme heating events, which
for the four selected bleaching sites was 0.158 °C weeks per lead to mass coral bleaching. Our findings are also likely
year. This result agrees with the predictions of flow patterns applicable to many coral reef regions worldwide, and in
that indicate that low water velocity regions within or particular reefs that are found in shallow and partially
adjacent to the bay that cause long water retention times are enclosed coastal regions with long water retention times.
more susceptible to heating under predicted atmospheric Based on the susceptibility of these Panamanian coral
warming conditions. Considering that most of the coral reefs reefs to bleaching, there is a need to consider the man-
were in a bleaching Warning or Alert Level 1 in the 2010 agement status of existing marine protected areas (MPA).
scenario, this trend of increasing SST with time is of sub- Current MPAs in this study site were designated within a
stantial environmental concern. Within the modeled 2010 limited region, located adjacent to Bleaching Site 2
scenario, regions that show Alert Level 1 do not necessarily (Fig. 1). While Site 2 was estimated to have Alert Level 1

123
Coral Reefs

status until 2024, indicating less severe bleaching, areas Davis K, Lentz S, Pineda J, Farrar J, Starczak V, Churchill J (2011)
close to Site 4 were predicted to be under ‘Warning,’ and Observations of the thermal environment on Red Sea platform
reefs: a heat budget analysis. Coral Reefs 30:25–36
therefore be less susceptible to thermal stress under pro- Doney SC, Ruckelshaus M, Duffy JE, Barry JP, Chan F, English CA,
jected warming scenarios. MPA design criteria often allo- Galindo HM, Grebmeier JM, Hollowed AB, Knowlton N,
cate higher priority to areas that are both likely to survive Polovina J, Rabalais NN, Sydeman WJ, Talley LD (2012)
thermal stress due to climate change and have high biodi- Climate change impacts on marine ecosystems. Ann Rev Mar
Sci 4:1–27
versity (McClanahan et al. 2007b). Using these predicted Donner SD (2009) Coping with commitment: projected thermal stress
scenarios, it is suggested that coral reefs located within on coral reefs under different future scenarios. PLoS One
regions with shorter water retention times that are exposed 4:e5712
to exchange flows with the open ocean, such as Site 4, Eakin CM, Lough JM, Heron SF (2009) Climate variability and change:
monitoring data and evidence for increased coral bleaching stress. In:
should be prioritized for MPA management. However, van Oppen M, Lough JM (eds) Coral bleaching: patterns, processes,
results suggest that we should expect large-scale coral causes and consequences. Springer, Heidelberg, pp 41–67
bleaching events, and likely coral die-off, in the coming Egbert GD, Erofeeva SY (2002) Efficient inverse modeling of
decades within this Caribbean reef. barotropic ocean tides. J Atmos Ocean Tech 19:183–204
Egbert GD, Bennet AF, Foreman MGG (1994) TOPEX/POSEIDON
tides estimated using a global inverse model. J Geophys Res
Acknowledgments We thank the scientific staff at the Smithsonian 99:24821–24852
Tropical Research Institute in Bocas del Toro, Panama, for providing ESRI (2010) DeLorme World Base Map 2010. ArcGIS Online. http://www.
meteorological and sea surface temperature data. This research was arcgis.com/home/item.html?id=b165c3df453e4be6b5ac4fdb241effbe
funded by the National Science Foundation (NSF-OCE 1151314) and by Ferrier-Pages C, Richard C, Forcioli D, Allemand D, Pichon M, Shick
the University of Virginia’s Deepening Global Education Grant Program. JM (2007) Effects of temperature and UV radiation increases on
the photosynthetic efficiency in four scleractinian coral species.
Biol Bull 213:76–87
Garratt JR (1977) Review of drag coefficients over oceans and
References continents. Mon Weather Rev 105:915–929
Gill AE (1982) Atmosphere-Ocean dynamics. Academic Press 662 pp
Aronson RB, Precht WF, Toscano MA, Koltes KH (2002) The 1998 Glynn PW (1993) Coral-reef bleaching - ecological perspectives.
bleaching event and its aftermath on a coral reef in Belize. Mar Coral Reefs 12:1–17
Biol 144:435–447 Green EP, Short FT (2003) World atlas of seagrasses. Prepared by the
Ateweberhan M, McClanahan TR (2010) Relationship between UNEP-WCMC. University of California Press, Berkley, USA
historical sea-surface temperature variability and climate Grena R (2012) Five new algorithms for the computation of sun
change-induced coral mortality in the western Indian Ocean. position from 2010 to 2110. Sol Energy 86:1323–1337
Mar Pollut Bull 60:964–970 Guzmán H, Barnes PAG, Lovelock CE, Feller IC (2005) A site
Baird AH, Bhagooli R, Ralph PJ, Takahashi S (2009) Coral description of the CARICOMP mangrove, seagrass and coral
bleaching: the role of the host. Trends Ecol Evol 24:16–20 reef sites in Bocas del Toro, Panama. Caribb J Sci 44:430–440
Baker A, Starger CJ, McClanahan T, Glynn PW (2004) Corals’ Hannah DM, Malcolm IA, Soulsby C, Youngson AF (2004) Heat
adaptive response to climate change. Nature 430:741 exchanges and temperatures within a salmon spawning stream in
Berkelmans R, Oliver JK (1999) Large-scale bleaching of corals on the Cairngorms, Scotland: seasonal and sub-seasonal dynamics.
the Great Barrier Reef. Coral Reefs 18:55–60 River Res Appl 20:635–652
Berkelmans R, Weeks SJ, Steinberg CR (2010) Upwelling linked to Hannah DM, Malcolm IA, Soulsby C, Youngson AF (2008) A
warm summers and bleaching on the Great Barrier Reef. Limnol comparison of forest and moorland stream microclimate, heat
Oceanogr 55:2634–2644 exchanges and thermal dynamics. Hydrol Process 22:919–940
Bindoff NL, Willebrand J, Artale V, Cazenave A, Gregory J, Gulev S, Harvell CD, Mitchell CE, Ward JR, Altizer S, Dobson AP, Ostfeld RS
Hanawa K, Le Quere C, Levitus S, Nojiri Y, Shum CK, Talley (2002) Climate warming and disease risks for terrestrial and
LD, Unnikrishnan A (2007) Observations: Oceanic climate marine biota. Science 296:2158–2162
change and sea level. In: Solomon S, Qin D, Manning M, Chen Heaps NS (1965) Storm surges on a continental shelf. Philos Trans A
Z, Marquis M, Averyt KB, Tignor M, Miller HL (eds) Climate Math Phys Eng Sci 257:351–383
change 2007: the physical science basis. Contribution of working Hoegh-Guldberg O (1999) Climate change, coral bleaching and the
Group 1 to the Fourth Assessment Report of the Intergovern- future of the world’s coral reefs. Mar Freshw Res 50:839–866
mental Panel on Climate Change. Cambridge University Press, Hoegh-Guldberg O (2005) Low coral cover in a high-CO2 world.
Cambridge and New York, pp 384–428 J Geophys Res C 110:C09S06
Boudreau BP, Jorgensen BB (2001) The Benthic boundary layer: Hoegh-Guldberg O (2011) The impact of climate change on coral reef
transport processes and biogeochemistry. Oxford University ecosystems. In: Dubinsky Z, Stambler N (eds) Coral reefs: an
Press, New York ecosystem in transition. Springer, Dordrecht, Netherlands,
Bradley P, Fisher WS, Bell H (2009) Development and implemen- pp 391–403
tation of coral reef biocriteria in U.S. jurisdictions. Environ Hoegh-Guldberg O, Mumby PJ, Hooten AJ, Steneck RS, Greenfield
Monit Assess 150:43–51 P, Gomez E, Harvell CD, Sale PF, Edwards AJ, Caldeira K,
Bruno JF, Selig ER, Casey KS, Page CA, Willis BL, Harvell CD, Knowlton N, Eakin CM, Iglesias-Prieto R, Muthiga N, Bradbury
Sweatman H, Melendy AM (2007) Thermal stress and coral RH, Dubi A, Hatziolos ME (2007) Coral reefs under rapid
cover as drivers of coral disease outbreaks. PLoS Biol 5:e124 climate change and ocean acidification. Science 318:1737–1742
Buddemeier RW, Baker AC, Fautin DG, Jacobs JR (2004) The Hughes TP, Baird AH, Bellwood DR, Card M, Connolly SR, Folke C,
adaptive hypothesis of bleaching. In: Rosenberg E (ed) Coral Grosberg R, Hoegh-Guldberg O, Jackson JBC, Kleypas J, Lough
health and disease. Springer-Verlag, Berlin, pp 427–444 JM, Marshall P, Nystrom M, Palumbi SR, Pandolfi JM, Rosen B,

123
 Coral Reefs

Roughgarden J (2003) Climate change, human impacts, and the Monsen NE, Cloern JE, Lucas LV, Monismith SG (2002) A comment
resilience of coral reefs. Science 301:929–933 on the use of flushing time, residence time, and age as transport
IMaRS-USF (2005) Millennium Coral Reef Mapping Project. UNEP time scales. Limnol Oceanogr 47:1545–1553
World Conservation Monitoring Centre. Cambridge, UK. http:// Muller EM, Rogers CS, Spitzack AS, van Woesik R (2008) Bleaching
data.unep-wcmc.org/datasets/13 increase likelihood of disease on Acropora palmata (Lamarck) in
IMaRS-USF and IRD (2005) Millennium Coral Reef Mapping Hawksnest Bay, St. John, US Virgin Islands. Coral Reefs
Project. UNEP World Conservation Monitoring Centre. Cam- 27:191–195
bridge, UK. http://data.unep-wcmc.org/datasets/13 Peñaflor EL, Skirving WJ, Strong AE, Heron SF, David LT (2009)
IPCC (2003) Future climate in world regions: an intercomparison of Sea-surface temperature and thermal stress in the Coral Triangle
model-based projections for the new IPCC emissions scenarios. over the past two decades. Coral Reefs 28:841–850
The Intergovernmental Panel on Climate Change. http://www. Prager EJ (1991) Numerical simulation of circulation in a Caribbean-
ipcc-data.org/ddc_visualisation.html type backreef lagoon. Coral Reefs 10:177–182
IUCN and UNEP-WCMC (2010) The World Database on Protected Pratchett MS, Wilson SK, Graham NAJ, Munday PL, Jones GP,
Areas (WDPA). UNEP-WCMC, Cambridge, UK. http://www. Polunin NVC (2009) Coral bleaching and consequences for
protectedplanet.net motile reef organisms: Past, present and uncertain future effects.
Jokiel PL, Brown EK (2004) Global warming, regional trends and In: van Oppen M, Lough J (eds) Coral bleaching: patterns and
inshore environmental conditions influence coral bleaching in processes, causes and consequences. Springer, Heidelberg,
Hawaii. Glob Chang Biol 10:1627–1641 pp 139–158
Jones AM, Berkelmans R, van Oppen MJH, Mieog JC, Sinclair W Reaser JK, Pomerance R, Thomas PO (2000) Coral bleaching and
(2008) A community change in the algal endosymbionts of a global climate change: scientific findings and policy recommen-
scleractinian coral following a natural bleaching event: field dations. Conserv Biol 14:1500–1511
evidence of acclimatization. Proc R Soc Lond B Biol Sci ReefBase (2013) A Global Information System for Coral Reefs.
275:1359–1365 http://www.reefbase.org
Kauffman KW, Thomson RC (2005) Water temperature variation and Rehr AP (2012) A decision support framework for science-based,
the meteorological and hydrographic environment of Bocas del multi-stakeholder deliberation: a coral reef example. Environ
Toro, Panama. Caribb J Sci 41:392–413 Manage 50:1204–1218
Kernkamp HWJ, Petit HAH, Gerritsen H, de Goede ED (2005) A Reidenbach MA, Monismith SG, Koseff JR, Yahel G, Genin A (2006)
unified formulation for the three-dimensional shallow water Boundary layer turbulence and flow structure over a fringing
equations using orthogonal co-ordinates: theory and applica- coral reef. Limnol Oceanogr 51:1956–1968
tions. Ocean Dynamics 55:351–369 Reynolds RW, Smith TM (1994) Improved global sea surface
Lesser MP (1996) Exposure of symbiotic dinoflagellates to elevated temperature analysis using optimal interpolation. J Clim
temperatures and ultraviolet radiation causes oxidative stress and 7:929–948
inhibits photosynthesis. Limnol Oceanogr 41:271–283 Reynolds RW, Zhang HM, Smith TM, Gentemann CL, Wentz F
Lesser MP, Farrell JH (2004) Exposure to solar radiation increases (2005) Impact of in situ and additional satellite data on the
damage to both host tissues and algal symbionts of corals during accuracy of a sea-surface temperature analysis for climate. Int J
thermal stress. Coral Reefs 23:367–377 Climatol 25:857–864
Lesser MP, Stochaj WR, Tapley DW, Shick JM (1990) Bleaching in Rowan R (2004) Coral bleaching: Thermal adaptation in reef coral
coral reef anthozoans: effects of irradiance, ultraviolet radiation, symbionts. Nature 430:742
and temperature on the activities of protective enzymes against Stelling GS (1984) On the construction of computational methods for
active oxygen. Coral Reefs 8:225–232 shallow water flow problems. Rijkswaterstaat Commun 35, The
Lesser GR, Roelvink JA, van Kester J, Stelling GS (2004) Develop- Hague, Netherlands
ment and validation of a three-dimensional morphological Strong AE, Arzayus F, Skirving W, Heron SF (2006) Identifying coral
model. Coast Eng 51:883–915 bleaching remotely via Coral Reef Watch–improved integration
Levitus S, Antonov JI, Boyer TP, Locarnini RA, Garcia HE, and implications for changing climate. In: Phinney JT, Hoegh-
Mishonov AV (2009) Global ocean heat content 1955–2008 in Guldberg O, Kleypas J, Skirving W, Strong A (eds) Coral reefs
light of recently revealed instrumentation problems. Geophys and climate change: science and management, coastal and
Res Lett 36:L07608 estuarine studies 61. American Geophysical Union, Washington,
Little AF, van Oppen MJ, Willis BL (2004) Flexibility in algal D.C, pp 163–180
endosymbioses shapes growth in reef corals. Science 304:1492–1494 Thur SM (2010) User fees as sustainable financing mechanisms for
Liu G, Strong AE, Skirving W, Arzayus LF (2006) Overview of marine protected areas: an application to the Bonaire National
NOAA coral reef watch program’s near-real time satellite global Marine Park. Mar Policy 34:63–69
coral bleaching monitoring activities. Proc 10th Int Coral Reef UNEP-WCMC, Short FT (2005) Global distribution of seagrasses. V2.0.
Symp 1:1783–1793 http://www.arcgis.com/home/item.html?id=de158acb9f884db3bff
Madsen J, Chambers P, James W, Koch E, Westlake D (2001) The 1a39dd056913b
interaction between water movement, sediment dynamics and UNEP-WCMC, WorldFish Centre, World Resources Institute, The
submersed macrophytes. Hydrobiologia 444:71–84 Nature Conservancy (2010) Global distribution of warmwater
Manzello DP, Brandt M, Smith TB, Lirman D, Hendee JC, Nemeth coral reefs, compiled from multiple sources, including the
RS (2007) Hurricanes benefit bleached corals. Proc Natl Acad Millennium Coral Reef Mapping Project. UNEP World Conser-
Sci USA 104:12035–12039 vation Monitoring Centre. Cambridge, UK. http://data.unep-
McClanahan TR, Graham NAJ, Calnan JM, MacNeil MA (2007a) wcmc.org/datasets/13
Toward pristine biomass: reef fish recovery in coral reef marine Walther GR, Post E, Convey P, Menzel A, Parmesan C, Beebee TJC,
protected areas in Kenya. Ecol Appl 17:1055–1067 Fromentin JM, Hoegh-Guldberg O, Bairlein F (2002) Ecological
McClanahan TR, Ateweberhan M, Muhando CA, Maina J, Moham- responses to recent climate change. Nature 416:389–395
med MS (2007b) Effects of climate and seawater temperature Warner JC, Sherwood CR, Arango HG, Signell RP (2005) Perfor-
variation on coral bleaching and mortality. Ecol Monogr mance of four turbulence closure models implemented using a
77:503–525 generic length scale method. Ocean Modeling 8:81–113

123
Coral Reefs

Webb BW, Zhang Y (1999) Water temperatures and heat budgets in Willmott CJ (1982) Some comments on the evaluation of model
Dorset chalk water courses. Hydrol Process 13:309–321 performance. Bull Am Meteorol Soc 63:1309–1369
Wells JR, Fram JP, Pawlak G (2012) Solar warming of near-bottom Zhang Z, Falter J, Lowe R, Ivey G, McCulloch M (2013)
water over a fringing reef. J Mar Res 70:641–660 Atmospheric forcing intensifies the effects of regional ocean
Wilkinson CR (1996) Global change and coral reefs: Impacts on warming on reef-scale temperature anomalies during a coral
reefs, economies and human cultures. Glob Chang Biol bleaching event. J Geophys Res 118:4600–4616
2:547–558

123