Access to this full-text is provided by Springer Nature.
Content available from Scientific Reports
This content is subject to copyright. Terms and conditions apply.
1
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
Three decades of heat stress
exposure in Caribbean coral reefs:
a new regional delineation to
enhance conservation
Aarón Israel Muñiz-Castillo1, Andrea Rivera-Sosa1, Iliana Chollett2, C. Mark Eakin
3,
Luisa Andrade-Gómez
4, Melanie McField5 & Jesús Ernesto Arias-González
1
Increasing heat stress due to global climate change is causing coral reef decline, and the Caribbean has
been one of the most vulnerable regions. Here, we assessed three decades (1985–2017) of heat stress
exposure in the wider Caribbean at ecoregional and local scales using remote sensing. We found a high
spatial and temporal variability of heat stress, emphasizing an observed increase in heat exposure
over time in most ecoregions, especially from 2003 identied as a temporal change point in heat
stress. A spatiotemporal analysis classied the Caribbean into eight heat-stress regions oering a new
regionalization scheme based on historical heat exposure patterns. The temporal analysis conrmed
the years 1998, 2005, 2010–2011, 2015 and 2017 as severe and widespread Caribbean heat-stress
events and recognized a change point in 2002–2004, after which heat exposure has been frequent in
most subsequent years. Major heat-stress events may be associated with El Niño Southern Oscillation
(ENSO), but we highlight the relevance of the long-term increase in heat exposure in most ecoregions
and in all ENSO phases. This work produced a new baseline and regionalization of heat stress in the
basin that will enhance conservation and planning eorts underway.
Reefs worldwide are being exposed to heat stress at greater frequency and intensity1–5. Heat stress disrupts the
symbiotic relationship between coral and the microscopic algae that inhabit the coral. is loss of symbionts in
the coral host is termed “bleaching” and impedes the coral’s ability to obtain energy via photosynthesis. It may
also lead to coral death unless temperatures improve and the densities of its symbiotic algae are restored6,7. Severe
heat stress acts as the main precursor to large-scale bleaching, many disease outbreaks, and consequent mortal-
ity3,4,6–11. Bleaching increases the vulnerability of corals to other anthropogenic stressors and can have devastating
impacts on reef biodiversity and ecosystem services6,7,12. ese ecological consequences are of signicant global
concern, as many nations depend on coral reefs ecosystem services, such as coastal protection, sheries and
tourism for their livelihood and survival13. Also, future projections predict that under the scenario that reects a
continuation of current emissions (RCP 8.5 used by the Intergovernmental Panel on Climate Change) coral reefs
are likely to be exposed to severe heat stress every year by mid-21st century2,14.
Heat stress is a fundamental stressor that must be characterized and prioritized to best identify potentially
resilient reefs for conservation. Along with other indicators (e.g. depth, connectivity, ocean currents), heat stress
can oer a portfolio of optimal reefs for conservation and restoration2,15–19. One common approach is to iden-
tify sites with a history of minimal past heat stress to seek possible refugia from climate change2,15–17. e other
approach includes seeking if past heat stress may have increased the tolerance of corals and therefore inuenced
coral adaptation20–26. Historical patterns of heat stress are also useful in placing projections of future climate
change in context2,14,27,28. Consequently, identifying regional variations in historical heat stress is crucial in
1Laboratorio de Ecología de Ecosistemas de Arrecifes Coralinos, Departamento de Recursos del Mar, Centro de
Investigación y de Estudios Avanzados del I.P.N. Mérida, 97310, Yucatán, Mexico. 2Smithsonian Marine Station,
Smithsonian Institution, Fort Pierce, Florida, 34949, USA. 3Coral Reef Watch, National Oceanic and Atmospheric
Administration, College Park, Maryland, 20740, USA. 4Unidad de Recursos Naturales, Centro de Investigación
Cientíca de Yucatán, A.C., Mérida, 97200, Yucatán, Mexico. 5Healthy Reefs for Healthy People, Smithsonian Marine
Station, Fort Pierce, Florida, 34949, USA. Correspondence and requests for materials should be addressed to A.I.M.-C.
(email: aaron.muniz@cinvestav.mx) or J.E.A.-G. (email: earias@cinvestav.mx)
Received: 25 January 2019
Accepted: 10 July 2019
Published: xx xx xxxx
OPEN
Content courtesy of Springer Nature, terms of use apply. Rights reserved
2
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
determining which areas have been exposed to the greatest and the least risk of coral bleaching in the past and a
minimum of what is likely in the future.
On a large spatiotemporal scale, one of the major drivers of heat stress causing bleaching is El Niño-Southern
Oscillation (ENSO)1,7,21. ENSO is a complex phenomenon and is one of the most forceful drivers of climate pat-
terns worldwide29,30. ENSO is linked to the Caribbean via a tropical atmospheric bridge, although the Caribbean
is also inuenced by the thermal inertia of Atlantic variability31–33. ENSO events building atop global heat stress
has corresponded with global bleaching events (1997–1998, 2010, 2014–2017)1,3–5,7,34 and El Niño has been linked
to heat stress, bleaching and other impacts in the Caribbean1,7,9,11,24,35–37. But ENSO has not always been the
driver of heat stress, as tropical forcing probably played a minor role in the 2005 Caribbean bleaching event38,39.
Additionally, heat stress is not solely related to the warm-phase, El Niño, since warm thermal anomalies are pres-
ent somewhere in both positive and negative ENSO phases. As a result, La Niña leads to coral bleaching in some
locations, and warming global ocean temperatures have caused La Niña years now to be warmer than they were
during El Niño events three decades ago1,3.
e Caribbean has historically been one of the areas most exposed to heat stress and is characterized by high
spatial variation in its thermal patterns2,17,40. ese heat stress patterns subsequently resulted in the observed
magnitude3,7,10,41–43 and the spatial footprint of coral bleaching across the Caribbean10. Long-term assessments
of heat stress in the basin can oer an understanding of past disturbance patterns related to the current state and
variation of coral cover and species composition3,6,7,12. ose heat stress patterns can be useful in identifying
potential “thermal refugia” (regions that escaped heat stress)15,16,19,44 or regions with frequent past heat stress
where surviving corals may have developed adaptation20–24,26. is information also helps to better understand
the potential impact of projections of future heat stress2,14,27,28. erefore, assessing historical variability becomes
critical to understand heat stress exposure, especially when constant and severe bleaching risk is predicted for
Caribbean reefs by 20502,14.
Here we apply a newly available SST dataset from 1985 to 20175 and provide a spatiotemporal contextualiza-
tion of the wider Caribbean heat stress. is study aimed to:
(a) Characterize the geographical extent and variability of heat stress in the Caribbean ecoregions45 during the
last three decades,
(b) classify the wider Caribbean into new heat-stress regions based on historical heat stress,
(c) assess the temporal variability of heat stress in the Caribbean ecoregions45 and its relation to past ENSO
events based on the Oceanic Niño Index-ONI.
Results
Spatiotemporal variability in overall heat stress. e ecoregions45 within the wider Caribbean exhib-
ited a high spatial variability of heat stress exposure (maximum Degree Heating Weeks, DHW) from 1985 to 2017
(Fig.1a,b). Heat stress within 20 km of coral reefs around the wider Caribbean ranged from 0.0 to 25.6 °C-weeks
across the entire time series. 83% of Caribbean reef area was exposed to “bleaching risk” (≥4 °C-weeks) at some
time between 1985 and 2017 (Fig.1c,d), and 42% of the area was exposed to “mortality risk” (≥8 °C-weeks) at
least once (Fig.1e,f). roughout the paper, we refer to these two thresholds because they are dened as the levels
of heat stress likely to cause coral bleaching and mortality2,10,46.
e ecoregions with the highest heat stress were the Southern Caribbean (SC), Eastern Caribbean (EC),
Southwestern Caribbean (SWC), Southern Gulf of Mexico (SGoM) and Western Caribbean (WC; Fig.1;
TablesS1–S8). These five ecoregions experienced significantly higher heat stress than the rest of the wider
Caribbean according to a heteroscedastic one-way ANOVA and post hoc tests for most indicators (TablesS5–S8).
ese ecoregions experienced exposure to elevated DHW values and bleaching and mortality risk events (Fig.1a–f;
TablesS1–S3). All these regions except for the EC showed an increase that ranged from 0.10 to 0.35 °C-weeks
per year, obtained from the trend analysis of annual maximum DHWs (Fig.1g,h; TableS4). e SC was the most
exposed to bleaching and mortality risk because most of the area within that ecoregion experienced more than
eight bleaching risk events and all the area showed at least one mortality risk event (Fig.1c–f). e SWC was
another ecoregion subjected to high heat stress, where most of the area experienced more than three bleaching
risk events and was exposed to at least one mortality risk event (Fig.1c–f). In contrast, the ecoregions least
exposed to heat stress were the Bahamian (BHM), Floridian (FL) and Greater Antilles (GA; Fig.1; TablesS1–S8).
ese ecoregions exhibited the greatest percentage of their areas without bleaching and mortality risk (Fig.1c–f).
However, even these ecoregions had high heat stress in some locations. e Florida Keys, Cuba, and areas of the
BHM showed high heat stress exposure and an increase of 0.1–0.2 °C-weeks per year (Fig.1a,c,e,g).
e most prominent heat-stress events in the wider Caribbean occurred during the years 1998, 2005, 2010,
2011, 2015 and 2017 (Fig.2). We found a high spatial variation in heat stress during the dierent major heat-stress
events (Fig.2c). e temporal patterns showed a constant exposure to heat stress from 2003 onwards, since this
year ~10% of the wider Caribbean has been exposed to bleaching risk annually (Fig.2d,e).
e most widespread event occurred in 2005 when 42% of the wider Caribbean suered its highest heat stress
(Fig.2a–c). e year 2010 was the second most widespread heat-stress event when 15% of the area reached its
maximum DHW (Fig.2a–c). e heat stress in 2010 was more intense than any other year, exposing the area of
the SC to values close to 25 °C-weeks, the highest DHW magnitude in the time series (Fig.2c,d). During these
two events, more than 50% of the wider Caribbean was exposed to bleaching risk and about 20% was exposed to
mortality risk (Fig.2e). e next warmest event for the entire basin occurred during 2015–2017, a variable but
long-lasting period, in which 25% of the area experienced its maximum heat stress (Fig.2a,b). In each of these
years, more than 20% of the wider Caribbean was exposed to bleaching risk and more than 5% of the area was
exposed to mortality risk (Fig.2e). Two other major heat-stress events were 1998 and 2011, in which 6–7% of the
wider Caribbean suered its maximum DHW (Fig.2a,b) and about 30% of the area was exposed to bleaching risk
in each of these years (Fig.2e).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
3
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
Heat-stress regions. e spatiotemporal variation of heat stress (cluster analysis using K-means and eight
optimal regions obtained using elbow criteria; Figs3a and S3) yielded eight spatially distinct heat-stress regions
(HSR) characterized by dierent time patterns of exposure levels (Figs3a–c and S4). e HSRs were consistent
with the heat stress patterns (Fig.3b), but did not follow the ecoregional delineation for the wider Caribbean - two
to three heat-stress regions were included within most ecoregions45 (Fig.3a). HSRs 1–5 were the most exposed
to elevated DHW, with a greater risk of bleaching and mortality, as well as a greater tendency to increase than the
other HSRs (Fig.3a,b; TablesS9–S16).
HSRs 1–3 were the most exposed to heat stress (Fig.3a; TablesS9–S12). ese HSRs were exposed to high
DHW values in several years, including 1995, 1998, and constant exposure since 2003, especially in 2003–2006,
2008, 2010–2011, and the last heat-stress event of 2014–2017 (Fig.3c). HSR 1, located along the Venezuelan coast
Figure 1. Spatial variability of heat stress exposure indicators in the wider Caribbean region from 1985–2017.
(a) Map showing heat stress values per pixel. (b) Histogram of the distribution of heat stress for ecoregions
and the wider Caribbean in the entire time series. (c) Map and (d) histogram of bleaching risk events
(≥4 °C-weeks). (e) Map and (f) histogram of mortality risk events (≥8 °C-weeks). (g) Map showing trend
of annual maximum DHW obtained by a Generalized Least Squares model (GLS), considering a temporal
autocorrelation; the grey pixels show non-signicant trend coecients (p-value > 0.05). (h) Histogram of the
annual trend of maximum DHW. Histograms for the ecoregions are ordered by statistical signicance supported
by a pairwise post hoc comparison of the heteroscedastic one-way ANOVA test (TablesS5–S8). e total
number of pixels (25,591) for the complete region represents an area of about 127,405 km2. e corresponding
numbers of pixels included in each ecoregion are shown in parenthesis. Maps were created using QGIS version
3.2.0 (https://www.qgis.org/en/site/)73.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
4
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
had the highest increase in annual maximum DHW and the most elevated frequency of bleaching and mortality
risk events (Fig.3a,b, TablesS9–S12). HSR 2 (Honduran and Nicaraguan Miskito Cays) and HSR 3 (lower Lesser
Antilles and western Venezuela) also exhibited high heat stress exposure (Fig.3a,b, TablesS9–S12).
Other HSRs considerably exposed to heat stress were HSR 4 (the Florida Keys, Bahamas, and southwestern
Cuba) and HSR 5 located in the southern Gulf of Mexico and the Gulf of Honduras (Fig.3a,b). ese areas
experienced high heat stress exposure and a considerable increase in the annual maximum DHW (Fig.3a,b;
TablesS9–S12), in these HSRs their maximum exposure to heat stress occurred during 2014–2017 (Fig.3c).
In contrast, HSRs 6–8 were least exposed to heat stress in the wider Caribbean (Fig.3a,b; TablesS9–S16). HSR
6 (upper Lesser Antilles) was the most exposed of HSRs 6–8, characterized by high heat stress exposure in 2005
and 2010 (Fig.3c) and suered the highest heat stress in the wider Caribbean during 2005. HSR 6 suered many
bleaching and mortality risk events, but the annual maximum DHW increased slowly (Fig.3b). HSR 7 (contain-
ing part of the Mesoamerican Reef, southern Cuba, Jamaica, Costa Rica and Panama) had low exposure to heat
stress, but a considerable increase in annual maximum DHW (Fig.3a,b; TablesS9–S16). Surprisingly, HSR 8
included the largest part of the wider Caribbean’s reef area (41.7%). HSR 8 was the area least exposed to heat stress
and was located mainly at northern latitudes (Fig.3a,b; TablesS9–S16).
Temporal cycles of heat stress and relationship to ENSO phases. Time series analyses (median of
the regional DHW values on a given day) conrmed that the strongest heat-stress events were observed during
1998–1999, 2004–2005, 2010–2011 and 2014–2017 (Fig.4a–h). e heteroscedastic one-way ANOVA test for
Figure 2. Spatiotemporal summary of heat-stress events in the wider Caribbean basin during 1985–2017. (a)
Percent of pixels with maximum DHW value in each year. (b) Year with maximum DHW value for the eight
ecoregions. (c) Major heat-stress events. (d) Temporal distribution of annual maxima (interquartile range and
median are represented with white box, outliers are represented with black points) and; (e) percentage of area
with bleaching risk (≥4 °C-weeks) and mortality risk (≥8 °C-weeks). Maps of annual maximum DHW for
the whole time series (1985–2017) can be found in the Supplementary FigsS1 and S2. Maps were created using
QGIS version 3.2.0 (https://www.qgis.org/en/site/)73.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
5
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
the time series showed that the SC, SGoM, WC, SWC and EC ecoregions had the greatest heat stress over the last
three decades (Fig.4a–h; TablesS17 and S18). ese areas were above the median of the wider Caribbean DHW
values in most years and during the strongest events, with values greater than 5 °C-weeks during the strongest
heat-stress events (Fig.4a–c; Fig.S5). e BHM and FL ecoregions showed median values higher than 5 °C-weeks
in 1997–1998, 2005, 2010 and 2014–2015 (Figs4f,g and S5). In the GA, the years 2005 and 2010 were the highest
heat-stress events, in which median regional values of ~3.5 °C-weeks were observed (Figs4h and S5).
Most ecoregions and the wider Caribbean have experienced constant heat stress since 2003. Change point
analysis identied the period of 2002–2004 as the temporal point when the time series changed signicantly
(Figs4a–h and S6; TableS19). is change point was dierent in the EC and FL, where it occurred between 1997
and 1998, and no signicant change point was observed in the GA (Fig.4e,g,h; TableS19). Moreover, the wavelet
analysis also showed that since 2003, the annual cycles of DHWs presented signicant periodicities in most sub-
sequent years (Fig.S7). e wavelet identied the frequencies and timing in which the major anomalies occurred,
Figure 3. Heat-stress regions and their maximum annual DHW during 1985–2017. (a) Reef locations within
heat-stress regions 1–8 (clusters) outlined by ecoregions. (b) Total annual maximum Degree Heating Weeks
(DHW), bleaching and mortality risk events and trends of annual maximum DHW. (c) Heat-stress regions 1–8
showing distribution of annual maxima, interquartile range and median are represented with white box, outliers
are represented with black points. e pink shadow represents the limit of mortality risk (≥8 °C-weeks). Map
was created using QGIS version 3.2.0 (https://www.qgis.org/en/site/)73.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
6
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
considered as signicant periods and time ranges in which the variation was higher than expected47. ese anal-
yses strengthened the results previously presented and recognized 1998, 2003–2006, 2008–2011 and 2014–2017
periods as the highest heat-stress events (Figs4a–h and S7).
To identify whether heat-stress events recognized in the wider Caribbean and across ecoregions may be
related to ENSO, we performed a cross-wavelet analysis to identify the signicant common periodicities between
the heat-stress events and the ONI48,49. 1998–2000 was the rst heat-stress period sharing common periodicities
with the ONI (Figs5b and S8). A strong El Niño occurred in 1997–1998 followed by a long-lasting La Niña event
in 1999–2000 (Figs4i and 5a). In 2005, ENSO had low inuence on heat stress as there was only a weak El Niño
followed by a brief, weak La Niña (Figs4i and 5a). e period from 2010 to 2012 showed the highest values
(darkest red) in the cross-wavelet, caused by the combination of high DHW and ONI variation (Fig.5b). e
2010–2011 period was classied as an El Niño event, followed by La Niña and a long neutral phase during 2012–
2013 (Figs4i and 5a). 2014–2017 also showed strong common periodicities with ENSO, when most ecoregions
were inuenced by the forceful 2015–2016 El Niño event (Figs5 and S8). e inuence observed in this last event
was remarkable even in the high latitude ecoregions such as the FL and BHM, where the common periodicities
with ONI were noted starting in 2014, perhaps due to the incipient El Niño in late 2014 (Figs5a,b and S8). Our
ecoregional results for the entire time series showed similar behavior with no dierences from the cross-wavelet
analysis. is consistency in the temporal heat stress patterns may have been related to the main events in the
wider Caribbean (1998, 2005, 2010, 2015 and 2017), which also correspond to the ecoregional level cross-wavelet
results (Figs5a,b and S8).
e cross-correlation analysis revealed a signicant positive correlation between El Niño (positive phase of
ENSO) and heat stress, this relationship presented the highest values in temporal lags of 6 to 12 months (p < 0.05,
Figure 4. Temporal patterns of Degree Heating Weeks (DHW) for ecoregions in the wider Caribbean and the
ONI during years 1985 to 2017. For each ecoregion (a–h) the vertical dotted black line shows the change point
analysis obtained via a Pettit test (TableS19), the horizontal dotted red line shows the limit of mortality risk
(≥8 °C-weeks), the black curve shows the median of the wider Caribbean DHW values on a given day and the
pink curve shows the median of the ecoregional DHW values on a given day. For ONI (i), red bars indicate El
Niño phases, blue bars indicate La Niña phases, and grey bars show neutral phases.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
7
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
Figs5c and S9). e 6–12 month time lag, with a signicant positive correlation, may be related to the delay that
occurs between the mature phase of El Niño in November to February and the heat stress peak that occurs in
August to December (Fig.S10). e cross-correlation analysis highlighted a signicant eight-month lag in the
wider Caribbean and in most ecoregions (Figs5c and S9). However, FL did not exhibit a signicant correlation
and SC showed a signicant negative and positive correlation at dierent time lags (Fig.S9).
e Generalized Linear Model (GLM) of annual variation of heat stress (annual hottest monthly average gen-
erated from the median of the regional DHW values on a given day) showed a signicant temporal increase in all
ecoregions (Table1; Figs6 and S11). e GLM obtained for the wider Caribbean and the ecoregions presented
a suitable t, with a lower Akaike Information Criterion (AICc) value than the Generalized Least Square (GLS)
models (TableS20), adding that the models obtained did not present temporal autocorrelation in the residuals.
During El Niño years, ecoregions generally experienced higher heat stress than the other ENSO phases (Figs6
and S11). e additive eect of the ENSO phases was signicant at the wider Caribbean level, and for the EC,
BHM, and GA ecoregions (Table1; Figs6 and S11). However, heat stress increased in all phases of ENSO, espe-
cially aer 2003, this long-term trend exceeded ENSO inuence in most ecoregions, nding that the additive
eect of ENSO phases was not signicant in four of the ve ecoregions most exposed to heat stress (Table1;
Fig.S11).
Discussion
Heat stress in the ecoregions was highly variable, with both spatial and temporal heterogeneity, but following a
general latitudinal gradient, as expected, across the wider Caribbean2,17,40. Generally, the ecoregions in the north-
ern Caribbean were the least exposed to heat stress and those in the south were the most exposed (Fig.1b). e
regions with the highest heat magnitude typically had an increase in heat stress through time and a high fre-
quency of heat-stress events. Time series analyses showed that the most relevant heat-stress events (1998, 2005,
2010–2011, 2014–2017) coincided with the most extreme bleaching episodes reported globally3,7,34 and in the
Caribbean3,7,10,41–43. e 2005 and 2010 events had the highest heat stress and can be considered the two periods
of greatest coral reef crisis in the Caribbean to date3,7,10,34,42,43.
The spatial variability of temporal heat stress exposure (annual maximum DHW) was used to develop
heat-stress regions (HSRs), a new scheme based on heat stress history that is more explanatory for heat stress pat-
terns than traditional ecoregions45. HSRs dene Caribbean areas that share a common history of exposure to heat
stress, providing a useful tool for spatial conservation and management15. In this sense, we recommend the use of
these HSRs at dierent scales (e.g. the wider Caribbean, within ecoregions or at country level). is new classica-
tion system can help identify regions exposed to recurring extreme heat stress such as HSR 1 and 2 (o Venezuela
and Miskito Cays, considered “historical hotspots”) where corals could potentially either suer repeated mor-
tality or develop adaptations that may increase resistance to bleaching20–26. Likewise, acclimatization studies are
needed in “emerging heat-stress regions”, regions that have experienced their greatest stress to date during the
Figure 5. Wider Caribbean DHW temporal patterns and ENSO relationship. (a) 1.5 years smoothed mean of
DHW (black line) and ONI (red line) time series. (b) Cross-wavelet showing the common power (color bar)
and phases (arrows). Phases arrow direction represents decreases of ONI and increases in DHW (le); increase
of ONI and increase in DHW (right). Black solid lines show the signicance of cross-wavelet power at 95%
condence. e ‘cone of inuence’ is represented by the white shadow; only results inside the cone of inuence
was ben considered and interpreted (outside the cone = high uncertainty). (c) cross-correlation between DHW-
ONI at dierent time lags. Red bars represent signicant positive correlation and blue bars represent signicant
negative correlation at 95% condence.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
8
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
latest mass-bleaching event (2014–2017)1,3,50. ese “emerging heat-stress regions” include some areas such as the
Mesoamerican Reef, southern Cuba and Florida Keys (HSR 4 and 5; Figs2 and 3). Within this heat-stress classi-
cation, the large HSR 8 region stands out for its low past heat stress exposure (a potential heat-stress refugium),
with reefs that have experienced few or no exposures to severe mortality risk events since suering moderate heat
stress and considerable bleaching in 200510. Studies suggest that while recent heat stress may inuence suscepti-
bility to bleaching25,26, this inuence decreases as the time since previous heat stress increases4.
Local-scale variability in oceanographic conditions such as depth, upwelling, currents, and water circulation
also inuences heat stress patterns at the local scale18–20,44,51. Regions such as northern Quintana Roo (HSR 7)
have lower heat stress due to the inuence of colder waters, high wave exposure and upwelling19,52. However,
upwelling has not provided refuge to the Caribbean’s most exposed region (HSR 1 in the Southern Caribbean19),
which experienced frequent and intense heat stress since 1990 (Figs3 and 4). Upwelling must be synchronous
with heat-stress events to reduce severe warming, making the timing of these events critical and adding complex-
ity to local-scale analyses of heat stress patterns and bleaching risk18,51. is complexity highlights the urgent need
for systematic coordinated Caribbean-wide bleaching monitoring programs that can provide a better understand-
ing of coral community responses to heat stress and environmental conditions.
Climate change projections of SST and heat stress that apply statistical downscaling analyses base their down-
scaling on historical data14,27. Given the spatiotemporal variability in heat stress found in this study, downscaling
eorts should try to include long time series in their analyses and only use spatial patterns that are stationary
through time. Additionally, given the stochastic nature and importance of episodic bleaching events, these pro-
jections should be updated frequently to capture new events. For example, some ecoregions strongly aected in
past years, such as the Eastern Caribbean with maximum heat stress in 2005, have experienced lower heat stress
in recent years, resulting in a low annual increase in heat stress. In contrast, “emerging heat-stress regions”, such
as the Southern Mesoamerican Reef and the Florida Keys were most exposed during 2014–2017, leading to a
signicantly increasing heat stress trend. However, events like these include a signicant stochastic component.
ese results suggest that the constant change in heat stress forms a problematic basis for long-term designation
of ‘resilient reefs’ or conservation areas more likely to survive the impacts of climate change. us, we recommend
caution in the use of heat stress patterns and thermal regimes for the prioritization of coral reef conservation
based on historical data16,17,44, particularly for those analyses that consider short term time series or only include
Ecoregion (explained deviance) Terms (df, dfr) Residual
Deviance F p-value
Wider Caribbean (0.527)
Null 8.6301
Years (1, 30) 6.0229 16.1568 0.00040
ENSO (2, 28) 4.082 6.0141 0.00672
Southern Caribbean (0.345)
Null 25.326
Years (1, 30) 18.834 10.909 0.00262
ENSO (2, 28) 16.583 1.8912 0.16966
Southern Gulf of Mexico (0.505)
Null 15.5578
Years (1, 30) 9.2556 19.2895 0.00015
ENSO (2, 28) 7.7055 2.3723 0.11175
Western Caribbean (0.562)
Null 13.4729
Years (1, 30) 7.1254 27.7206 0.00001
ENSO (2, 28) 5.8964 2.6837 0.08584
Southwestern Caribbean (0.448)
Null 18.0882
Years (1, 30) 11.2214 18.8147 0.00017
ENSO (2, 28) 9.9758 1.7064 0.19986
Eastern Caribbean (0.497)
Null 21.33
Years (1, 30) 14.829 13.9972 0.00084
ENSO (2, 28) 10.726 4.4176 0.02150
Bahamian (0.526)
Null 10.062
Years (1, 30) 8.5275 8.5552 0.00676
ENSO (2, 28) 4.7682 10.4791 0.00040
Floridian (0.236)
Null 11.9031
Years (1, 30) 9.8 5.819 0.02265
ENSO (2, 28) 9.0903 0.9818 0.38716
Greater Antilles (0.372)
Null 6.4482
Years (1, 30) 5.3653 6.3445 0.01776
ENSO (2, 28) 4.0447 3.8688 0.03285
Table 1. Analysis of the deviance obtained for Generalized Linear Model (GLM) with tests of the signicance
of the additive terms of years and phases of ENSO, with their respective degrees of freedom (df) and degrees
of freedom of residuals (dfr). e statistics for F tests and the p-value obtained for the Caribbean and the
ecoregions are presented. Values of p in bold are those considered statistically signicant.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
9
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
certain events (e.g. 1998, 2005 or 2010). We encourage a precautionary approach to selecting portfolios of conser-
vation areas, which includes reefs exposed to variable characteristics, such as those with high-frequency (daily or
weekly) variation in heat stress or temperature20, those with more constant heat stress exposure (potentially now
acclimated)21–25, and those that have (to date) experienced constant low heat stress2,15,17, as these statistics could
change with the next major heat-stress event.
e heat stress increased in the Caribbean since 2002–2004, in agreement with previous work1,2,19. is was
a change point aer which heat stress has been higher than in previous decades. is temporal pattern is slightly
apparent in the largest available global coral bleaching database34, in which it is possible to observe that from 2003
to 2010 about 50% of the reefs sampled per year in the Caribbean had moderate (11–50%) to severe (>50%)
bleaching (Supplementary Fig.S12). However, consistent reporting of coral bleaching episodes throughout the
ecoregions is limited, making it dicult to validate the ecological impacts of the spatiotemporal patterns of heat
stress3,34. Also, the high past exposure in some areas may have contributed to acclimatization processes or histori-
cal environmental ltering that may have eliminated the most susceptible individuals21–26, contributing to the lack
of relationship between current heat stress patterns and the local bleaching response. In this sense, we highlight
the importance of large ecoregional monitoring programs, such as the Healthy Reefs Initiative, which coordi-
nates regular reef monitoring and emergency response monitoring for beaching events, including the 2015–
2017 event53, with a publication focused on these data in preparation. Emerging heat stress has also occurred in
regions with insucient biological monitoring eorts; therefore, biodiversity loss related to bleaching and coral
diseases may have gone unreported in these areas (e.g., Miskito Cays in the HSR 2)16. is lack of information
is of particular concern given that major disease outbreaks have occurred during or aer heat-stress events in
the Caribbean8,9,11, highlighting the importance of monitoring aected areas during and aer heat-stress events.
Our results suggest that three out of four major heat-stress events in the Caribbean (1998, 2010–2011 and
2014–2017) have been inuenced by El Niño1,50. During these three Caribbean heat-stress events, bleaching, dis-
eases and a decrease in coral growth rates have all been associated with El Niño3,9,11,36,37. is relationship between
El Niño and heat stress showed a lag of 6–12 months, which partially corresponded with previously reported lag
times of 3–6 months for SST31–33. is lag could be associated with the delay in the climatological forcing of the
mature phase of ENSO (December to February, during the austral summer)29 until the appearance of heat stress
in the Caribbean during the boreal summer31–33 (Supplementary Fig.S10). Moreover, at the wider Caribbean level
and in the WC ecoregion, a signicant correlation was observed in a time lag of about two years, which may be
due to the eect of long-lasting events such as the 2014–2017. In this period an incomplete formation of a strong
El Niño in 2014–2015 was reported, followed by the 2015–2016 strong and long-lasting El Niño, which was linked
to a warm event that lasted until 201750.
Although some major Caribbean heat-stress events have been associated with El Niño, the long-term trend in
rising temperatures has caused heat stress during all ENSO phases - a pattern that has been recognized on reefs
globally3. Our results showed that this long-term trend is even more important in the most exposed ecoregions,
with four of the ve most exposed ecoregions showing no signicant additive eect of ENSO, while their overall
increase in heat stress was signicant (Table1; Fig.S11). Since the 1998 El Niño all subsequent El Niño events,
with the exception of the 2015–2016 El Niño, have been of lower intensity. However, even these weak or moderate
El Niño events can be associated with high exposure to heat stress as has been observed on coral reefs globally1,3.
For example, the most widespread heat-stress event in the Caribbean occurred in 2005, which was a relatively
weak El Niño event. e change in the heat stress regime since 2003 and the long-term trend observed could be
linked to other low-frequency patterns such as the recent Atlantic Multidecadal Oscillation (AMO) warm sig-
nal2,36,38,39 and anthropogenic climate change1–3,30,36,38. Both the AMO and climate change have been recognized as
important drivers in recent heat stress in the Caribbean, causing negative impacts on coral growth36 and climate
change has been strongly associated with slowing coral growth elsewhere54. is pattern of exposure to regular
Figure 6. Eect of time and ENSO phases in annual maximum of the monthly averages of wider Caribbean
DHW. (a) Conditional plot of time eect in annual heat stress, the color of points represents the dominant
ENSO phases in each year: neutral (black), La Niña (blue) and El Niño (red). (b) Cross-sectional plots
illustrating the t of the wider Caribbean annual heat stress with an additive interaction between time and
ENSO phases. (c) Box plots showing distribution of annual heat stress during ENSO phases. e ENSO
phase category was identied from the ONI time series (http://origin.cpc.ncep.noaa.gov/products/analysis_
monitoring/ensostu/ONI_v5.php), classifying El Niño years as those with anomalies above 0.05 °C, La Niña
years as those below −0.05 °C and neutral years as those in the range of −0.05 to 0.05 °C.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
10
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
and increasing heat stress not only poses a risk of coral bleaching and associated mortality but the potential nega-
tive eects of heat stress extend to reduce the overall functionality and ecosystem services provided by Caribbean
reefs.
This work produced a new contextualization of heat stress in the basin that will enhance conservation
and planning eorts currently underway. Given humanity’s critical dependence on marine resources in the
Caribbean13, the need to better understand and plan for future bleaching and disease events is paramount. We
highlight the relevance of multi-scale and retrospective analyses of heat stress in the contextualization of the
vulnerability of corals to bleaching in the wider Caribbean. It should be noted that the high spatial and temporal
variation found in heat stress exposure may aect the geographic patterns of potential adaptation or sensitivity
of corals to heat stress in the wider Caribbean. We also emphasize the potential impact of the last heat-stress
event (2014–2017) on some Caribbean ecoregions, particularly in the “emergent heat-stress regions”. Although
additional research is needed to identify the cause of low-frequency patterns on Caribbean heat stress, our results
provide evidence of a signicant change point in increasing heat stress since 2003. is chronic long term heat
stress in combination with acute heat-stress events may ultimately have an even greater impact on the condition
of Caribbean corals, by increasing their vulnerability to other stressors such as the devastating Stony Coral Tissue
Loss disease now aecting the wider Caribbean11,55,56.
Methods
Reef locations. Heat stress on coral reefs was characterized by analyzing the pixels located within 20 km of
reef locations within the wider Caribbean (32.7°N–8.4°N, 59.2°–97.0°W). By including contiguous areas, there is
a limitation within the analysis on the ecoregional and wider Caribbean scales, as zones with the absence of coral
reefs may be included. However, this 20-km buer was considered the best scale because it could identify oceanic
processes related to heat stress at the reef (100 m to 10 km) and regional scales (>10 km)57. is buer also allows
a better comparison with previous work, carried out applying a spatial resolution in a range from 4.5 km to 50
km1,2,4,10,19,42,43, recognized as the resolution range at which is possible to identify massive bleaching events46. Reef
locations were obtained from the Global Distribution of Coral Reefs58. is is the most comprehensive, published,
global dataset of warm-water coral reefs compiled from multiple sources.
Historical heat stress data. e spatiotemporal variation in daily Sea Surface Temperature (SST) from
1985 to 2017 was obtained from the NOAA’s Coral Reef Watch Program “CoralTemp” dataset, the latest and most
complete global satellite-derived dataset at a resolution of 5 km (0.05°) available for 1985 to present5 (https://cor-
alreefwatch.noaa.gov/product/5). e Maximum Monthly Mean (MMM) was also obtained from the Coral Reef
Watch Program version 3.1 dataset at 5 km (https://coralreefwatch.noaa.gov/satellite/bleaching5km/index.php),
the MMM is a value of SST that represents the warmest monthly climatological mean from 1985 to 2012 for each
location46. We then calculated the coral bleaching HotSpot (HS) and Degree Heating Weeks (DHW) metrics. HS
represent daily positive anomalies above the MMM (Equation1)46. DHW quantify heat stress by summing HS
above 1 °C over 84-days (12 weeks), divided by 7 to express values per week (Equation2)46, and calculated daily.
Analyses were conducted in R version 3.4.159 using the “raster”60 and “sp”61,62 libraries.
=
−>
≤.
HS
SST MMM SST MMM
SST MMM
,
0,
(1)
dailydaily
daily,
∑
=≥°
=
=
DHWHSifHS
1
7(, 1C)
(2)
i
j
ii
1
84
Oceanic Niño index data. e El Niño Southern Oscillation cycles and variation were determined using
the NOAA´s Oceanic Niño Index (ONI) version ve (http://origin.cpc.ncep.noaa.gov/products/analysis_mon-
itoring/ensostu/ONI_v5.php). is time series dataset provides the monthly average anomalies from 1950 to
date. ese monthly values were based on a 3-month running anomaly, calculated centered on a reference of
30-year base periods updated every 5 years (e.g. for 2000–2005 the reference is the 1985–2015 base period). All
ONI values calculated aer 2005 used the period 1985–2015 as a baseline. e spatial reference zone was situated
in the Tropical Pacic Ocean (5°N–5°S, 120°–170°W; Niño 3.4 region).
Data analyses. Spatiotemporal variation of heat stress. e annual maximum DHW was the main indicator
used to evaluate the exposure to heat stress and represents the maximum heat stress occurred in the year2,10,15,46.
We calculated the heat stress value observed for each pixel and year for the entire time series (FigsS1 and S2).
e ve main metrics calculated for each pixel were: a) the maximum DHW value per pixel for the entire time
series, b) the frequency of annual maximum DHW values ≥ 4 °C-weeks (a predictor of coral “bleaching risk”) per
pixel, c) the frequency of annual maximum DHW values ≥ 8 °C-weeks (a predictor of bleaching-induced mor-
tality or “mortality risk”) per pixel2,10,15,46, d) the year in which the maximum DHW occurred, and e) the trend
of the annual maximum DHW (dened below) per pixel. Analyses were conducted in R version 3.4.159 using the
“raster”60 and “sp”61,62 libraries.
e trend of annual maximum DHW was calculated with a Generalized Least Squares model (GLS), intro-
ducing to the regression a structure of temporal autocorrelation (AR1, which represents the covariance of order
1 considering the temporal similarity between the nearest years)63. Because we calculated the trend from annual
values, the GLS model did not consider seasonality. Once the slope of the regression was obtained, the signi-
cance of the slope was calculated at a 95% condence, considering as null hypothesis that the tendency was equal
Content courtesy of Springer Nature, terms of use apply. Rights reserved
11
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
to zero. In all pixels in which the slope was not signicant, the value of zero was set to represent a null slope. e
analyses were performed from the functions available in the “nlme” library64 of program R59.
To determine the dierences in the maximum DHW, the frequency of bleaching risk and mortality risk,
and the trend of annual maximum DHW among the ecoregions and the Caribbean, a heteroscedastic one-way
ANOVA for trimmed means test (0.10) and the corresponding pairwise post hoc comparison were performed.
is analysis included a comparison among the mean DHWs for each ecoregion and only considered the data
found from the 10th to the 90th percentile. ese tests were performed from the functions available in the
“WRS2” library65 of program R59.
Heat-stress regions. e regionalization of heat stress was performed by a clustering analysis with the K-means
algorithm through the unsupervised classication function present in the “RStoolbox” library66. e maximum
annual DHWs during the years 1985–2017 were used as input to the clustering procedure. To identify the opti-
mal number of groups, we used the graphic elbow criterion. is evaluation illustrated a curve of the remaining
variation from the addition of each given number of groups, revealing a relationship of the variance among added
groups and the total variance. In this way, we chose the least number of groups that explained the greatest spati-
otemporal variation. In order to visualize the arrangement of each of the pixels and their corresponding groups
resulting from the K-means algorithm, they were plotted on a two-dimensional plot of the rst two components
obtained from a Principal Component Analysis using the function present in the “FactoMineR” library67.
To test the dierence in the total annual maximum DHW for each year and the other exposure indicators
among the heat-stress regions (HSR), we performed a heteroscedastic one-way ANOVA for trimmed means
test (0.10), along with the corresponding pairwise post hoc comparison. ese analyses included a comparison
among the mean of each HSR and only considered the data found from the 10th to the 90th percentiles. ese
tests were performed from the functions available in the “WRS2” library65 of program R59.
Temporal cycles of heat stress and relationship to ENSO phases. Spatiotemporal daily data were summarized to
describe the temporal patterns at an ecoregional scale by calculating the median of the regional DHW values on
a given day. We tested the dierence of the regional median values among the dierent ecoregions, for this, we
considered all the values present in the days found within the months from September to November (recognized
as the season with greatest regional DHW values) in all the time series. e test was performed using a hetero-
scedastic one-way ANOVA for trimmed means and the corresponding pairwise post hoc comparison. We only
considered the data found from the 10th to the 90th percentiles. ese tests were performed from the functions
available in the “WRS2” library65 of program R59.
To identify patterns in the frequency of months or years, and for subsequent comparisons, the time series of
the median regional DHW values on a given day was averaged over each month. Using this monthly average as
a lower frequency indicator, Pettit’s non-parametric test68 was applied to identify whether there was a signicant
change point in the time series at the monthly scale in each of the ecoregions and in the wider Caribbean. is
test is a non-parametric comparison of the rank values of the sequence similar to the Mann-Whitney test and
identies a time point at which there is a signicant change in the variation and magnitude of monthly values.
ese analyses were performed using a p-value = 0.05.
e monthly frequency of heat-stress events and the relationship between heat stress and the ONI (both at
ecoregional and wider Caribbean scales), were characterized by wavelet and cross-wavelet analyses. e frequen-
cies and time in which the main anomalies occurred were identied by a wavelet analysis47–49. e cross-wavelets
analysis identied the common periodicities in the heat stress and ONI time series and assessed if they are
in phase (i.e., both time series increase in synchrony) or anti-phase (i.e., time series increases while the other
decreases)47. e frequencies and times considered as signicant were selected based on a Chi-Square test for
both techniques. For the statistical signicance in the case of wavelets, the null hypothesis was that the time
series was stationary at a given frequency over time, although in the cross-wavelet, the null hypothesis states that
time series had no variation in common and do not have signicant shared periodicities. For both analyses, we
rst applied a low-pass lter using the monthly mean to the daily DHW time series to match the temporal reso-
lution of the ONI to the monthly time series. To comply with the statistical assumption of normality needed for
this analysis47, we transformed the DHW data using a logarithmic transformation, this transformation allowed
us to improve the distribution of the data by decreasing the dierences in the values observed. Wavelet analy-
ses were conducted with the “biwavelet” library69, using the Morlet mother wavelet function and bias-corrected
cross-wavelet power with a 95% condence level47–49.
In addition, we calculated the cross-correlation function between the DHW and ONI time series to identify
the existing correlation considering dierent time lag periods between the time series. For this analysis, we con-
sider a maximum lag of 38 months, to provide at least 10 cycles in the entire time series (33 years). e statistical
signicance of the cross-correlation was calculated considering a 95% condence level. is analysis was per-
formed by the “tseries” library70.
To identify a temporal trend and determine if the ENSO phases had a signicant eect on the annual hottest
monthly average DHW values, generated from the median of DHW values across each region on a given day,
we compared a Generalized Linear Model (GLM) with a GLS model considering temporal autocorrelation. e
annual hottest monthly average DHW was considered as the dependent variable, considering as explanatory var-
iables the years and the category corresponding to the ENSO phase (neutral, La Niña and El Niño) introduced in
the model as additive terms. e dominant ENSO phase category by year was identied from the ONI time series
(http://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostu/ONI_v5.php). El Niño years were con-
sidered those with anomalies above 0.05 °C, La Niña years as those below −0.05 °C and Neutral years as those in
the range of −0.05 to 0.05 °C, these values had to be present in a range equal to or greater than ve months to be
Content courtesy of Springer Nature, terms of use apply. Rights reserved
12
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
designated as the dominant phase in each year71. In the GLS model, the temporal autocorrelation structure AR1
was introduced, which represents the covariance of order 1 (temporal similarity between the nearest years). In the
GLM model, the Gamma error family was chosen with a logarithm link function that adequately characterizes
continuous variables and is similar to the exponential curve. Once the models were made, graphical evaluations
of the residuals and the partial autocorrelation function were conducted, as well as a comparison between the
values of the Akaike Information Criterion of second order (for relatively small samples)72 for the two models
obtained by ecoregion and at the wider Caribbean level. e GLS model was made from the “nmle” library64,
while the other analyses were made from dierent functions available in the R program59.
Data Availability
Daily SST and the MMM data are available from NOAA CRW program CoralTemp Dataset version 3.1: https://
coralreefwatch.noaa.gov/satellite/coraltemp.php. e ONI time series data are available from NOAA: http://or-
igin.cpc.ncep.noaa.gov/products/analysismonitoring/ensostu/ONIv5.php. e main data used for the gures
and analyses were submitted to the NOAA National Centers for Environmental Information (NCEI).
References
1. Lough, J. M., Anderson, . D. & Hughes, T. P. Increasing thermal stress for tropical coral reefs: 1871–2017. Sci. Rep. 8, 6079 (2018).
2. Heron, S. F., Maynard, J. A., van Hooidon, . & Eain, C. M. Warming Trends and Bleaching Stress of the World’s Coral eefs
1985–2012. Sci. Rep. 6, 38402 (2016).
3. Hughes, T. P. et al. Spatial and temporal patterns of mass bleaching of corals in the Anthropocene. Science (80-.). 359, 80–83 (2018).
4. Hughes, T. P. et al. Global warming and recurrent mass bleaching of corals. Nature 543, 373–377 (2017).
5. Sirving, W. J. et al. e relentless march of mass coral bleaching: a global perspective of changing heat stress. Coral Reefs, https://
doi.org/10.1007/s00338-019-01799-4 (2019).
6. Hoegh-Guldberg, O. et al. Coral reefs under rapid climate change and ocean acidication. Science 318, 1737–42 (2007).
7. Baer, A. C., Glynn, P. W. & iegl, B. Climate change and coral reef bleaching: An ecological assessment of long-term impacts,
recovery trends and future outloo. Estuar. Coast. Shelf Sci. 80, 435–471 (2008).
8. andall, C. J. & van Woesi, . Contemporary white-band disease in Caribbean corals driven by climate change. Nat. Clim. Chang.
5, 1–5 (2015).
9. andall, C. J. & Van Woesi, . Some coral diseases trac climate oscillations in the Caribbean. Sci. Rep. 7, 1–8 (2017).
10. Eain, C. M. et al. Caribbean corals in crisis: ecord thermal stress, bleaching, and mortality in 2005. PLoS One 5 (2010).
11. Precht, W. F., Gintert, B. E., obbart, M. L., Fura, . & Van Woesi, . Unprecedented Disease-elated Coral Mortality in
Southeastern Florida. Sci. Rep, https://doi.org/10.1038/srep31374 (2016).
12. Hughes, T. P. et al. Global warming transforms coral reef assemblages. Nature 556, 492–496 (2018).
13. Pendleton, L. et al. Coral eefs and People in a High-CO2 World: Where Can Science Mae a Dierence to People? PLoS One 11,
e0164699 (2016).
14. van Hooidon, ., Maynard, J. A., Liu, Y. & Lee, S. . Downscaled projections of Caribbean coral bleaching that can inform
conservation planning. Glob. Chang. Biol. 21, 3389–3401 (2015).
15. Beyer, H. L. et al. is-sensitive planning for conserving coral reefs under rapid climate change. Conserv. Lett. e12587, https://doi.
org/10.1111/conl.12587 (2018).
16. Chollett, I., Enríquez, S. & Mumby, P. J. edening ermal egimes to Design eserves for Coral eefs in the Face of Climate
Change. PLoS One 9, e110634 (2014).
17. Selig, E. ., Casey, . S. & Bruno, J. F. New insights into global patterns of ocean temperature anomalies: Implications for coral reef
health and management. Glob. Ecol. Biogeogr. 19, 397–411 (2010).
18. Bayratarov, E., Pizarro, V., Eidens, C., Wile, T. & Wild, C. Bleaching susceptibility and recovery of Colombian Caribbean corals in
response to water current exposure and seasonal upwelling. PLoS One 8, 1–11 (2013).
19. Chollett, I. & Mumby, P. J. eefs of last resort: Locating and assessing thermal refugia in the wider Caribbean. Biol. Conserv. 167,
179–186 (2013).
20. Safaie, A. et al. High frequency temperature variability reduces the ris of coral bleaching. Nat. Commun. 9 (2018).
21. Coles, S. L. & Brown, B. E. Coral bleaching - Capacity for acclimatization and adaptation. Advances in Marine Biology 46, 183–223
(2003).
22. Guest, J. . et al. Contrasting patterns of coral bleaching susceptibility in 2010 suggest an adaptive response to thermal stress. PLoS
One 7, 1–8 (2012).
23. owan, . Coral bleaching: ermal adaptation in reef coral symbionts. Nature 430, 742–742 (2004).
24. Scheufen, T., rämer, W. E., Iglesias-Prieto, . & Enríquez, S. Seasonal variation modulates coral sensibility to heat-stress and
explains annual changes in coral productivity. Sci. Rep. 7, 1–15 (2017).
25. Hughes, T. P. et al. Ecological memory modies the cumulative impact of recurrent climate extremes. Nat. Clim. Chang. 9, 40–43
(2019).
26. Maynard, J. A., Anthony, . . N., Marshall, P. A. & Masiri, I. Major bleaching events can lead to increased thermal tolerance in
corals. Mar. Biol. 155, 173–182 (2008).
27. Wol, N. H., Mumby, P. J., Devlin, M. & Anthony, . . N. Vulnerability of the Great Barrier eef to climate change and local
pressures. Glob. Chang. Biol. 24, 1978–1991 (2018).
28. Langlais, C. E. et al. Coral bleaching pathways under the control of regional temperature variability. Nat. Clim. Chang. 7, 839–844
(2017).
29. Timmermann, A. et al. El Niño–Southern Oscillation complexity. Nature 559, 535–545 (2018).
30. Chiamoto, Y. et al. Silful multi-year predictions of tropical trans-basin climate variability. Nat. Commun. 6, 1–7 (2015).
31. lein, S. A., Soden, B. J. & Lau, N.-C. emote Sea Surface Temperature Variations during ENSO: Evidence for a Tropical
Atmospheric Bridge. J. Clim. 12, 917–932 (1999).
32. Giannini, A., Chiang, J. C. H., Cane, M. A., ushnir, Y. & Seager, . e ENSO teleconnection to the Tropical Atlantic Ocean:
Contributions of the remote and local SSTs to rainfall variability in the Tropical Americas. J. Clim. 14, 4530–4544 (2001).
33. Eneld, D. B. & Mayer, Da Tropical Atlantic sea surface temperature variability and its relation to El Niño-Southern Oscillation. J.
Geophys. Res. Ocean. 102, 929–945 (1997).
34. Donner, S. D., icbeil, G. J. M. & Heron, S. F. A new, high-resolution global mass coral bleaching database. PLoS One 12, 1–17
(2017).
35. Spillman, C. M., Alves, O. & Hudson, D. A. Seasonal Prediction of ermal Stress Accumulation for Coral Bleaching in the Tropical
Oceans. Mon. Weather Rev. 139, 317–331 (2011).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
13
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
36. Lizcano-Sandoval, L. D., Marulanda-Gómez, Á., López-Victoria, M. & odriguez-amirez, A. Climate Change and Atlantic
Multidecadal Oscillation as Drivers of ecent Declines in Coral Growth ates in the Southwestern Caribbean. Front. Mar. Sci. 6,
1–10 (2019).
37. Carilli, J. E., Norris, . D., Blac, B., Walsh, S. M. & McField, M. D. Century-scale records of coral growth rates indicate that local
stressors reduce coral thermal tolerance threshold. Glob. Chang. Biol. 16, 1247–1257 (2010).
38. Donner, S. D., nutson, T. . & Oppenheimer, M. Model-based assessment of the role of human-induced climate change in the 2005
Caribbean coral bleaching event. Proc. Natl. Acad. Sci. 104, 5483–5488 (2007).
39. Trenberth, . E. & Shea, D. J. Atlantic hurricanes and natural variability in 2005. Geophys. Res. Lett. 33, 1–4 (2006).
40. Chollett, I., Müller-arger, F. E., Heron, S. F., Sirving, W. & Mumby, P. J. Seasonal and spatial heterogeneity of recent sea surface
temperature trends in the Caribbean Sea and southeast Gulf of Mexico. Mar. Pollut. Bull. 64, 956–965 (2012).
41. McWilliams, J. P., Côté, I. M., Gill, Ja, Sutherland, W. J. & Watinson, A. . Accelerating impacts of temperature-induced coral
bleaching in the Caribbean. Ecology 86, 2055–2060 (2005).
42. Alemu I, J. B. & Clement, Y. Mass coral bleaching in 2010 in the Southern Caribbean. PLoS One 9, (2014).
43. Oxenford, Ha et al. Quantitative observations of a major coral bleaching event in Barbados, Southeastern Caribbean. Clim. Change
87, 435–449 (2008).
44. iegl, B. & Piller, W. E. Possible refugia for reefs in times of environmental stress. Int. J. Earth Sci. 92, 520–531 (2003).
45. Spalding, M. D. et al. Marine Ecoregions of the World: A Bioregionalization of Coastal and Shelf Areas. Bioscience 57, 573 (2007).
46. Liu, G. et al. eef-scale thermal stress monitoring of coral ecosystems: New 5-m global products from NOAA coral reef watch.
Remote Sens. 6, 11579–11606 (2014).
47. Grinsted, A., Moore, J. & Jevrejeva, S. Application of the cross wavelet transform and wavelet coherence to geophysical time series.
Nonlinear Process. Geophys. 11, 561–566 (2004).
48. Torrence Christopher, P. & Compo Gilbert, A. Practical Guide to Wavelet Analysis. Bull. Am. Meteorol. Soc. 79, 61–78 (1998).
49. Liu, Y., San Liang, X. & Weisberg, . H. ectication of the Bias in the Wavelet Power Spectrum. J. Atmos. Ocean. Technol. 24,
2093–2102 (2007).
50. Eain, C. et al. Ding, Dong, e Witch is Dead (?)– ree Years of Global Coral Bleaching 2014–2017. Reef Encount. 32, 33–38
(2017).
51. Chollett, I., Mumby, P. J. & Cortés, J. Upwelling areas do not guarantee refuge for coral reefs in a warming Ocean. Mar. Ecol. Prog.
Ser. 416, 47–56 (2010).
52. Varela, ., Costoya, X., Enriquez, C., Santos, F. & Gómez-Gesteira, M. Dierences in coastal and oceanic SST trends north of
Yucatan Peninsula. J. Mar. Syst. 182, 46–55 (2018).
53. Mceld, M. et al. 2018 Report Card for the Mesoamerican Reef, at, http://www.healthyreefs.org (2018).
54. Lough, J. M. & Cantin, N. E. Perspectives on massive coral growth rates in a changing ocean. Biol. Bull, https://doi.org/10.1086/
BBLv226n3p187 (2014).
55. Atlantic and Gulf apid eef Assessment (AGA) Program. Stony coral tissue loss disease. at http://www.agrra.org/coral-disease-
outbrea/
56. Meyer, J. L. et al. Microbial community shis associated with the ongoing stony coral tissue loss disease outbrea on the Florida eef
Tract. bioRxiv 626408, https://doi.org/10.1101/626408, (2019).
57. Lowe, . J. & Falter, J. L. Oceanic forcing of coral reefs. Ann. Rev. Mar. Sci. 7, 43–66 (2015).
58. UNEP-WCMC & Centre, W. F. Global Distribution of Coral eefs. World Wide Web Electron. Publ. http//data.unep-wcmc.org/
datasets 1–4 (2010).
59. Core Team. : A language and environment for statistical computing. (2017).
60. Hijmans, . J. aster: Geographic Data Analysis and Modeling. R package version 2.6–7 (2017).
61. Pebesma, E. J. & Bivand, . S. Classes and methods for spatial data in . News 5. 2 (2005).
62. Bivand, . S., Pebesma, E. J. & Gomez-ubio, V. Applied spatial data analysis with ; Second edition. (2013).
63. Weatherhead, E. C. et al. Factors aecting the detection of trends: Statistical considerations and applications to environmental data.
J. Geophys. Res. Atmos. 103, 17149–17161 (1998).
64. Pinheiro, J., Bates, D., Deboy, S. S. D. & . C. T. nlme: Linear and Nonlinear Mixed Eects Models. pacage version 3.1-131,
https://CAN.-project.org/pacage=nlme. R Packag. version 3.1-131, https//CAN.-project.org/pacage=nlme, https://doi.
org/10.1016/j.tibs.2011.05.003 (2017).
65. Mair, P. & Wilcox, . . obust statistical methods in using the WS2 pacage. J. Stat. Sow, https://doi.org/10.18637/jss.v000.i00
(2017).
66. Leutner, B., Horning, N., Schwalb-Willmann, J. & Hijmans, . J. Stoolbox: Tools for emote Sensing Data Analysis. pacage
version 0.2.3 (2018).
67. Lê, S., Josse, J. & Husson, F. FactoMine: An Pacage for Multivariate Analysis. J. Stat. Sow. 25 (2008).
68. Pettitt, A. N. A Non-Parametric Approach to the Change-Point Problem. Appl. Stat. 28, 126 (1979).
69. Gouhier, T. C., Grinsted, A. & Simo, V. Pacage ‘biwavelet’. Conduct Analyses, Bivariate Wavelet (V.0.20.17). 1–39 (2018).
70. Trapletti, A. & Horni, . tseries: Time Series Analysis and Computational Finance. R Packag. version 0.10-42 at, https://cran.
biodis.org/web/pacages/tseries/tseries.pdf (2017).
71. Gergis, J. L. & Fowler, A. M. Classication of synchronous oceanic and atmospheric El Niño-Southern Oscillation (ENSO) events
for palaeoclimate reconstruction. Int. J. Climatol. 25, 1541–1565 (2005).
72. Burnham, . P. & Anderson, D. . Model selection and multimodel inference: a practical information-theoretic approach, Second
Edition. Book, https://doi.org/10.1016/j.ecolmodel.2003.11.004 (2002).
73. QGIS Development Team. QGIS Geographic Information System. Open Source Geospatial Foundation at, https://www.qgis.org/en/
site/ (2018).
Acknowledgements
is paper is part of the fulllment requirements of the Ph.D. degree of AIMC in the postgraduate program of
Recursos Marinos of the Centro de Investigaciones y Estudios Avanzados (CINVESTAV) Unidad Mérida. is
program is acknowledged for providing four years of a CONACYT fellowship with grant number 340074 and
666908, to support the Ph.D. degree of AIMC and ARS respectively. We also thank CINVESTAV-IPN for the
funds to support researchers given to the corresponding authors and for the Mixed Funds for doctoral students
CINVESTAV-FOMIX 2018, granted to AIMC and ARS. e scientic results and conclusions, as well as any
views or opinions expressed herein, are those of the author and do not necessarily reect the views of NOAA or
the Department of Commerce.
Author Contributions
A.I.M.C. conceived the study with input from all authors. A.I.M.C. conducted all the data analysis, gures, and
supplementary materials. A.I.M.C. and A.R.S. wrote the manuscript. I.C., C.M.E., L.A.G., M.M., and J.E.A.G
assisted in writing the manuscript. All authors reviewed, edited and approved the manuscript.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
14
SCIENTIFIC REPORTS | (2019) 9:11013 | https://doi.org/10.1038/s41598-019-47307-0
www.nature.com/scientificreports
www.nature.com/scientificreports/
Additional Information
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-019-47307-0.
Competing Interests: e authors declare no competing interests.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© e Author(s) 2019
Content courtesy of Springer Nature, terms of use apply. Rights reserved
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
Available via license: CC BY 4.0
Content may be subject to copyright.
Content uploaded by C. Mark Eakin
Author content
All content in this area was uploaded by C. Mark Eakin on Aug 12, 2019
Content may be subject to copyright.