Post on 08-Jul-2020
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
21Volume 7 No 1 February 2014 Journal of Operational Oceanography
Reef Temp Nex t Generation: A new operational system for monitoring
reef thermal stress
INTRODUCTION
Coral bleaching has been observed sporadically
since 1982 on the Great Barrier Reef (GBR), with
mass bleaching events occurring in 1997/1998
and 2002 and a more localised but severe event in
2006 affecting the southern GBR.1,2,3 Mass coral bleaching
events are caused by anomalously warm ocean tempera-
tures1,4,5, which can result in mortality when thermal stress is
prolonged and/or severe. These major bleaching events tend
to occur during the summer and early-autumn months when
ocean temperatures are warmest.1 Coral bleaching is expected
to increase in frequency and severity as ocean temperatures
increase under climate change1,6,7,8, which poses many chal-
lenges for reef management worldwide.9,10
Corals have a symbiotic relationship with dinoflagellates
known as zooxanthellae. Zooxanthellae provide corals with
photosynthetically derived carbohydrates and, in return, corals
provide zooxanthellae with protection and inorganic nutrients
from metabolic and respiratory processes. If ocean tempera-
tures increase beyond the thermal tolerance of the coral, the
expulsion of zooxanthellae from the coral host can occur. The
loss of zooxanthellae, and associated photosynthetic pigments,
causes the corals to appear pastel or white; hence the term
coral bleaching.11 Anomalously high ocean temperatures have
also been linked to coral disease outbreaks.1,12,13 Corals can
recover and be recolonised by zooxanthellae once stress levels
decline and more favourable conditions return.3 However,
the length of the recovery period depends on the severity of
bleaching. Extreme or prolonged stress levels can result in the
death of coral colonies. Reefs with widespread mortality can
take one to two decades to fully recover.14,15
The original ReefTemp system16 (hereafter RT1) was
developed as an experimental system in 2007 through a
collaboration between Australian agencies; Commonwealth
LA Garde BAppSci, BSc (Hons), Climate Prediction,
National Climate Centre, Bureau of Meteorology, Melbourne, Australia
CM Spillman PhD, BSc, BE (Hons), Centre for Australian Weather and
Climate Research (CAWCR), Bureau of Meteorology, Melbourne Australia
SF Heron PhD, BSc (Hons) Coral Reef Watch, National Oceanic and Atmospheric
Administration (NOAA) and Marine Geophysical Laboratory, Physics Department, School of
Engineering and Physical Sciences, James Cook University, Townsville, Australia
RJ Beeden MAppSci, BSc (Hons), Climate Change Group,
Great Barrier Reef Marine Park Authority (GBRMPA), Townsville, Australia
The expected increase in the frequency of mass coral bleaching under climate change
underlines the importance of thermal stress monitoring systems for coral reef manage-
ment. ReefTemp Next Generation (RTNG) is a sophisticated remote sensing application
designed to operationally monitor the ocean temperatures that can lead to coral bleach-
ing across the Great Barrier Reef. Products are derived from state-of-the-art satellite
data; and newly calculated climatologies and management thresholds for bleaching are
presented. RTNG is a key component of the Great Barrier Reef Marine Park Authority’s
Early Warning System, which informs management action and response strategies.
22
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Journal of Operational Oceanography Volume 7 No 1 February 2014
Scientific and Industrial Research Organisation (CSIRO),
the Bureau of Meteorology (the Bureau) and the Great
Barrier Reef Marine Park Authority (GBRMPA). Because
of its fine spatial resolution (0.018°; see Tables 1 and 2 for
system details and products), RT1 was designed to provide
GBRMPA managers with a reef-scale assessment of tem-
perature conditions across the vast 344,000km2 expanse of
the GBR Marine Park. Used on a daily-to-weekly basis, par-
ticularly during summer, RT1 was designed to monitor envi-
ronmental conditions such as sea surface temperature (SST)
and accumulated thermal stress across the GBR. Near real
time remote sensing at this scale improves the efficiency of
resource deployment and improves capacity to detect health
impacts. Maynard et al.17 suggest that providing early warn-
ing of conditions that may lead to coral bleaching events
will both open communications early with stakeholders and
the public, and assist planning and implementation of effi-
cient processes to assess and monitor ecological impacts. In
addition, RT1 has been employed to assess patterns of ther-
mal stress exposure and has also informed research on the
vulnerability and resilience of reefs under a range of climate
change scenarios.
For the past 5 years, RT1 has been employed by the
GBRMPA as part of its strategic framework to assess the
threat of coral bleaching.17 As part of this strategy, RT1 is
used in conjunction with SST outlooks18,19 generated from
the Bureau’s seasonal forecast model POAMA (Predictive
Ocean Atmosphere Model for Australia), and tools provided
by the NOAA Coral Reef Watch monitoring seasonal outlook
systems20, to form the GBRMPA’s Early Warning System.
Together these tools support management decisions and
response plans, ultimately leading to accurate and timely
information prior to coral bleaching events.17,21
However, RT1 is an experimental system, run in
research mode at CSIRO with no operational support. In
addition, products created by the RT1 system are based
on the Bureau’s legacy 14-day Advanced Very High
Resolution Radiometer (AVHRR) mosaic composite SST
product16,22, which is soon to be replaced by new daily
satellite SST products developed under the Integrated
Marine Observing System23 (IMOS). With funding from
the National Plan for Environmental Information (NPEI)
eReefs program, a new operational (ie, 24/7 supported)
version of RT1, known as ReefTemp Next Generation24
(RTNG), was designed, developed and launched at the
Bureau during late 2012. RTNG will replace RT1, aug-
menting the existing suite of products with a new set of
products for reef management.
RTNG produces SST and thermal stress products that
take advantage of the latest scientific advances and new
state-of-the-art, multi-sensor composite AVHRR satel-
lite SST products.23 RTNG uses IMOS real-time daily
night-only High Resolution Picture Transmission (HRPT)
AVHRR composite SST products for the Australian region23
at a resolution of 0.02° × 0.02°. Table 1 and Table 2 sum-
marise the main features of RT1 and RTNG coral bleach-
ing systems and their product suites respectively. A full
technical description of the RTNG system is also provided
by Garde et al.24 In developing RTNG, both the design and
delivery of thermal stress products were redesigned in close
consultation with GBRMPA, incorporating feedback to
ensure that the needs of reef management were addressed.
RTNG product suites are uploaded daily online, provid-
ing reef managers with the latest available data on reef
conditions (http://www.bom.gov.au/environment/activities/
reeftemp/reeftemp.shtml).
Table 1: Summaries of ReefTemp coral bleaching thermal stress monitoring systems: Legacy (RT1) and Next Generation (RTNG).
Please note: Degree Heating Days (DHD) are a measure of accumulated thermal stress and Mean Positive Summer Anomalies
(MPSA) are a measure of average stress severity.
LegacyÊReefTemp(2007)16
ReefTemp:ÊNextÊGeneration(2012)24
SST Input Product Bureau Legacy 14-day mosaic SST IMOS 1-day L3S SST
Grid Resolution 0.018° × 0.018° 0.02° × 0.02°
SST Dataset 14-day (day + night) composite 1-day night-time
Climatology Dataset31 23
31
Climatology Baseline 1993–2003 2002–2011
Climatology Calculation 65th percentile, day + night Monthly means, night only
Climatology Resolution 0.042° × 0.036° 0.02° × 0.02°
DHD/MPSA analysis period 1 December to 28/29 February 1 December to 31 March
Product Deployment Experimental at CSIROOperational at the Bureau
(24/7 support)
23
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Volume 7 No 1 February 2014 Journal of Operational Oceanography
LegacyÊReefTemp16
(2007)ReefTempÊNextÊGeneration24
(2012)
1-day IMOS SST
(IMOS climatology)
1-day IMOS SST
(CSIRO climatology)
14-day mosaic IMOS SST
(IMOS Climatology)
14-day mosaic IMOS SST
(CSIRO climatology)
REEF TEMP NEXT GENERATION (RTNG)
RTNG 1-day SST product: IMOS Satellite DataRTNG system details and products are summarised in Tables
1 and 2. Within IMOS, the Bureau produces high-resolution
satellite SST products for the Australian region. Raw data
measured by AVHRR sensors on-board National Oceanic
and Atmospheric Administration (NOAA) polar-orbiting sat-
ellites (NOAA-11, 12, 14, 15, 16, 17, 18, and 19) have been
broadcast via HRPT to several ground receiving stations
across Australia and Antarctica. A combination of available
data on any given day are then used to produce real-time
daily AVHRR skin SST data files beginning 2002.23,25 The
IMOS SST data were calibrated with in-situ measurements
to the ocean skin temperature (around 10-20 micron depth)
by first regressing the AVHRR brightness temperatures
against drifting buoy SST observations over the Australian
region, followed by conversion from buoy depths to the
cool skin by subtraction of 0.17°C.23 From the raw data,
each Level 2 pre-processed (L2P) geolocated single swath
dataset is re-mapped to a standard Level 3 un-collated grid-
ded product (L3U). A new Level 3 collated product (L3C) is
created from a combination of multiple swaths from a single
sensor. Finally a super-collated, multi-sensor, multi-swath
product (L3S) is produced. An overview of the data format
is provided by the Group for High-Resolution Sea Surface
Temperature (GHRSST) and can be viewed online at http://
imos.org.au. In addition, Casey et al.25 provides detailed
information regarding the content and format of these prod-
ucts. IMOS L3S grids do not contain any spatial or temporal
smoothing, nor use interpolation to fill in missing data within
the dataset. RTNG is based on the L3S 1-day night-only SST
data (Table 1). The night-only SST restriction was chosen
to avoid observations that are affected by diurnal warming
Table 2: Summary of ReefTemp coral bleaching thermal stress products: Legacy (RT1) and Next Generation (RTNG)
24
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Journal of Operational Oceanography Volume 7 No 1 February 2014
of the thin surface layer caused by solar heating, solar glare
effects26 and to therein represent the temperature variability
observed at the depths of reef organisms, particularly during
thermal stress events.27,28
An important stage in the creation of the L3S SST grids
is the identification of clouds. Clouds prevent an accurate
measure of the underlying ocean surface temperature, and
tend to lower the observed temperature and result in an
overall negative temperature bias at the affected grid cell.23
However, it is also crucial to ensure that pixels are not unnec-
essarily masked out during this process, as spatial coverage is
also important.23 The IMOS processing system uses a set of
brightness temperature, uniformity and two-channel thresh-
olds to identify cloud.23 At each IMOS SST grid location,
quality level (QL) data are also provided which is assigned
based on proximity to cloud (in terms of distance), satellite
zenith angle and day or night pass23. IMOS L3S metadata
states that QL values 0 through 5 represent no and bad data
(QL 0 and QL 1) then worst, low, acceptable and best quality
(QL 2 to QL 5), respectively.23 This enables users to select
the desired quality level dependent upon the needs of each
specific application.
Comparing brightness temperatures observed from the
Multifunctional Transport Satellites (MTSAT-2) platform
with the daily IMOS L3S SST indicates that the cloud
identification process is successful. Fig 1 shows MTSAT-2
brightness temperatures and IMOS L3S 1-day SST and QL
data over the GBR domain for two example dates. On 1
February 2011, extensive and persistent cloud cover was
observed (brightness temperature range 190-260K), associ-
ated with Severe Tropical Cyclone Yasi29 (Fig 1a), which
caused extensive damage to approximately 90,000km2
(~26%) of the GBR Marine Park.30 MTSAT-2 observed
brightness temperatures for 8 January 2012 shows that the
cloud cover was minimal across the domain (brightness
temperature range 280-290K; Fig 1d).
Fig 1: (a) Infrared brightness temperatures acquired by MTSAT-2 platform at 16:32 UTC, 1 February 2011, showing tropical
are illustrated as solid white lines in (a) and (d), and black lines in (b), (c), (e) and (f).
25
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Volume 7 No 1 February 2014 Journal of Operational Oceanography
In Fig 1b and 1e, white shading represents regions where
the IMOS SST could not be measured due to cloud cover. As
expected, due to cloud cover extent, data coverage is lower
for 1 February 2011 (Fig 1b), as compared with 8 January
2012 (Fig 1e). However, despite greater coverage, a large
number of SST grid cells have been flagged as low quality
(QL 3) for 8 January 2012, with best quality (QL 5) pixels
confined to the north-western and south-eastern regions of the
Coral Sea (Fig 1f). For this specific date, the likely explana-
tion for the swath of low quality data is high satellite sensor
viewing angle (see distinct swath edges in Fig 1f). If it had
been decided that only the most robust data points (QL 5
threshold only) be considered by the analysis, minimising the
potential for cloud contamination, it is clear that data cover-
age would be significantly sacrificed. For this reason and
after substantial examination of the SST and QL data, it was
decided that RTNG would use SST data with QL 3, 4 and 5.
Development of RTNG 14-day SST Mosaic As the daily IMOS L3S dataset does not incorporate any fill-
ing of spatio-temporal gaps, a potential issue was raised dur-
ing development that poor coverage could lead to an under-
estimate of the total accumulated thermal stress. This is of
major concern to GBRMPA regarding the timely monitoring
of thermal stress episodes. One potential solution to the issue
of inadequate spatial coverage was to generate a new SST
mosaic and provide derived gap filled products in addition
to the 1 day products (Table 2). This was based on the IMOS
L3S 1-day night-only SST data and followed the methodol-
ogy of the RT1 SST mosaic16,22; ie, filling each missing data
grid cell in the current day using the most recent daily SST
for that grid cell from up to 13 days prior.
Comparing the new RTNG 14-day SST mosaic product
(Fig 2a and c) with the IMOS SST L3S 1-day product (Fig
1b and e) for the same dates, it is evident that the mosaic
product provides markedly improved spatial coverage over
the GBR region. However, it is important to note that the
new SST field can be spatially inhomogeneous as a result
of the varying dates when the grid data were recorded. As a
result, during periods of warming and cooling, it is possible
that pixels within the mosaic product may contain values
that do not represent current conditions. This may result in
the appearance of unrealistic SST gradients at small scales.
Furthermore, persisting temperatures which trigger thermal
stress thresholds has the potential to over predict accumulated
thermal stress. However, from the perspective of reef man-
agement, it is preferable to over predict (and respond) than
to under predict (and not be aware of a bleaching event12,17).
To quantify the inhomogeneity of the 14-day SST mosaic,
a new Pixel Age product was developed to report the age of
Fig 2: RTNG 14-day (a) SST mosaic
and (b) mosaic Pixel Age grids for 1
February 2011. (c) and (d) as in (a)
and (b) for 8 January 2012.
26
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Journal of Operational Oceanography Volume 7 No 1 February 2014
filled data points. This product provides a means to assess
the temporal relevance of each SST value within the mosaic
product. The RTNG 14-day SST mosaic product created
for 1 February 2011 (Fig 2b) contains SST grids from up to
13 days prior, across most of the Coral Sea. In contrast, the
mosaic generated for 8 January 2012 (Fig 2d) contains more
contemporary data and more likely provides a better repre-
sentation of the SST conditions on the reported date.
To quantify the sparse spatial coverage of the IMOS L3S
1-day night-time only dataset, the persistence of missing data
was examined by Garde et al24 at each grid cell. Measured
in days, the longest consecutive period with no SST data
was found over the period 2002 to 2012, for each grid cell.
The study showed that isolated regions of the GBR have, at
times, lacked IMOS L3S SST data for more than 14 days.24
For some locations near Papua New Guinea and within the
inshore reef regions, the longest consecutive period with no
SST data was in excess of 28 days.24 The lack of SST data is
most likely due to persistent cloud cover24, with greater gap
counts measured in the northern Coral Sea. However, there
is a possibility that the absence of satellite retrievals or the
unavailability of data of sufficient quality during this period
could also be the cause. However, regardless of the cause,
a lack of SST data will affect the daily cumulative thermal
stress values.24 Grid cells where SST data are missing for
more than 14 consecutive days will also result in missing data
within the 14-day SST mosaic grid.
SST ClimatologyThe long-term average monthly SST conditions (the SST
climatology) are used to calculate SST anomalies and ther-
mal stress levels across the GBR. However, the values within
the SST climatology are sensitive to the period it is calcu-
lated, which can have an effect on reported stress levels. Two
options were considered to represent the long-term monthly
SST conditions for the GBR: (1) the existing CSIRO clima-
tology employed in RT1; and (2) a new climatology devel-
oped from the IMOS L3S 1-day SST data used in RTNG.
Reef managers requested that two separate suites of products
be generated using the two climatologies. This permits daily
comparison of stress metrics derived from two different base-
lines, and thus allowing management to test and update alert
levels for action given observed conditions.
The RT1 monthly climatology was derived by CSIRO
from composite 3-day day-and-night-time HRPT AVHRR
SST data from NOAA polar-orbiting satellites for the period
1993-2003.16,31 The SST data were calibrated to observa-
tions at depths around 20-30cm from global drifting buoys.
Exclusion of anomalous values caused by cloud and diurnal
surface warming was considered by taking the SST value at
the 65th percentile of the cumulative frequency distribution
within each group of monthly data.16
The new RTNG monthly climatology was derived from
night-time IMOS L3S 1-day SST data, by calculating the
average temperature for each month during the 10 year period
of January 2002 to December 2011 (the period for which data
were available). This climatology is calculated by averaging
the available data at an individual pixel for a particular month
across all years, regardless of the amount of missing data at
that location during the period. The data coverage per pixel
over the climatological period is presented in Garde et al.24
To ensure consistency with the real-time SST data, only SST
data of QL 3, 4 and 5 were included.
Comparing the CSIRO (1993-2003) and IMOS (2002-
2011) climatologies for December, January, February and
March over the GBR region (Fig 3a to d) shows that the
IMOS climatology is slightly warmer than the CSIRO clima-
tology across most of the GBR region during these months,
with the exception of the far northern and inshore regions of
the GBR. This finding is in spite of the 0.17°C cold offset
between the IMOS skin SST and CSIRO sub-skin SST and
the use of night-time only data for the IMOS climatology.
March shows the greatest difference with SSTs located in
the central Coral Sea in excess of 1.5°C warmer in the IMOS
climatology than the CSIRO climatology.
There are two explanations for the IMOS climatology
being warmer on average than the CSIRO climatology. First,
the climatologies are calculated for two different decades.
SSTs in the Australian region were the warmest on record
during the recent decade (2001 to 2010) as compared to the
1961–1990 average using NOAA Extended Reconstruction
SSTs.32 Research has shown that SSTs averaged along
Australia’s NE tropical coasts had a warming trend of 0.12°C/
decade over the 1950-2007 period.33 Secondly, the cloud
quality control methods differ between the IMOS and CSIRO
products. Griffin et al31 states that approximately 40% of thin
cloud at night is not detected by the Saunders and Kribel
cloud detection algorithm34, which was used in the CSIRO
dataset. This would result in a cooler SST for the CSIRO SST
climatology although the effect could be, at least partially,
offset by using the 65th percentile median SST. For the IMOS
SST and climatology, a new threshold test based on bright-
ness temperatures measured from thermal infrared channels
was found to be very effective at identifying cloud.23 Here it
is postulated that the observed climatology differences could
be due to these combined factors. However, it is worth high-
lighting that the two climatologies showed very little overall
difference (within ± 0.5°C) across the GBR Marine Park,
which is of importance to primary management users of the
RTNG product.
Given the difference between the two climatologies, both
the CSIRO (1993-2003; used in RT1) and IMOS (2002-2011;
new state-of-the-art SST dataset) climatologies are utilised to
provide two parallel suites of products in the RTNG operational
system (Table 2). There is ongoing scientific research as to
which climatological baseline is best used to indicate thermal
stress and thus predict coral bleaching14, particularly in the con-
text of global warming and coral adaptation. In future, products
based on the legacy CSIRO climatology may be phased out,
should the IMOS climatology be deemed as a more appropri-
ate baseline for coral stress by reef managers and stakeholders.
Thermal stress productsEach day during the extended summer period, 1 December
to 31 March, RTNG calculates the following thermal stress
metrics across the GBR region: Sea Surface Temperature
Anomaly (SSTA), Degree-Heating Days (DHD) and Mean
Positive Summer Anomaly (MPSA).16 As stated above, to
satisfy requests of reef managers and allow operational evalu-
ation of the new RTNG system, thermal stress products are
27
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Volume 7 No 1 February 2014 Journal of Operational Oceanography
calculated using each SST input (1-day, 14-day mosaic prod-
uct) and each climatology (IMOS, CSIRO) to create four par-
allel suites of products (1-day IMOS; 1-day CSIRO; 14-day
mosaic IMOS; 14-day mosaic CSIRO; Table 2).
SSTA fields are calculated by subtracting the correspond-
ing monthly climatology from the SST values, to indicate
the deviation of current SST conditions from the long-term
monthly mean.
SSTA SSTx y x y, ,= − climatology thismonth
Degree-Heating Days (DHD) is a measure of the accu-
mulated thermal stress, calculated progressively from 1
December through to 31 March each year (inclusive). A DHD
of 1°C-days is equivalent to an SSTA value of 1°C for one
day; ie, SST exceeds the corresponding monthly climatol-
ogy by 1°C for one day at a particular grid point (x, y). Only
positive SSTA values are accumulated during the period, viz.
DHD SSTA SSTAx y x yt Dec
t today
xst, , ,==
=
∑0
1
1where ,,y > 0°C
If no SST value is available at a particular pixel, there is
no accumulation to the relevant DHD value; ie, DHD value
remains the same as the previous day. RTNG also includes a
complementary product called DHD counts, which indicates the
number of days where the SSTA was greater than 0°C at each
grid cell; ie, days on which DHD was incremented (Table 2).
DHD values can represent a broad range of thermal stress;
for example, 3 days at 1°C above the local long-term aver-
age results in the same DHD value as 1 day at 3°C, with the
latter representing more intense stress to corals.16 For this
reason RTNG produces the Mean Positive Summer Anomaly
(MPSA35; formerly called Heating Rate in RT116). This is the
number of DHDs divided by the DHD count and it provides a
measure of the average severity of thermal stress:
MPSADHD
DHDx y
x y
count x y
,
,
,
=
The way the MPSA is calculated means that differences
between the 1-day and 14-day mosaic-based MPSA values
are unlikely to be as significant as in the DHD products.
Interpretation of thermal stressIn RT1, DHD (MPSA) values above 60 °C-days (1.7 °C)
and 100 °C-days (2.4 °C) were statistically linked to moder-
ate and severe bleaching observations on the Great Barrier
Reef, respectively.16 In RTNG, as the 14-day mosaic products
Fig 3: Difference between monthly
climatology values (IMOS–CSIRO)
in the RTNG GBR domain for (a)
December, (b) January, (c) February
and (d) March.
28
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Journal of Operational Oceanography Volume 7 No 1 February 2014
derived from 1-day IMOS SST data were created to closely
mimic the legacy SST mosaic product used in RT1, we
assume that the RT1 bleaching thresholds are thus applica-
ble to RTNG 14-day DHD and MPSA products. However,
the RTNG 1-day and the 14-day mosaic DHD products will
potentially show different accumulated totals, primarily aris-
ing from differences in spatial coverage, ie in locations where
data are missing in the 1-day IMOS SST product but are gap
filled in the 14-day mosaic. For example, if the SST at a loca-
tion is above the monthly climatology and is persisted for the
following four cloudy days in the 14-day mosaic, then the
accumulation of the 14-day DHD will be five times that of the
1-day DHD during that five day period. The DHD products
may differ even at the beginning of summer (1 December)
Fig 4: 1-day and 14-day mosaic product comparison for the 2002-2012 period covering the GBR domain. Degree-Heating Days
(DHD) using the (a) IMOS and (b) CSIRO climatology baseline. Mean Positive Summer Anomaly (MPSA) using the (c) IMOS
and (d) CSIRO climatology baseline. Dashed black line is the line of best fit (equation shown on each plot). Orange and red
dashed lines represent moderate and severe bleaching thresholds as calculated by Maynard et al16, respectively. See Table 3 for
new thresholds. Note the different 1-day DHD scale used in (a) and (b).
29
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Volume 7 No 1 February 2014 Journal of Operational Oceanography
if there is no corresponding 1-day SST field and the 14-day
mosaic product SST contains a persisted value from late
November.
This discrepancy was first noted by Garde et al24 who
showed that during the 2011-2012 Australian summer, 1-day
DHD values were significantly lower than the corresponding
14-day DHD values. Further to this, over the extended period
of 2002-2012, 1-day DHD values very rarely exceeded 50
°C-days anywhere in the GBR region, compared to 14-day
DHD values which were up to 200 °C-days, regardless of
climatology used (Fig 4a and b). Therefore the original
RT1 moderate (60 °C-days) and severe (100 °C-days) coral
bleaching thresholds16 are not appropriate when evaluating
thermal stress using the new IMOS 1-day SST based products
and thus must be revised.
In order to determine a new set of coral bleaching
thresholds appropriate for the use with the 1-day products,
the relationship between 1-day and corresponding 14-day
product (DHD, MPSA) values was determined (Fig 4).
DHD and MPSA values at the end of each summer
(31 March) for each year of 2002 to 2012 for all grid cells
were binned into increments of 1 °C-days and 0.1 °C for the
14-day (horizontal) DHD and MPSA products, respectively
(n = 11 years × number of wet cells). The average of the
associated 1-day value pair (vertical) for each bin was then
determined.
Bin pairs exhibited a linear relationship in the high-den-
sity regions of the distribution (ie, all but the highest values
of the DHD/MPSA). Linear regression of these bin pairs
was undertaken using only bins that contained >1% of the
population maximum value (defined as the bin with greatest
number of pairs). This ensured any outliers did not influence
the regression, resulting in very high correlations (r>0.99;
Table 3). Regression parameters indicate that the 1-day DHD
bleaching thresholds are approximately 20% of the 14-day
DHD thresholds (60 and 100 °C-days16), regardless of cli-
matology used (Fig 4a to b). New 1-day DHD thresholds
are 11.6 (12.2) and 18.9 (20.2) °C-days for moderate and
severe bleaching for IMOS (CSIRO) climatologies (Table
3). This suggests that on average, positive SSTA values were
persisted for five days in the 14-day mosaic product for every
day of data in the 1-day product. For the MPSA products,
the 1-day MPSA thresholds are roughly 80% of the 14-day
MPSA thresholds (1.7 and 2.4 °C for moderate and severe
bleaching, respectively16), again regardless of climatology
used (Fig 4c to d). New 1-day MPSA thresholds are 1.4 (1.5)
and 2.0 (2.1) °C for moderate and severe bleaching for IMOS
(CSIRO) climatologies (Table 3). The reduction in MPSA
values indicate that on average, higher SST values were
persisted for longer in the 14-day mosaic. This is reasonable
given that increases in SST will predominantly result from
closely spaced cloud-free days.36
Fig 5 and Fig 6 represent DHD and MPSA values as of the
end of the 2012 extended summer respectively. It can be seen
that when the colour schemes are appropriately scaled (as in
Fig 5b and e for DHD; and Fig 6b and e for MPSA) there is
improved thermal stress spatial pattern agreement between
the 1-day and 14-day products. However, it is evident that the
DHD products differ depending on the use of the IMOS (Fig
5a to c) or CSIRO climatology (Fig 5d to f). As discussed
previously, the IMOS climatology is generally warmer than
the CSIRO climatology for most of the GBR (Fig 3). It is
therefore not surprising that the DHD values were higher
when using the CSIRO climatology. The MPSA maps (Fig 6)
highlight small isolated regions of high MPSA values (orange
to red colours) within the GBR Marine Park off the Cape
York Peninsula, as well as south of Papua New Guinea. In
this example (extended 2012 summer), these values are likely
erroneous. Importantly, it is apparent that all four MPSA
products show similar spatial structure, identifying regions
of acute stress.
FUTURE WORKIn its current form, RTNG provides the foundation for
the development of additional coral products. GBRMPA
management have expressed interest in updating RTNG
products to assess the risk of coral diseases such as White
Syndrome and Black Band disease outbreaks, which can
be related in part to temperature stress.13 Another area of
Table 3: Newly derived RTNG thermal stress thresholds for 1-day products. RT1 thresholds defined as 60 and 100 °C-days
(DHD), and 1.7 and 2.4 °C (MPSA), for moderate and severe bleaching, respectively16. The new RTNG thresholds divided by the
original corresponding RT1 thresholds are denoted as percentages in brackets.
Degree-HeatingÊDays(DHD;Ê¡C-days)
MeanÊPositiveÊSummerÊAnomaly(MPSA;Ê¡C)
Climatology IMOS CSIRO IMOS CSIRO
Regression y = 0.183x + 0.572 y = 0.199x + 0.247 y = 0.799x + 0.086 y = 0.838x + 0.075
Correlation 0.993 0.998 0.996 0.999
Moderate Threshold
(RT1: 60 °C-days; 1.7 °C-days)
11.6
(19%)
12.2
(20%)1.4
(82%)
1.5
(88%)
Severe Threshold
(RT1: 100 °C-days; 2.4 °C-days)
18.9
(19%)
20.2
(20%)
2.0
(83%)
2.1
(88%)
30
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Journal of Operational Oceanography Volume 7 No 1 February 2014
future development would be to expand the RTNG system
to include additional geographic regions. Regions of inter-
est include northern and western coastal Australian waters,
including Ningaloo Reef, in addition to the western Pacific
Ocean.
Given the potential grid spatial inhomogeneity relating
to the use of a mosaic SST product, it is advised that further
research be targeted at developing a more accurate back-filled
daily SST product. Accuracy could be improved using spatio-
temporal interpolation techniques; eg, the SSTA from the
previous day could be damped (reducing over-prediction of
the thermal stress metrics). However, sensitivity and perform-
ance evaluations would be needed to measure the accuracy of
such techniques.
Finally, a verification study is needed to assess the skill
of the RTNG system in its identification of coral bleach-
ing events using coral health surveys and field data. This
would also assess the performance of the new 1-day DHD
and MPSA thresholds. It should be noted that the use of two
different climatologies will also impact DHD and MPSA,
though to a lesser extent, in that products utilising the warmer
IMOS climatology will accumulate more slowly than those
based on the cooler legacy CSIRO climatology. Quantifying
these impacts and implications for bleaching thresholds are
the subject of future work. These improvements are strongly
recommended as they will significantly increase both the
system functionality and the number of end user applications,
leading to further product validation and refinement.
Fig 5: GBR DHD calculated for the 2012 extended summer (1 December to 31 March inclusive) using IMOS climatology with
(a) IMOS L3S 1-day SST (b) IMOS L3S 1-day SST with rescaled colour scheme and (c) 14-day IMOS SST mosaic. (d), (e) and
(f) as in (a), (b) and (c) but using CSIRO climatology.
31
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Volume 7 No 1 February 2014 Journal of Operational Oceanography
CONCLUSIONRTNG is a sophisticated real-time, reef-scale remote sensing
system that monitors ocean temperatures and coral thermal
stress across the GBR. Now employing the latest in high
resolution gridded daily L3S multi-sensor composite satellite
SST data, RTNG presents and communicates daily thermal
stress across the GBR at a very high resolution. The RTNG
development during 2012 has resulted in an operationally
supported Bureau system with multiple enhancements and
new product suites arising from extensive consultation with
GBRMPA and RT1 users.
The use of new cutting edge data inputs and shifting
climatological periods in RTNG has required revised guide-
lines to interpretation of thermal stress products. New 1-day
DHD thresholds are approximately 20% of legacy RT1
DHD thresholds and have been defined as 11.6 (12.2) and
18.9 (20.2) °C-days for moderate and severe bleaching for
IMOS (CSIRO) climatologies. Corresponding new MPSA
thresholds are 1.4 (1.5) and 2.0 (2.1) °C for moderate and
severe bleaching for IMOS (CSIRO) climatologies. These
revised coral bleaching thresholds for RTNG 1-day DHD and
MPSA products provide reef management with the means of
better monitoring and communicating thermal stress events.
The RTNG aids reef managers in generating and actioning
appropriate strategic plans for minimising potential impacts
of coral bleaching on reef health. The development and
operational support for these types of systems is extremely
important, as the frequency and severity of coral bleaching is
expected to increase due to climate change.
ACKNOWLEDGMENTSFunding was provided by eReefs/NPEI. Data were
sourced from IMOS. IMOS is supported by the Australian
Fig 6: GBR MPSA calculated for 31 March 2012 using IMOS climatology with (a) IMOS L3S 1-day SST (b) IMOS L3S 1-day SST
with rescaled colour scheme and (c) 14-day IMOS SST mosaic. (d), (e) and (f) as in (a), (b) and (c) but using CSIRO climatology.
32
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Journal of Operational Oceanography Volume 7 No 1 February 2014
Government through the National Collaborative Research
Infrastructure Strategy and the Super Science initiative. Raw
HRPT AVHRR SST data was collected by the Australian
Bureau of Meteorology in collaboration with Australian
Institute of Marine Science (AIMS) and CSIRO Marine and
Atmospheric Research.
The authors would like to acknowledge Dr Oscar Alves,
Dr Debbie Hudson, Dr Helen Beggs, Aurel Griesser, Elaine
Miles, and Griff Young (Centre for Australian Weather
and Climate Research, Australian Bureau of Meteorology)
for their guidance and helpful discussion as well as Leon
Majewski, Christopher Griffin and George Kruger (Bureau
of Meteorology) for the preparation of IMOS SST data.
Thanks are also extended to GBRMPA for their support and
feedback during the RTNG development process. Thanks also
to the reviewers of this manuscript. The manuscript contents
are solely the opinions of the authors and do not constitute a
statement of policy, decision, or position on behalf of NOAA
or the U.S. Government.
REFERENCES1. Hoegh-Guldberg O. 1999. Climate change, coral
bleaching and the future of the world’s coral reefs. Marine
and Freshwater Research 50: 839–866.
2. Berkelmans R, De’ath G, Kininmonth S and Skirving
W. 2004. A comparison of the 1998 and 2002 coral bleaching
events on the Great Barrier Reef: spatial correlation, pat-
terns and predictions. Coral Reefs 23: 74–83.
3. Weeks SJ, Anthony KRN, Bakun A and Feldman GC.
2008. Improved predictions of coral bleaching using seasonal
baselines and higher spatial resolution. Limnology and
Oceanography 53: 1369–1375.
4. Brown BE. 1997. Coral bleaching: causes and conse-
quences. Coral Reefs 16: DOI: 10.1007/s003380050249.
5. Lesser MP. 2004. Experimental biology of coral reef
ecosystems. Journal of Experimental Marine Biology and
Ecology 300: 217–252.
6. Donner SD, Skiving W, Little CM, Oppenheimer M
and Hoegh-Guldberg O. 2005. Global assessment of coral
bleaching and required rates of adaptation under climate
change. Global Change Biology 11: 2251–2265.
7. Hennessy K, Fitzharris B, Bates BC, Harvey N,
Howden SM, Hughes L, Salinger J and Warrick R. 2007.
Australia and New Zealand. Climate Change 2007: Impacts,
Adaptation and Vulnerability. Contribution of Working Group
II to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change, Parry ML, Canziani OF, Palutikof
JP, van der Linden PJ and Hanson CE, Eds., Cambridge
University Press, Cambridge, UK, 507–540.
8. Frieler K, Meinshausen M, Golly A, Mengel M, Lebek
K, Donner SD and Hoegh-Guldberg O. 2012. Limiting global
warming to 2°C is unlikely to save most coral reefs. Nature
Climate Change 2(9): 1–6. DOI: 10.1038/nclimate1674.
9. Hoegh-Guldberg O, Mumby PJ, Hooten AJ, Steneck RS,
Greenfield P, Gomez E, Hardvell CD, Sale PF, Edwards AJ,
Caldeira K, Knowlton N, Eakin CM, Iglesias-Prieto R, Muthiga
N, Bradbury RH, Dubi A and Hatziolos ME. 2007. Coral
Reefs Under Rapid Climate Change and Ocean Acidification.
Science 318: 1737–1742, DOI: 10.1126/science.1152509.
10. Donner SD. 2009. Coping with commitment: projected
thermal stress on coral reefs under different future scenarios.
PLoS One 4(6): e5712. DOI: 10.1371/journal.pone.0005712.
11. Jones RJ, Hoegh-Guldberg O, Larkum AWD, and
Schreiber U. 1998. Temperature-induced bleaching of cor-
als begins with impairment of the CO2 fixation mechanism in
zooxanthellae. Plant, Cell and Environment 21(2): 1219–1230.
12. Beeden R, Maynard JA, Marshall PA, Heron SF,
and Willis BL. 2012. A framework for responding to coral
disease outbreaks that facilitates adaptive management.
Environmental Management 49(1): 1–13.
13. Heron SF, Maynard JA, Willis BL, Christensen T,
Harvell D, Vargas-Angel B, Beeden R, Sziklay J, Aeby G,
Franklin EC, Skirving W, Eakin M, Burgess T, Li J, Liu G,
Lucas E, Rauenzahn J and Strong A. 2012. Developments
in understanding relationships between environmental
conditions and coral disease. Proceedings of the 12th
International Coral Reef Symposium, Cairns, Australia,
9–13 July 2012. 9–13.
14. Baker A, Glynn PW, and Riegl B. 2008. Climate
change and coral reef bleaching: An ecological assessment
of long-term impacts, recovery trends and future outlook.
Estuarine, Coastal and Shelf Science 80(4): 435–471.
15. Wilkinson C. 2008. Status of coral reefs of the
world. Global Coral Reef Monitoring Network and Reef and
Rainforest Research Centre, Townsville, Australia, 296 p.
16. Maynard JA, Turner PJ, Anthony KRN, Baird
AH, Berkelmans R, Eakin CM, Johnson J, Marshall
PA, Packer GR, Rea A and Willis BL. 2008. ReefTemp:
An interactive monitoring system for coral bleaching
using high-resolution SST and improved stress predic-
tors. Geophysical Research Letters, 35(5): L05603, DOI:
10.1029/2007GL032175.
17. Maynard JA, Johnson JE, Marshall PA, Eakin CM,
Goby G, Schuttenberg H, Spillman CM. 2009. A strategic
framework for responding to coral bleaching events in a
changing climate. Environmental Management, 44(1): 1–11,
DOI: 10.1007/s00267-009-9295-7.
18. Spillman CM and Alves O. 2009. Dynamical seasonal
prediction of summer sea surface temperatures in the Great
Barrier Reef. Coral Reefs 28(1): 197–206.
19. Spillman CM, Alves O, and Hudson DA. 2012. Predicting
thermal stress for coral bleaching in the Great Barrier Reef
using a coupled ocean-atmosphere seasonal forecast model.
International Journal of Climatology. 33: 1001–1014
20. Liu G, Eakin CM, Rauenzahn JL, Christensen
TRL, Heron SF, Li J, Skirving W, Strong AE, Burgess T.
2012. NOAA Coral Reef Watch’s Decision Support System
for Coral Reef Management. Proceedings of the 12th
International Coral Reef Symposium, Cairns, Australia,
9–13 July 2012. pp5.
21. Spillman C, Heron S, Jury M, and Kenneth A.
2011. Climate Change and Carbon Threats to Coral Reefs:
National Meteorological and Ocean Services as Sentinels.
Bull. Amer. Meteor. Soc., 92, 1581–1586.
22. Rea A. 2004. Recent Improvements to the NOAA
AVHRR SST Product at the Australian Bureau of Meteorology.
See http://imos.org.au/srsdoc.html.
23. Paltoglou G, Beggs H and Majewski L. 2010. New
Australian High Resolution AVHRR SST Products from the
33
Reef Temp Nex t Generation: A new operational system for monitoring reef thermal stress
Volume 7 No 1 February 2014 Journal of Operational Oceanography
Integrated Marine Observing System. Extended Abstracts of
the 15th Australasian Remote Sensing and Photogrammetry
Conference, Alice Springs, 13–17 September.
24. Garde LA, Spillman CM, Majewski L, Griffin C,
Kruger G, and Beggs H. 2013. ReefTemp Next Generation.
Centre for Australian Weather and Climate Research
(CAWCR) Technical Report. pp 44.
25. Casey K, Donlon C, and the GHRSST Science Team.
2010. The recommended GHRSST Data Specification (GDS)
2.0, Revision 4, available from the GHRSST International
Project Office, 123 pp.
26. Gentemann CL, Donlon CJ, Stuart-Menteth A, and
Wentz FJ. 2003. Diurnal signals in satellite sea surface
temperature measurements, Geophysical Research Letters,
30(3): 1140 DOI: 10.1029/2002GL016291.
27. Donlon CJ, Minnett PJ, Gentemann C, Nightingale
TJ, Barton IJ, Ward B and Murray MJ. 2002. Toward
improved validation of satellite sea surface skin tempera-
ture measurements for climate research. Journal of Climate
15(4): 353–369.
28. Skirving WJ, Strong AE, Liu G, Arzayus F, Liu C,
Sapper J. 2006. Extreme events and perturbations of coastal
ecosystems: Sea surface temperature change and coral
bleaching. 9: 11–25.
29. Bureau of Meteorology. 2011. Severe Tropical
Cyclone Yasi [online]. http://www.bom.gov.au/cyclone/
history/yasi.shtml.
30. Great Barrier Reef Marine Park Authority. 2011.
Impacts of tropical cyclone Yasi on the Great Barrier Reef:
a report on the findings of a rapid ecological impact assess-
ment, July 2011, GBRMPA, Townsville, pp 21.
31. Griffin DA, Rathbone CE, Smith GP, Suber KD
and Turner PJ. 2004. A decade of SST satellite data.
Final report for the National Oceans Office, Contract
NOOC2003/020, 1, 8.
32. Bureau of Meteorology. 2011. Annual Climate
Summary 2010. January 2011. pp 20.
33. Lough JM. 2008. Shifting climate zones for Australia’s
tropical marine ecosystems. Geophysical Research Letters,
35(14): L14708 DOI: 10.1029/2008GL034634.
34. Saunders RW, and Kriebel KT. 1988. An improved
method for detecting clear sky and cloudy radiances from
AVHRR data. International Journal of Remote Sensing, 9(1):
123–150.
35. Maynard JA, Anthony KRN, Harvell CD, Burgman
MA, Beeden R, Sweatman H, Heron SF, Lamb JB and Willis
BL. 2011. Predicting outbreaks of climate-driven coral dis-
ease in the Great Barrier Reef. Coral Reefs. 30(2): DOI:
10.1007/s00338-010-0708-0.
36. Leahy SM, Kingsford MJ, Steinberg CR. 2013. Do
Clouds Save the Great Barrier Reef? Satellite Imagery
Elucidates the Cloud-SST Relationship at the Local
Scale. PLoS One 8(7): e70400. DOI: 10.1371/journal.
pone.0070400.