Climate Change Indicators in the United States Technical Documentation Sea Surface Temperature

Sea Surface Temperature

Identification

1. Indicator Description

This indicator describes global trends in sea surface temperature (SST) from 1880 to 2015. SST is a key
indicator related to climate change because it describes conditions at the boundary between the
atmosphere and the oceans, which is where the transfer of energy between the atmosphere and oceans
takes place. As the oceans absorb more heat from the atmosphere, SST is expected to increase. Changes
in SST can affect circulation patterns and ecosystems in the ocean, and they can also influence global
climate through the transfer of energy back to the atmosphere.

Components of this indicator include:

Global average SST from 1880 to 2015 (Figure 1)

A global map showing variations in SST change from 1901 to 2015 (Figure 2)

2. Revision History

April 2010: Indicator published.

December 2012: Updated indicator with data through 2011.

August 2013: Updated indicator on EPA's website with data through 2012.

May 2014: Updated Figure 1 with data through 2013. Added Figure 2 to show spatial patterns.

June 2015: Updated indicator on EPA's website with data through 2014.

August 2016: Updated indicator with data through 2015.

Data Sources

3. Data Sources

Figure 1 is based on the Extended Reconstructed Sea Surface Temperature (ERSST) analysis developed
by the National Oceanic and Atmospheric Administration's (NOAA's) National Centers for Environmental
Information (NCEI). The reconstruction model used here is ERSST version 4 (ERSST.v4), which covers the
years 1880 to 2015 and was described in Huang et al. (2015), Liu et al. (2015), and Huang et al. (2016).
Figure 2 has been adapted from a map in the Intergovernmental Panel on Climate Change's (IPCC's) Fifth
Assessment Report (IPCC, 2013). The original map appears in IPCC's Working Group I report as Figure
SPM.l and Figure 2.21, and it shows temperature change over land as well as over the ocean. The data
originally come from NCEI's NOAA Global Surface Temperature (NOAAGIobalTemp) data set, formerly
known as the Merged Land-Ocean Global Surface Temperature Analysis Dataset (MLOST), which
combines land-surface air temperature data with SST data from ERSST.v4.

Technical Documentation: Sea Surface Temperature

-------
ERSST.v4 is based on a large set of temperature measurements dating back to the 1800s. This data set is
called the International Comprehensive Ocean-Atmosphere Data Set (ICOADS), and it is compiled and
maintained by NOAA.

4. Data Availability

NCEI and the National Center for Atmospheric Research (NCAR) provide access to monthly and annual
SST and error data from the ERSST.v4 reconstruction in Figure 1, as well as a mapping utility that allows
the user to calculate average anomalies over time and space (NOAA, 2016a). EPA used global data (all
latitudes), which can be downloaded from:

ftp://ftp.ncdc.noaa.gov/pub/data/noaaglobaltemp/operational/timeseries.

Specifically, EPA used the ASCII text file "aravg.ann.ocean.90S.90N.v4.0.0.201601.asc", which includes
annual anomalies and error variance. A "readme" file in the same FTP directory explains how to use the
ASCII file. The ERSST.v4 reconstruction is based on in situ measurements, which are available online
through ICOADS (NOAA, 2016b).

Figure 2 is an updated version of a map that was published in IPCC (2013). Underlying gridded data and
documentation are available at: ftp://ftp.ncdc.noaa.gov/pub/data/noaaglobaltemp/operational.

Underlying ICOADS data are available at: http://icoads.noaa.gov.

Methodology

5. Data Collection

Both components of this indicator—global average SST since 1880 and the map of variations in SST
change since 1901—are based on in situ instrumental measurements of water temperature worldwide.
When paired with appropriate screening criteria and bias correction algorithms, in situ records provide a
reliable long-term record of temperature. The long-term sampling was not based on a scientific sampling
design, but was gathered by "ships of opportunity" and other ad hoc records. Records were particularly
sparse or problematic before 1900 and during the two World Wars. Since about 1955, in situ sampling
has become more systematic and measurement methods have continued to improve. SST observations
from drifting and moored buoys were first used in the late 1970s. Buoy observations became more
plentiful following the start of the Tropical Ocean Global Atmosphere (TOGA) program in 1985.

Locations have been selected to fill in data gaps where ship observations are sparse.

A summary of the relative availability, coverage, accuracy, and biases of the different measurement
methods is provided by Reynolds et al. (2002). Sampling and analytical procedures are documented in
several publications that can be accessed online. NOAA has documented the measurement, compilation,
quality assurance, editing, and analysis of the underlying ICOADS sea surface data set at:
http://icoads.noaa.gov/publications.html.

Although SST can also be interpreted from satellite imagery, ERSST.v4 does not include satellite data. In
the original update from ERSST v2 to v3, satellite data were added to the analysis. However, ERSST.v4
does not include satellite data because the addition of satellite data left a residual cold bias, even
though the satellite data were corrected with respect to the in situ data. The bias was strongest in the

Technical Documentation: Sea Surface Temperature

-------
middle and high latitude Southern Hemisphere where in situ data were sparse. The residual bias led to a
modest decrease in the global warming trend and modified global annual temperature rankings.

6. Indicator Derivation

Figure 1. Average Global Sea Surface Temperature, 1880-2015

This figure is based on the ERSST, a reconstruction of historical SST using in situ data. The reconstruction
methodology has undergone several stages of development and refinement. This figure is based on the
most recent data release, version 4 (ERSST.v4).

This reconstruction involves filtering and blending data sets that use alternative measurement methods
and include redundancies in space and time. Because of these redundancies, this reconstruction is able
to fill spatial and temporal data gaps and correct for biases in the different measurement techniques
(e.g., uninsulated canvas buckets, intakes near warm engines, uneven spatial coverage). Locations have
been combined to report a single global value, based on scientifically valid techniques for averaging over
areas. Specifically, data have been averaged over 5-by-5-degree grid cells as part of the
NOAAGIobalTemp data product (www.ncdc.noaa.gov/data-access/marineocean-data/noaa-global-
surface-temperature-noaaglobaltemp). Daily and monthly records have been averaged to find annual
anomalies. Thus, the combined set of measurements is stronger than any single set. Fundamental ERSST
reconstruction methods are documented in more detail by Smith et al. (2008). Smith and Reynolds
(2005) discuss and analyze the similarities and differences between various reconstructions, showing
that the results are generally consistent. For example, the long-term average change obtained by this
method is very similar to those of the "unanalyzed" measurements and reconstructions discussed by
Rayner et al. (2003).

This figure shows the extended reconstructed data as anomalies, or differences, from a baseline
"climate normal." In this case, the climate normal was defined to be the average SST from 1971 to 2000.
No attempt was made to project data beyond the period during which measurements were collected.

Additional information on the compilation, data screening, reconstruction, and error analysis of the
reconstructed SST data can be found at: www.ncdc.noaa.gov/data-access/marineocean-data/extended-
reconstructed-sea-surface-temperature-ersst.

Figure 2. Change in Sea Surface Temperature, 1901-2015

This map is based on gridded data from NOAAGIobalTemp, which in turn draws SST data from ERSST.v4.
ERSST's analytical methods are described above for Figure 1.

EPA replicated and updated the map in IPCC (2013) by calculating trends for each grid cell using the
same Interactive Data Language (IDL) code that the authors of IPCC (2013) used. A long-term trend was
calculated for each grid cell using linear regression. Trends have been calculated only for those cells with
more than 70 percent complete records and more than 20 percent data availability during the first and
last 10 percent of years (i.e., the first and last 11 years). The slope of each grid cell's trend (i.e., the rate
of change per year) was multiplied by the number of years in the period to derive an estimate of total
change. Parts of the ocean that are blank on the map did not meet these data availability thresholds.
Black plus signs (+) indicate grid cells where the long-term trend is significant to a 90-percent confidence
level.

Technical Documentation: Sea Surface Temperature

-------
EPA displayed only the ocean pixels on the map (no land-based data) because this indicator focuses on
SST. EPA also converted the results from Celsius to Fahrenheit.

7. Quality Assurance and Quality Control

Thorough documentation of quality assurance and quality control (QA/QC) methods and results is
available in the technical references for ERSST.v4 (www.ncdc.noaa.gov/data-access/marineocean-
data/extended-reconstructed-sea-surface-temperature-ersst).

Analysis

8. Comparability Over Time and Space

Presenting the data at a global and annual scale reduces the uncertainty and variability inherent in SST
measurements, and therefore the overall reconstruction in Figure 1 is considered to be a good
representation of global SST. This data set covers the Earth's oceans with sufficient frequency and
resolution to ensure that overall averages are not inappropriately distorted by singular events or missing
data due to sparse in situ measurements or cloud cover. The confidence interval shows the degree of
accuracy associated with the estimates over time and suggests that later measurements may be used
with greater confidence than pre-20th century estimates.

Figure 2 is based on several data products that have been carefully assembled to maximize consistency
over time and space. Areas with insufficient data for calculating trends have been excluded from the
map.

Continuous improvement and greater spatial resolution can be expected in the coming years as
historical data are updated. For example, there is a known bias during the World War II years (1941-
1945), when almost all measurements were collected by U.S. Navy ships that recorded ocean intake
temperatures, which can give warmer numbers than the techniques used in other years. Future efforts
will adjust the data more suitably to account for this bias.

Researchers Smith and Reynolds (2005) have compared ERSST with other similar reconstructions using
alternative methods. These comparisons yield consistent results, albeit with narrower uncertainty
estimates. Hence, the graph presented in Figure 1 may be more conservative than would be the case
had alternative methods been employed.

9. Data Limitations

Factors that may impact the confidence, application, or conclusions drawn from this indicator are as
follows:

1. The 95-percent confidence interval in Figure 1 is wider than in other methods for long-term
reconstructions; in mean SSTs, this interval tends to dampen anomalies.

2. The geographic resolution of Figure 1 is coarse for ecosystem analyses, but reflects long-term
and global changes as well as shorter-term variability.

Technical Documentation: Sea Surface Temperature

-------
3. The reconstruction methods used to create both components of this indicator removed most
random noise in the data; however, the anomalies are also dampened when and where data
were too sparse for a reliable reconstruction. The 95-percent confidence interval in Figure 1
reflects this dampening effect and uncertainty caused by possible biases in the observations.

4. Data screening results in loss of multiple observations at latitudes higher than 60 degrees north
or south. Effects of screening at high latitudes are minimal in the context of the global average;
the main effect is to lessen anomalies and widen confidence intervals. This screening does
create gaps in the Figure 2 map, however.

10. Sources of Uncertainty

The ERSST.v4 model has largely corrected for measurement error, but some uncertainty still exists.
Contributing factors include variations in sampling methodology by era as well as geographic region, and
instrument error from both buoys as well as ships.

The ERSST.v4 global reconstruction (Figure 1) includes an error variance for each year, which is
associated with the biases and errors in the measurements and treatments of the data. NOAA has
separated this variance into three components: high-frequency error, low-frequency error, and bias
error. For this indicator, the total variance was used to calculate a 95-percent confidence interval (see
Figure 1) so that the user can understand the impact of uncertainty on any conclusions that might be
drawn from the time series. For each year, the square root of the error variance (the standard error) was
multiplied by 1.96, and this value was added to or subtracted from the reported anomaly to define the
upper and lower confidence bounds, respectively.

Processing principles and procedures, including error estimates, for the gridded NOAAGIobalTemp data
set (as shown in Figure 2) have been described in Smith et al. (2008). Uncertainty measurements are
also available for some of the underlying data. For example, several articles have been published about
uncertainties in ICOADS in situ data; these publications are available from:

http://noc.ac.uk/publication/nl71590. See Box 2.1 in IPCC (2013) for additional discussion about the
challenge of characterizing uncertainty in long-term climatic data sets.

11. Sources of Variability

SST varies seasonally, but Figure 1 has removed the seasonal signal by calculating annual averages.
Temperatures can also vary as a result of inter-annual climate patterns, such as the El Nino-Southern
Oscillation. Figure 2 shows how patterns in SST vary regionally.

12. Statistical/Trend Analysis

Figure 1 shows a 95-percent confidence interval that has been computed for each annual anomaly.
Analysis by Smith et al. (2008) confirms that the increasing trend apparent from Figure 1 over the 20th
century is statistically significant. EPA reached the same conclusion based on least-squares linear
regression of the data, finding slopes of +0.010°F/year (1880-2015) and +0.013°F/year (1901-2015),
both with p < 0.0001 (highly significant). Figure 2 shows long-term linear trends for individual grid cells
on the map, and "+" symbols indicate cells where these trends are significant at a 90-percent level based
on least-squares linear regression—an approach that is consistent with the original IPCC source map
(IPCC, 2013).

Technical Documentation: Sea Surface Temperature

-------
References

Huang, B., V.F. Banzon, E. Freeman, J. Lawrimore, W. Liu, T.C. Peterson, T.M. Smith, P.W. Thorne, S.D.
Woodruff, and H. Zhang. 2015. Extended Reconstructed Sea Surface Temperature version 4 (ERSST.v4):
Part I: Upgrades and intercomparisons. J. Climate 28(3):911-930.

Huang, B., P.W. Thorne, T.M. Smith, W. Liu, J. Lawrimore, V.F. Banzon, H. Zhang, T.C. Peterson, and M.
Menne. 2016. Further exploring and quantifying uncertainties for Extended Reconstructed Sea Surface
Temperature (ERSST) version 4 (v4). J. Climate 29(9):3119-3142.

IPCC (Intergovernmental Panel on Climate Change). 2013. Climate change 2013: The physical science
basis. Working Group I contribution to the IPCC Fifth Assessment Report. Cambridge, United Kingdom:
Cambridge University Press, www.ipcc.ch/report/ar5/wgl.

Liu, W., B. Huang, P.W. Thorne, V.F. Banzon, H. Zhang, E. Freeman, J. Lawrimore, T.C. Peterson, T.M.
Smith, and S.D. Woodruff. 2015. Extended Reconstructed Sea Surface Temperature version 4 (ERSST.v4):
Part II: Parametric and structural uncertainty estimations. J. Climate 28(3):931-951.

NOAA (National Oceanic and Atmospheric Administration). 2016a. Extended reconstructed sea surface
temperature (ERSST.v4). National Centers for Environmental Information. Accessed March 2016.
www.ncdc.noaa.gov/data-access/marineocean-data/extended-reconstructed-sea-surface-temperature-
ersst.

NOAA (National Oceanic and Atmospheric Administration). 2016b. International comprehensive ocean-
atmosphere data sets (ICOADS). Accessed March 2016. http://icoads.noaa.gov.

Rayner, N.A., D.E. Parker, E.B. Horton, C.K. Folland, L.V. Alexander, D.P. Rowell, E.C. Kent, and A. Kaplan.
2003. Global analyses of sea surface temperature, sea ice, and night marine air temperature since the
late nineteenth century. J. Geophys. Res. 108:4407.

Smith, T.M., and R.W. Reynolds. 2005. A global merged land-air-sea surface temperature reconstruction
based on historical observations (1880-1997). J. Climate 18(12):2021-2036.
www.ncdc.noaa.gov/monitoring-references/docs/smith-revnolds-2005.pdf.

Smith, T.M., R.W. Reynolds, T.C. Peterson, and J. Lawrimore. 2008. Improvements to NOAA's historical
merged land-ocean surface temperature analysis (1880-2006). J. Climate 21(10):2283-2296.
www.ncdc.noaa.gov/sites/default/files/attachments/lmprovements-NQAAs-Historical-Merged-land-
Qcean-Temp-Analvsis-1880-2006 O.pdf.

Technical Documentation: Sea Surface Temperature

6

-------