Antarctic Sea Ice

Identification

1.	Indicator Description

This indicator tracks changes in the February and September average extent of sea ice on the Southern
Ocean around Antarctica since 1979. The extent of area covered by Antarctic sea ice is considered a
useful indicator of global climate because a warmer climate will generally reduce the amount of sea ice
present, although climate can also affect sea ice in other more complex ways. This indicator also
provides a comparison to Arctic sea ice and a general sense of global sea ice conditions. The trends in
global sea ice extent are negative overall in every season and every month, which provides a direct
contribution toward decreasing the Earth's reflectivity (known as albedo) (Parkinson, 2014).

2.	Revision History

August 2016: Indicator published.

Data Sources

3.	Data Sources

This indicator is based on monthly average sea ice extent data provided by the National Snow and Ice
Data Center (NSIDC). NSIDC's data are derived from satellite imagery collected and processed by the
National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC).

4.	Data Availability

Users can access monthly map images, geographic information system (GlS)-compatible map files, and
gridded daily and monthly satellite data, along with corresponding metadata, at:
http://nsidc.org/data/seaice index/archives.html. From this page, users can also download monthly
extent and area data. From this page, select "FTP Directory" under the "Monthly Extent and
Concentration Images" heading, which will lead to a public FTP site

(ftp://sidads.colorado.edu/DATASETS/NOAA/GQ2135). To obtain the February or September monthly
data that were used in this indicator, select the "Feb" or "Sep" directory, find the file names that start
with "S" (Southern Hemisphere), and choose the "...area.txt" file with the data. To see a different
version of the graph in Figure 1 (plotting percent anomalies rather than square miles), return to the
parent directory and open the "...plot.png" image.

NSIDC's Sea Ice Index documentation page (http://nsidc.org/data/docs/noaa/g02135 seaice index)
describes how to download, read, and interpret the data. It also defines database fields and key
terminology. Gridded source data developed by NASA GSFC can be found at:
http://nsidc.org/data/nsidc-0051.html and: http://nsidc.org/data/nsidc-0081.html.

Technical Documentation: Antarctic Sea Ice

1


-------
Methodology

5.	Data Collection

This indicator is based on maps of sea ice extent in the ocean around Antarctica, which were developed
using brightness temperature imagery in the microwave wavelengths collected by satellites. Data from
October 1978 through June 1987 were collected using the Nimbus-7 Defense Meteorological Satellite
Program (DMSP) Scanning Multi Channel Microwave Radiometer (SMMR) instrument, and data since
July 1987 have been collected using a series of successor DMSP Special Sensor Microwave/lmager
(SSM/I) instruments. In 2008, the DMSP Special Sensor Microwave Imager and Sounder (SSMIS) replaced
the SSM/I as the source for sea ice products. These instruments can identify the presence of sea ice
because sea ice and open water have different passive microwave signatures.

The satellites that supply data for this indicator orbit the Earth continuously, collecting images that can
be used to generate daily maps of sea ice extent. They are able to map the Earth's surface with a
resolution of 25 kilometers. The resultant maps have a nominal pixel area of 625 square kilometers.
Because of the curved map projection, however, actual pixel sizes range from 382 to 664 square
kilometers.

The satellites that collect the data cover most of the Antarctic region in their orbital paths; however, the
sensors cannot collect data from a circular area immediately surrounding the South Pole due to orbit
inclination. A similar spatial gap requires correction for the Arctic Sea Ice indicator, but it does not affect
the Antarctic Sea Ice indicator, where the "pole hole" is entirely over land.

For documentation of passive microwave satellite data collection methods, see the summary and
citations at: www.nsidc.org/data/docs/noaa/g02135 seaice index.

6.	Indicator Derivation

Satellite data are used to develop daily ice extent and concentration maps using an algorithm developed
by NASA. Data are evaluated within grid cells on the map. Image processing includes quality control
features such as two weather filters based on brightness temperature ratios to screen out false positives
over open water, an ocean mask to eliminate any remaining sea ice in regions where sea ice is not
expected, and a coastal filter to eliminate most false positives associated with mixed land/ocean grid
cells.

From each daily map, analysts calculate the total "extent" and "area" covered by ice. These terms are
defined differently as a result of how they address those portions of the ocean that are partially but not
completely frozen:

• Extent is the total area covered by all pixels on the map that have at least 15-percent ice

concentration, which means at least 15 percent of the ocean surface within that pixel is frozen
over. The 15-percent concentration cutoff for extent is based on validation studies that showed
that a 15-percent threshold provided the best approximation of the "true" ice edge and the
lowest bias. In practice, much of the area covered by sea ice exceeds the 15-percent threshold,
so using a higher cutoff (e.g., 20 or 30 percent) would yield different totals but similar overall
trends (for example, see the Arctic analysis by Parkinson et al., 1999).

Technical Documentation: Antarctic Sea Ice

2


-------
• Area represents the actual surface area covered by ice. If a pixel's area were 600 square

kilometers and its ice concentration were 75 percent, then the ice area for that pixel would be
450 square kilometers. At any point in time, total ice area will always be less than total ice
extent.

EPA's indicator addresses extent rather than area. Both of these measurements are valid ways to look at
trends in sea ice, but in this case, EPA chose to look at the time series for extent because it is more
complete than the time series for area. In addition, extent is consistent with the Arctic Sea Ice indicator,
where "pole hole" limitations made it necessary to focus on extent rather than area.

NASA's processing algorithm includes steps to deal with occasional days with data gaps due to satellite
or sensor outages. These days were removed from the time series and replaced with interpolated values
based on the total extent of ice on the surrounding days.

From daily maps and extent totals, NSIDC calculated monthly average extent in square kilometers. EPA
converted these values to square miles to make the results accessible to a wider audience. By relying on
monthly averages, this indicator smooths out some of the variability inherent in daily measurements.

Figure 1 shows trends in February and September average sea ice extent. February is when Antarctic sea
ice typically reaches its annual minimum, after melting during the summer months. By looking at the
month with the smallest extent of sea ice, this indicator focuses attention on the time of year when
limiting conditions would most affect wildlife in the Antarctic region. Antarctic sea ice typically reaches
its annual maximum in late September or early October, after cold winter months freeze new ice.
September has the highest monthly average extent. Presenting the month with the greatest extent of
sea ice highlights the extent to which the Antarctic region recovers melted sea ice.

This indicator does not attempt to estimate values from before the onset of regular satellite mapping in
October 1978 (which makes 1979 the first year with February and September data for this indicator). It
also does not attempt to project data into the future.

For documentation of the NASA Team algorithm used to process the data, see Cavalieri et al. (1984) and:
http://nsidc.org/data/nsidc-0051.html. For more details about NSIDC methods, see the Sea Ice Index
documentation and related citations at: http://nsidc.org/data/docs/noaa/g02135 seaice index.

Other months of the year were considered for this indicator, but EPA chose to focus on February and
September, which represent the annual minimum and maximum extent of sea ice. Other months of the
year have similar patterns, as illustrated by Figure TD-1, which shows mean values for all months based
on the same NSIDC data source.

Technical Documentation: Antarctic Sea Ice

3


-------
Figure TD-1. Antarctic Sea Ice Extent for Each Month, 1979-2015/2016

o

1975 1980 1985 1990 1995 2000 2005 2010 2015 2020

Year

	September

	October

	August

—July

	November

	June

	December

	May

	April

	January

	March

—February

Data source: NSIDC: http://nsidc.orQ/data/seaice index/archives.html. Accessed July 2016.

7. Quality Assurance and Quality Control

Image processing includes a variety of quality assurance and quality control (QA/QC) procedures,
including steps to screen out false positives (i.e., ice is detected where it is not actually present). These
procedures are described in NSIDCs online documentation at:

http://nsidc.org/data/docs/noaa/g02135 seaice index as well as in some of the references cited
therein.

NSIDC Antarctic sea ice data have three levels of processing for QC. NSIDCs most recent data come from
the Near Real-Time SSM/I Polar Gridded Sea Ice Concentrations (NRTSI) data set. NRTSI data go through
a first level of calibration and QC to produce a preliminary data product. The final data are processed by
NASA's GSFC, which uses a similar process but applies a higher level of QC. Switching from NRTSI to
GSFC data can result in slight changes in the total extent values—on the order of 50,000 square
kilometers or less for total sea ice extent.

GSFC processing requires several months of lag time. At the time EPA published this report, the GSFC
data for 2016 had not yet been finalized.

Technical Documentation: Antarctic Sea Ice

4


-------
Analysis

8.	Comparability Over Time and Space

This indicator is based on data collection methods and processing algorithms that have been applied
consistently over time and space. NASA's satellites cover the entire area of interest.

9.	Data Limitations

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

1.	Variations in sea ice are not entirely due to changes in atmospheric or ocean temperature.

Other conditions, such as fluctuations in oceanic and atmospheric circulation, precipitation
change, and natural annual and decadal variability, can also affect the extent of sea ice. These
complex non-temperature factors are thought to exert a more significant influence on sea ice in
the Antarctic region than they do in the Arctic (IPCC, 2013).

2.	Many factors can diminish the accuracy of satellite mapping of sea ice. Although satellite
instruments and processing algorithms have improved somewhat over time, applying these new
methods to established data sets can lead to trade-offs in terms of reprocessing needs and
compatibility of older data. Hence, this indicator does not use the highest-resolution imagery or
the newest algorithms. Trends are still accurate, but should be taken as a general representation
of trends in sea ice extent, not an exact accounting.

3.	As described in Section 6, the threshold used to determine extent—15-percent ice cover within
a given pixel—represents an arbitrary cutoff without a particular scientific significance.
Nonetheless, studies have found that choosing a different threshold would result in similar
overall trends. Thus, the most important part of Figure 1 is not the absolute extent reported for
any given year, but the size and shape of the trend over time.

10.	Sources of Uncertainty

NSIDC has calculated standard deviations along with each monthly ice concentration average. NSIDC's
Sea Ice Index documentation (http://nsidc.org/data/docs/noaa/g02135 seaice index) describes several
analyses that have examined the accuracy and uncertainty of passive microwave imagery and the NASA
Team algorithm used to create this indicator. For example, a 1991 analysis estimated that ice
concentrations measured by passive microwave imagery are accurate to within 5 to 9 percent,
depending on the ice being imaged. Another study suggested that the NASA Team algorithm
underestimates ice extent by 4 percent in the winter and more in summer months. A third study that
compared the NASA Team algorithm with new higher-resolution data found that the NASA Team
algorithm underestimates ice extent by an average of 10 percent. For more details and study citations,
see: http://nsidc.org/data/docs/noaa/g02135 seaice index. Certain types of ice conditions can lead to
larger errors, particularly thin or melting ice. For example, a melt pond on an ice floe might be mapped
as open water. The instruments also can have difficulty distinguishing the interface between ice and
snow or a diffuse boundary between ice and open water. Using the February minimum minimizes many
of these effects because melt ponds and the ice surface become largely frozen by then. These errors do
not affect trends and relative changes from year to year.

Technical Documentation: Antarctic Sea Ice

5


-------
NSIDC has considered using a newer algorithm that would process the data with greater certainty, but
doing so would require extensive research and reprocessing, and data from the original instrument (pre-
1987) might not be compatible with some of the newer algorithms that have been proposed. Thus, for
the time being, this indicator uses the best available science to provide a multi-decadal representation
of trends in Antarctic sea ice extent. The overall trends shown in this indicator have been corroborated
by numerous other sources, and readers should feel confident that the indicator provides an accurate
overall depiction of trends in Antarctic sea ice overtime.

11.	Sources of Variability

Many factors contribute to variability in this indicator. In constructing the indicator, several choices have
been made to minimize the extent to which this variability affects the results. The apparent extent of
sea ice can vary widely from day to day, both due to real variability in ice extent (growth, melting, and
movement of ice at the edge of the ice pack) and due to ephemeral effects such as weather, clouds and
water vapor, melt on the ice surface, and changes in the character of the snow and ice surface. The
intensity of the Southern Annular Mode (a specific pattern of variability in atmospheric circulation) may
also have a year-to-year impact on Antarctic sea ice. Certain conditions could either promote or hinder
the northward drift of ice into warmer waters that would speed melting.

According to NSIDC's documentation at: http://nsidc.org/data/docs/noaa/g02135 seaice index, extent
is a more reliable variable than ice concentration or area. The weather and surface effects described
above can substantially impact estimates of ice concentration, particularly near the edge of the ice pack.
Extent is a more stable variable because it simply registers the presence of at least a certain percentage
of sea ice in a grid cell (15 percent). For example, if a particular pixel has an ice concentration of 50
percent, outside factors could cause the satellite to measure the concentration very differently, but as
long as the result is still greater than the percent threshold, this pixel will be correctly accounted for in
the total "extent." Monthly averages also help to reduce some of the day-to-day "noise" inherent in sea
ice measurements.

12.	Statistical/Trend Analysis

EPA used ordinary least-squares linear regression to identify trends in February and September ice
extent to support statements in the Key Points. Over the full period shown in Figure 1, February extent
increased at a rate of 0.0051 million square miles per year (p = 0.025) and September extent increased
at a rate of 0.0082 million square miles per year (p < 0.001). Thus, the results are significant to a 95-
percent level. Other publications have also performed linear regressions on these data and reported
statistically significant increases. For example, Parkinson and Cavalieri (2012) reported increases in
Southern Hemisphere sea ice extent that were statistically significant to a 99-percent level for winter
(July-September), spring (October-December), and annual averages (after adjusting for the seasonal
cycle), and significant to a 95-percent level for the fall (April-June).

References

Cavalieri, D.J., P. Gloersen, and W.J. Campbell. 1984. Determination of sea ice parameters with the
NIMBUS-7 SMMR. J. Geophys. Res. 89(D4):5355-5369.

Technical Documentation: Antarctic Sea Ice

6


-------
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.

Parkinson, C.L., D.J. Cavalieri, P. Gloersen, H.J. Zwally, and J.C. Comiso. 1999. Arctic sea ice extents,
areas, and trends, 1978-1996. J. Geophys. Res. 104(C9):20,837-20,856.

Parkinson, C.L., and D.J. Cavalieri. 2012. Antarctic sea ice variability and trends, 1979-2010. Cryosphere
6:871-880.

Parkinson, C.L. 2014. Global sea ice coverage from satellite data: Annual cycle and 35-yr trends. J.
Climate 27(24):9377.

Technical Documentation: Antarctic Sea Ice

7


-------