Fun without touching: satellite remote sensing of sea ice

Over the past 30 years polar-orbiting satellites have revolutionised our ability to monitor the vast and data-sparse 'Great White Hell' on climatically-relevant temporal and spatial scales. Over that time, they have given us great insight into the complex large-scale interactions of air, ocean and atmosphere in the Southern Ocean, and the variable role of sea ice as a habitat for microscopic phytoplankton to whales.

The sheer magnitude of Antarctic sea ice extent lends itself perfectly to measurement from space. Each year Antarctica's sea ice cover puts on the greatest geophysical show on Earth, expanding and contracting between an area of about 3 million km2 in February and about 19 million km2 in September–October. In so doing, it has a profound and immensely complex yet highly variable impact on high-latitude air-sea-ice physical and biogeochemical interaction processes, ocean water-mass modification and circulation, marine ecology and the global climate system. The seasonal sea ice cycle also has a profound effect on human activities.

Sea-ice concentration

Monthly mean images of sea ice concentration (%) from February (minimum sea ice extent) and September (maximum sea ice extent) 1996, derived from the Special Sensor Microwave Imager onboard the US Defence Meteorological Satellite Program satellites. With reliable data dating back to 1978, daily passive microwave images of sea ice concentration and extent have bcome a 'bread and butter' tool of Southern Ocean climate research. Image courtesy of the NASA Earth Observing System Distributed Active Archive Center, National Snow and Ice Data Centre, University of Colorado.

The magnitude and variability of these impacts depends on a number of characteristics of the ice cover, including its areal extent, concentration (degree of compaction), thickness distribution, motion/deformation, and snow cover thickness. On longer time scales, modeling studies suggest that global change may be readily detectable through changes in Antarctic sea-ice extent and thickness (or volume).

Fortunately, sea ice research is no longer limited to point measurements that are greatly limited in space and time. Over the past 30 years polar-orbiting satellites have revolutionised our ability to monitor the vast and data-sparse 'Great White Hell' on climatically-relevant temporal and spatial scales. Over that time, they have given us great insight into the complex large-scale interactions of air, ocean and atmosphere in the Southern Ocean, and the variable role of sea ice as a habitat for species from microscopic phytoplankton to whales. Only satellites can cost-effectively provide systematic, repetitive, continuous and reliable coverage for large-scale monitoring studies. With satellite sensor technology has come great advances in ground segment and computing technology. With the advent of the internet, this allows rapid access and dissemination of data.

The first Earth-observing satellites launched in the early 1960s gave us tantalising glimpses of Antarctic sea ice through breaks in the cloud cover, but it was not until the launch of a microwave instrument (the Electrically Scanning Microwave Radiometer, or ESMR) in 1972 onboard NASA's Nimbus-5 satellite that we gained the first overview of the incredible seasonal waxing and waning cycle of sea ice encircling Antarctica . These were also our first images showing significant interannual variability in areal sea ice extent, enabling routine extraction of another key geophysical parameter from the data – sea ice concentration.

By sensing the Earth's surface at centimetre-scale wavelengths through atmospheric 'windows', passive microwave radiometers have the key ability to penetrate both cloud cover and polar darkness. As a result, they have become a workhorse of sea-ice research. In this case, 'passive' describes the fact that the sensor detects radiation thermally-emitted from the surface, the intensity of which depends on both the temperature and the 'emissivity' of the radiating portion of that surface material. Emissivity is the ratio of the radiation emitted by a surface to the radiation emitted by a perfect (blackbody) radiator at the same temperature. Importantly, a large microwave-emissivity contrast exists between open ocean and sea ice, allowing the latter to be readily detected from space.

Subsequent sensors, such as the current Special Sensor Microwave-Imager, or SSM-I, added a multi-spectral capability, enabling more accurate measurements. Collecting data over a wide swath, such sensors offer complete coverage of the entire Antarctic sea ice cover once daily, albeit at a fairly coarse spatial resolution (pixels 12.5-25 km). Swath refers to the width of the ground track covered by the sensor, and spatial resolution is the smallest object or narrowest line that a sensor can detect.

Extending back 30 years, passive microwave data enable monitoring of trends in the sea ice cover, and have highlighted on a regional basis significant interannual variability in sea-ice extent. The data have also recently been used by studies relating anomalies in regional sea-ice concentration and extent to El Niño–Southern Oscillation events and the eastward propagation of the Antarctic Circumpolar Wave around the continent.

Time series of sea-ice concentration and extent data are critical for identifying possible decadal fluctuations resulting from significant changes in high-latitude oceanic and atmospheric circulation. This vitally important time series continues today with data from the Advanced Microwave Scanning Radiometer (AMSR-E) onboard NASA's Aqua satellite, launched in May 2002. With additional channels, this sensor will routinely provide additional key information on snow-ice interface temperature and snow-cover thickness.

Active microwave sensors
Active microwave sensors, which include synthetic aperture radars (SAR), radar altimeters and wind scatterometers, transmit a centimetre-wavelength signal and measure the strength of its return, or backscatter, from a target surface. This backscatter contains information about the nature of that target. The SAR is a high data rate sensor that allows us to observe the surface at a spatial resolution of tens of metres. Current SARs are aboard Radarsat-1 and Envisat, and collect data over a 400-500 km-wide swath – again uninterrupted by clouds and darkness.

Satellite image 1

Part of a Radarsat ScanSAR image of the Mertz Glacier tongue region, August 4 1999, showing new (frazil) ice formation in the Mertz Glacier polynya. Also prominent are grounded icebergs. This region is an important site for the formation of very dense and cold Antarctic Bottom Water, which is a key component of the global ocean circulation system. Copyright Radarsat International/NASA 1999.

Unlike their passive microwave cousins, SARs can resolve morphological features such as leads and floes, and provide information on the structure of the pack. As geolocation of SAR imagery can be relatively accurate, sequential images can be superimposed to determine the relative movement of sea ice in the intervening period. Maps can be constructed of sea-ice motion vectors using cross-correlation techniques to identify and track floes. (Note that sea-ice motion is also derived from passive microwave and visible-thermal IR images.)

Better understanding of large-scale ice motion is a key to enabling us to obtain a more accurate parameterisation of the relative roles of ice motion and interaction (dynamics) and thermodynamics in determining the thickness distribution of Antarctic sea ice. Rapid ice formation occurs in leads during divergent conditions, with pressure ridge building occurring during subsequent ice convergence.

SAR data are also being used to map the fast-ice cover around Antarctica's coastal margins, determine the effect of grounded icebergs on sea ice distributions, and estimate the extent of open water and thin ice types in the study of coastal polynyas (areas of open water within the sea ice cover). The latter is expected to result in more accurate estimates of ocean–atmosphere heat fluxes and sea-ice production rates – key climate-related processes.

Another important, though largely under-utilised, sea ice research tool is the radar scatterometer. Designed to measure wind speed and direction over open ocean, it also has the potential to deliver low resolution (25-50 km) but important hemispheric sea ice information on a daily basis. Products from these data include daily maps of ice-type (age) classification, extent, motion, and the timing and progression of regional sea ice melt. These data are currently research and development only, but show great potential.

Visible-thermal infrared radiometers
Microwave sensors cannot provide information on sea ice albedo and surface temperature. Albedo is defined as the ratio of electromagnetic (EM) radiation reflected from an object to the total amount incident upon it (for a given portion of the EM spectrum), and is an important climate variable. Measuring albedo and surface temperature is the realm of radiometers operating at visible to near-IR (about 0.4-1.4 microns) and thermal IR (about 3.0-12.0 microns) wavelengths respectively (1 micron = 1 x 10-6 m).

The medium-resolution radiometer is another workhorse, and offers excellent spatial coverage each time the satellite passes overhead. Examples are the NOAA Advanced Very-High Resolution Radiometer (AVHRR, resolution about 1-4 km, swath about 2000 km) and the new Moderate Resolution Imaging Spectrometer (MODIS, resolution 0.25-1.0 km, swath about 2,300 km) onboard NASA's Terra and Aqua satellites. Unfortunately from a sea ice perspective, these sensors are severely limited by cloud cover and, in the case of visible-near-infrared sensors, polar darkness. However, such is the regularity of coverage that gaps can usually be found in the cloud cover; and cloud cover is itself an important climate variable.

Another illustration of recent technological advances is the improved spectral resolution of MODIS operating in 36 discrete bands, compared to 5-6 relatively broad bands for the NOAA AVHRR. This increased sensitivity should potentially enable the detection of subtle differences in the surface signature related to ice thickness, for example.

Other visible to near-IR sensors are important because they can detect and monitor ocean colour related to the presence of phytoplankton in open water. Once calibrated, these data yield estimates of ocean primary production, and are providing insight into the patchy distribution of phytoplankton blooms in the Southern Ocean. Current ocean colour sensors are the MODIS and the SeaWiFS onboard the SeaStar satellite – all give excellent coverage, are wide-swath (>1000 km), and medium resolution (0.25-4.5 km).

Satellite image 3

A composite ocean colour image of the Southern Ocean produced from data from the SeaWiFs sensor onboard the SeaStar satellite, for the period September 1997 to August 2000. The colours relate to chlorophyll a concentrations in the upper ice-free ocean, with red being the highest concentration. Sea ice and ice sheets are masked out in white. Courtesy of NASA.

Another current trend is towards higher spatial and spectral resolution. The commercial QuickBird and IKONOS satellites can, for example, produce visible-near IR data at a an ultra-high spatial resolution of 0.61-1 m, albeit at the expense of good coverage (i.e. the swath is <15 km). Among today's other high-resolution sensors, the SPOT HRG can deliver data at a 2.5-20 m resolution (swath 60 km), while the Landsat Enhanced Thematic Mapper (ETM) has a resolution of 15-60 m and includes a night-time (thermal infrared) imaging capability (Figure 5). Big Brother can indeed be watching – on a cloud-free day, that is!!

Satellite Image 4

Landsat 7 Enhanced Thematic Mapper image of the Cape Adare region (image centered on 70.92°S, 171.62°E), December 12 1999 (about 21:00 UTC). The ETM simultaneously collects high-resolution data in 8 bands (0.45-12.5 microns) across a 180 x 183 km scene. Note that individual floe assemblages are resolved. Courtesy of US Geological Survey and Eosat International.

Modelling and operational uses
Satellite data are now of key importance in providing initial data to drive coupled ocean-ice-atmosphere models and also different data to validate the latter. Model simulations of future climate are highly sensitive to the manner in which the high latitudes are represented, and improved satellite data support improved model performance. Moreover, the resolutions of both passive microwave satellite data and sophisticated models have converged to the stage where assimilation of satellite data into operational sea-ice forecasting models is becoming a reality.

This exciting trend is aided by developments in the ground sector, including higher-performance computers, which have resulted in the near real-time availability of high-level geophysical products from many of these data.

Satellite remote sensing plays an important role in operational activities in the sea-ice zone. Vessels operating in Antarctic sea ice can now receive high-quality satellite data in real time. Polar-orbiting satellites also play a key role in transmitting data from remote sensor packages, such as automatic weather stations, drifting buoys or instruments attached to wildlife.

Conclusions
Satellite remote sensing, though a wonderful research tool, is not a panacea. Uncertainties remain in interpreting more complex data. Satellite data require calibrating and validating, ideally with dedicated surface measurements, a need driving next year's first voyage of the RV Aurora Australis. Sea ice thickness, a key climate parameter, can be inferred from satellite data or estimated by combining satellite observations with models, but its direct measurement from space remains an elusive 'holy grail'. Future missions are addressing this issue, including the sophisticated dual radar altimeters on board the European Space Agency's 2004 CryoSat. Developed to determine the (high) freeboard of multiyear sea ice in the Arctic, the accuracy of this instrument in determining ice thickness in the Southern Ocean remains to be seen.

An important 'take-home' message is that no one perfect, multi-purpose satellite sensor exists for sea-ice research. Due to technological trade-offs (for example, between spatial–temporal coverage and spatial resolution), each sensor possesses inherent strengths and weaknesses. A combination of data from different yet complementary sensors is a powerful research tool, particularly when added to contemporary surface observations and models. The imminent launch of new and improved sensors bodes well, not only for continuing key time series such as ice concentration and extent but also providing improved and additional data.

Satellite remote sensing will never entirely replace surface measurements. Rather, it will always serve to greatly extend them, both spatially and temporally. Satellites cannot, for example, directly measure ocean properties under an ice cover. Intermediate-scale measurements from helicopters and fixed-wing aircraft will continue to be important, and data collection from autonomous unmanned vehicles (both airborne and underwater) will play an increasing role in sea-ice research. But, as John Mobbs of the Australian Defence Force Academy put it, satellite remote sensing is still about the best way to stay informed by being out of touch. Or to put it another way, it's the most fun you can have without touching.

Rob Massom,
Polar Waters Program,
Antarctic CRC

This page was last modified on 5 October 2002.