POLinSAR: Advances in radar remote sensing
An advanced technique for analysing radar images shows tremendous promise for scientists studying forests, agriculture, ice and other terrain types, but experts at a recent ESA workshop cautioned that research work is needed before practical applications can be developed.
More than 120 scientists and researchers from 20 countries gathered recently at ESA’s ESRIN facility in Frascati, Italy, for a three-day workshop to share the latest results on scientific tests and potential applications of SAR polarimetry and polarimetric interferometry.
With ten radar satellites expected to be in orbit by the end of the decade, radar imagery and POLinSAR techniques could find practical uses ranging from detecting buried landmines, providing early warnings of threats to agricultural areas, and certifying forests for logging to monitoring compliance with international treaties on global warming and helping national governments to protect the biodiversity and ecological balance of their forests, wetlands and other natural assets.
"We are just at the starting point of revealing new applications," commented Yves-Louis Desnos, workshop chairman and ESA’s Head of R&D Support for the Exploitation and Science Division of the Earth Observation Programme.
From polarimetry to polarimetric interferometry
The workshop addressed the latest research developments that are moving from classical SAR studies to polarimetric SAR (or SAR polarimetry, as it’s also called). This features the capability to sense the Earth through various polarisation combinations. "SAR polarimetry clearly has many potential applications in agriculture monitoring, oil slicks, forest and sea ice classification," comments Henning Skriver, associate professor with the Technical University of Denmark (DTU). "It also needs further scientific development."
Known by the unwieldy acronym, POLinSAR, the technique combines varying the orientation, or polarisation, of radar signals (POLarimetry) with the analysis of the phase differences in the signal to produce differential range and range-change measurements (interferometry) from two or more images captured by synthetic aperture radars (SARs). Taken together, polarimetry and interferometry offers the potential to see the Earth in three dimensions.
POLinSAR analysis techniques emerged in the mid-1990s and are developing quickly. "POLinSAR has matured from a ‘first-results’ to a 'science' status," said Konstantinos Papathanassiou, a researcher with the Institute for Radio Frequency Technology and Radar Systems at German Aerospace Centre (DLR). "You can do things that were not possible before and are needed now."
Seeing the forest, measuring the trees
The capability of radar to penetrate ground cover and 'see' the underlying terrain, coupled with POLinSAR techniques to detect forest canopies, make classifying trees and how tall they are using SAR imagery a possibility. This may sound of interest only to a narrow band of scientists, until one realizes that determining the types and heights of trees in a forest are critical ingredients in determining its biomass – a quantitative estimate of the entire amount of organic material in a particular habitat.
In turn, a forest’s biomass, together with how it changes over time, are key elements in determining an area’s capacity to act as a carbon sink, soaking up carbon-based gases and cleansing the atmosphere of major types of pollution blamed for the greenhouse effect and global warming. Producing accurate forest biomass estimates, and how they are changing over time, are a critical challenge to environmental scientists to whom national governments are turning for help in meeting their international obligations to stabilise greenhouse-gas emissions under the Kyoto Protocol.
As a result, a number of tests are underway to assess how SAR imagery could be used to produce accurate biomass estimates. Shane Cloude, with the UK-based AEL Consultants, presented the results of an ESA-funded project that employed polarimetric interferometry to measure the height of trees with data captured by an airplane-mounted SAR. A forest of Scotch pine trees, naturally enough in Scotland, was chosen to serve as a testing area.
The results of the test were promising, within about a ten percent margin of error. Cloude reported that evaluation "consistently underestimated" the height of the trees by about a half-metre when the SAR data were matched against so-called 'ground truth – actual on-site measurements.
"It’s tough to do, and we are further along with single-species forests," Cloude said. "Forest biomass estimates are very important for measuring compliance with Kyoto targets, but we are far away from the political dimensions of using POLinSAR techniques for monitoring compliance."
New model for classifying agriculture
At a roundtable discussion on the forest-related study results presented at the conference, several participants urged that research be expanded to include more types of forests. Dirk Hoekman, with the Department of Environmental Sciences at Wageningen University in the Netherlands, pointing out that the recent Indonesian forest fires released a gigaton of carbon into the atmosphere, or one-third of the total annual increase of global carbon gases. More attention from POLinSAR researchers should be paid to tropical and peat moss forests, he said.
Hoekman's own paper on a new model for classifying the types of agriculture in SAR images captured the attention of some of the scientists at the meeting because of the promise of improved accuracy of the results. Accurately classifying agricultural types and other terrain features such as land cover and sea ice in SAR imagery is a problem that has plagued researchers and hampered the development of reliable applications. Tests of crop classifications from imagery taken over the Dutch Flevoland agricultural test site indicate that new ways to classify the radar data with levels of accuracy of 90.4% for C-band and 88.7 % for L-band, according to Hoekman’s paper presented to the conference.
"The new classification method probably will solve one of the problems we have had with existing methods," said DTU’s Skriver. "These methods have not been able to fully cope with the natural variation within individual classes and, hence, have limited the classification accuracy."
Sessions reveal potential applications
Workshop presentations, including lively poster session during which researchers were limited to two minutes to present a synopsis of their work and extend an invitation to speak at greater length, gave a glimpse at the diversity of research now underway in the SAR science community. Among the highlights:
An evaluation the best SAR polarisation methods to use for detecting land mines, including modern plastic versions that are nearly invisible to conventional ground-penetrating radars. Juan Lopez-Sanchez from Spain’s University of Alicante explained how a test conducted at a European microwave lab found that circular or near-circular SAR polarisations were best to distinguish buried mines from the surrounding soil. A complete analysis of the test results is currently underway, he added.
Polarimetric SAR techniques could improve the detection and classification of oil slicks. Polarization coherence and the polarimetric anisotropy have been suggested as ways to classify terrain surface roughness, stated Joaquim Fortuny-Guasch with Italy’s Joint Research Centre. After an examination of imagery of areas of the North Sea, the English Channel, northern Germany, and Pantelleria Island in southern Italy, preliminary results indicated that surface roughness is highly correlated with changes in the thickness and/or the texture of the oil layer.
Hugh Corr with the British Antarctic Survey presented details of how ground-based SARs and polarimetric techniques have been used to look through ice more than 2.5 km thick to study crystal formations and distinguish subsurface ice flows.
- Interferometric studies using SAR data under an ESA-sponsored project is revealing how the land around Polish coal mines is subsiding at a rate up to a half-metre per month, according to Zbigniew Perski, scientist with Poland’s University of Silesia. "Today, the determination of the areas affected by terrain deformation risks becomes on of the most promising applications for operational use of SAR data," he stated.
Plans for new space SARs
With their all-weather and nighttime capabilities, coupled with these promising new analytical methods for interpreting the results, space-based radars are providing new insights into the Earth’s land, waters and atmosphere. Participants at the POLinSAR workshop heard details of upcoming satellite missions that will incorporate new advances in SAR technology, faster revisit times, and a more complete inventory of those spots on Earth of interest to the scientists and the public.
Gordon Staples with Radarsat International detailed current planning for the Radarsat-2 follow-on to the existing orbiting Canadian Radarsat spacecraft. Planned for a 2004 launch, the spacecraft is being funded through a cooperative arrangement between the Canadian government and Radarsat International's parent company, MacDonald Dettwiler. It will feature a high-resolution, 3-metre resolution mode and the capability to scan both right and left along the satellite’s track by physically rotating the spacecraft one a day. After an evaluation of 32 application areas for radar imagery, Staples said that the new satellite offers the best improvements in examining crop types, crop conditions and providing information about sea ice.
The TerraSAR-X satellite will be the first German SAR for scientific and commercial applications, said DLR's Alberto Moreira. Planned for a 2005 launch, the X-band SAR features a 1-metre “spotlight” mode, with an experimental 0.6-metre resolution using the spacecraft’s 300-MHz-bandwidth transmit capability, Moreira said. Another unique feature of the spacecraft is the capability to split the antenna receiving the return radar signal into two separate channels to allow for the tracking of moving targets for measuring water currents and monitoring traffic over a large area.
"The X-band applications include mostly sea ice, snow cover and urban planning," Moreira said, with other uses including agricultural, map-making and risk assessment studies for floods, fires and storms.
A four-satellite constellation of X-band SARs is planned as the Cosmo-SkyMed program, developed by the Italian Space Agency with the cooperation of France’s space agency, CNES. The dual, military/civilian mission follows Italy’s HELIOS 1 military observation satellite that is ending in 2004. With the first launch planned in 2005, Cosmo-SkyMed will be coordinated with optical satellites, feature revisit times of a few hours and metre and sub-metre resolutions.
ESA’s Envisat Mission Manager Henri Laur provided workshop participants with a status report on ESA’s latest Earth observation mission. He stressed the synergy among the 10 sensors carried by the Envisat satellite, point out that Envisat’s advanced SAR that can switch between horizontal and vertical polarisations of the radar signal can acquire imagery simultaneously with the MERIS imaging spectrometer to allow a more complete 'top and bottom' analysis of various natural phenomena such as hurricanes.
Responding to questions concerning the availability of Envisat imagery, Laur said that all Envisat instruments are completely activated and operating, all ASAR products activated to be released, and data are being delivered to scientific users.
A new Envisat ground station, located in Svalbard at the northern edge of Norway, brings additional capability to the network for delivering Envisat data to users, the ESA official added. The arrival on-station of Artemis, ESA’s inter-satellite data relay spacecraft, will allow data to be received directly at ESA’s ESRIN facility, Laur said, with tests of the system to be conducted in February and March.
POLinSAR: stay tuned for more to come
The ESA POLinSAR workshop clearly demonstrated the strides taken by the radar research community in the past few years in radar polarimetry techniques and applications, and that more work is needed to move from the lab to practical uses.
"The workshop showed that the performance of several remote-sensing applications increases by introducing polarimetric information," commented DLR’s Papathanassiou. "The discussions also highlighted the limitations and problems in establishing applications."