74
R. Romeiser et al.
ground- or ship-based remote sensing systems (e.g. Palmer, 1991; Essen et al., 2000;
Plant et al., 2005) cannot be deployed easily.
5.2 How to Measure Currents by SAR
Although the idea of exploiting Doppler centroid anomalies of conventional SAR
raw data for current retrievals was formulated more than 30 years ago (Shuchman,
1979), it did not get much attention until an impressive demonstration with
ENVISAT ASAR data was published in 2005 (Chapron et al., 2005). In the meantime, the ATI technique had been proposed by Goldstein and Zebker (1987) and
demonstrated in several experiments. Initially, the ATI technique promised current
measurements at full SAR resolution, while a Doppler centroid anomaly analysis
seemed to reduce the SAR to a coarse-resolution real aperture radar, but as a result
of suboptimal system parameters of available spaceborne ATI systems on the one
hand and the development of optimised methods for Doppler centroid estimates on
the other hand, differences between actual results of the two techniques are much
smaller than one might expect. Let us have a brief look at the theoretical background.
5.2.1 Along-Track Interferometry
The ATI technique is based on interferometric combination of two complex SAR
images of the same scene, which are acquired with a short time lag on the order
of milliseconds. Phase differences between pixels of the two images are proportional to Doppler shifts of the backscattered signal. To obtain two interferometric
SAR (InSAR) images with a short time lag from a moving platform, one needs two
antennas separated by a corresponding distance in flight direction. Accordingly, the
technique is called along-track interferometry, not to be confused with cross-track
interferometry (XTI) for topographic mapping. As mentioned above, the ATI concept was first proposed by Goldstein and Zebker (1987). First airborne ATI results
were shown by Goldstein et al. (1989). Thompson and Jensen (1993) presented
results of another experiment and demonstrated the importance of correcting ATIderived velocity fields for contributions of wave motions. They were able to estimate
required corrections theoretically. Further airborne ATI experiments were carried
out, for example, by Ainsworth et al. (1995), Graber et al. (1996), Siegmund et al.
(2004), Bjerklie et al. (2005), Romeiser (2005), and Toporkov et al. (2005).
The time lag τ between the two ATI images is determined by the along-track
antenna separation L and platform velocity V. Depending on the transmit/receive
sequence of the antennas, one obtains τ = L/V or τ = L/2 V, where L or L/2,
respectively, is called effective baseline. For current measurements, τ needs to be
sufficiently long to obtain significant phase signatures from current variations of
interest and sufficiently short to avoid phase ambiguities and a decorrelation of
the backscattered signal. The decorrelation time depends on radar frequency and
wind/wave conditions. According to model results of Romeiser and Thompson
(2000), decorrelation times at X band (10 GHz) and L band (1 GHz) are on the
R. Romeiser et al.
ground- or ship-based remote sensing systems (e.g. Palmer, 1991; Essen et al., 2000;
Plant et al., 2005) cannot be deployed easily.
5.2 How to Measure Currents by SAR
Although the idea of exploiting Doppler centroid anomalies of conventional SAR
raw data for current retrievals was formulated more than 30 years ago (Shuchman,
1979), it did not get much attention until an impressive demonstration with
ENVISAT ASAR data was published in 2005 (Chapron et al., 2005). In the meantime, the ATI technique had been proposed by Goldstein and Zebker (1987) and
demonstrated in several experiments. Initially, the ATI technique promised current
measurements at full SAR resolution, while a Doppler centroid anomaly analysis
seemed to reduce the SAR to a coarse-resolution real aperture radar, but as a result
of suboptimal system parameters of available spaceborne ATI systems on the one
hand and the development of optimised methods for Doppler centroid estimates on
the other hand, differences between actual results of the two techniques are much
smaller than one might expect. Let us have a brief look at the theoretical background.
5.2.1 Along-Track Interferometry
The ATI technique is based on interferometric combination of two complex SAR
images of the same scene, which are acquired with a short time lag on the order
of milliseconds. Phase differences between pixels of the two images are proportional to Doppler shifts of the backscattered signal. To obtain two interferometric
SAR (InSAR) images with a short time lag from a moving platform, one needs two
antennas separated by a corresponding distance in flight direction. Accordingly, the
technique is called along-track interferometry, not to be confused with cross-track
interferometry (XTI) for topographic mapping. As mentioned above, the ATI concept was first proposed by Goldstein and Zebker (1987). First airborne ATI results
were shown by Goldstein et al. (1989). Thompson and Jensen (1993) presented
results of another experiment and demonstrated the importance of correcting ATIderived velocity fields for contributions of wave motions. They were able to estimate
required corrections theoretically. Further airborne ATI experiments were carried
out, for example, by Ainsworth et al. (1995), Graber et al. (1996), Siegmund et al.
(2004), Bjerklie et al. (2005), Romeiser (2005), and Toporkov et al. (2005).
The time lag τ between the two ATI images is determined by the along-track
antenna separation L and platform velocity V. Depending on the transmit/receive
sequence of the antennas, one obtains τ = L/V or τ = L/2 V, where L or L/2,
respectively, is called effective baseline. For current measurements, τ needs to be
sufficiently long to obtain significant phase signatures from current variations of
interest and sufficiently short to avoid phase ambiguities and a decorrelation of
the backscattered signal. The decorrelation time depends on radar frequency and
wind/wave conditions. According to model results of Romeiser and Thompson
(2000), decorrelation times at X band (10 GHz) and L band (1 GHz) are on the