76
R. Romeiser et al.
5.2.2 Doppler Centroid Anomaly Analysis
Direct instantaneous frequency determination from the phase history analysis of
single antenna returns is a standard methodology to process SAR images (e.g.
Madsen, 1989). Single-antenna Doppler estimates can indeed be directly obtained
from the measured return signal spectral peak frequencies, i.e. Doppler centroids.
Such estimates are commonly used to focus the SAR image. However, when
compared to geometrically predicted Doppler frequencies (considering the relative motion between satellite and rotating earth), systematic differences or Doppler
shift anomalies were reported for conventional SAR ocean scenes (Chapron et al.,
2002). Analysis of global Wave Mode data from ENVISAT ASAR (one 10 km ×
6 km “imagette” every 100 km) proved that these Doppler anomalies originated
from geophysical conditions. Their analysis works best for homogeneous scenes,
exhibiting small image intensity variations, and yields estimates with a spatial resolution of about 8 km × 4 km for ENVISAT ASAR Wide Swath Mode images
(swath width ≈ 400 km) and 1 km × 1 km for ERS and ENVISAT Image Mode
products (swath width ≈ 100 km). Note that the azimuthal resolution can be significantly better than the real aperture resolution of the radar antenna on the order of
5 km, since the Doppler centroid analysis can be combined with some amount of
SAR processing. Like in ATI data processing, there is a tradeoff between the effective spatial resolution and the relative rms error (compared to the local expectation
value) of retrieved radial velocity maps. From ASAR Wave Mode data, one velocity
estimate is obtained every 100 km along the track.
As interpreted (Chapron et al., 2005; Johannessen et al., 2008), the Doppler
anomaly is associated with an overall bulk velocity including the mean velocity of the radar detected surface scatters and the desired ocean surface current.
The Doppler anomaly has been found to depend on radar frequency, incidence
angle, polarisation, and environmental conditions, mostly wind speed and direction
(Mouche et al., 2008). The mean velocity of the radar detected scatterers is generally larger for HH than VV polarisation and decreases with the radar wavelength
and for incidence angles greater than 30 ◦ . Although no direct comparison has been
performed so far, the physical mechanisms that lead to differences between actual
surface currents and uncorrected Doppler velocities obtained from Doppler centroid
anomalies seem to be the same as the ones that affect ATI data (Thompson and
Jensen, 1993; Romeiser and Thompson, 2000; Romeiser, 2005).
The partitioning of the Doppler anomaly to the different contributions is a challenging problem, but the technique is robust and has the potential to meet high
spatial resolution requisites. In all cases, the measurement of Doppler anomalies
can complement the generation of conventional SAR images, so that geometrical
and dynamical properties of the ocean scene can be derived together. Under
favourable and well known environmental conditions, it is highly feasible to
clearly identify mesoscale and submesoscale features and to infer absolute surface velocities along the radar line-of-sight direction, as pointed out by Johannessen
et al. (2008).
R. Romeiser et al.
5.2.2 Doppler Centroid Anomaly Analysis
Direct instantaneous frequency determination from the phase history analysis of
single antenna returns is a standard methodology to process SAR images (e.g.
Madsen, 1989). Single-antenna Doppler estimates can indeed be directly obtained
from the measured return signal spectral peak frequencies, i.e. Doppler centroids.
Such estimates are commonly used to focus the SAR image. However, when
compared to geometrically predicted Doppler frequencies (considering the relative motion between satellite and rotating earth), systematic differences or Doppler
shift anomalies were reported for conventional SAR ocean scenes (Chapron et al.,
2002). Analysis of global Wave Mode data from ENVISAT ASAR (one 10 km ×
6 km “imagette” every 100 km) proved that these Doppler anomalies originated
from geophysical conditions. Their analysis works best for homogeneous scenes,
exhibiting small image intensity variations, and yields estimates with a spatial resolution of about 8 km × 4 km for ENVISAT ASAR Wide Swath Mode images
(swath width ≈ 400 km) and 1 km × 1 km for ERS and ENVISAT Image Mode
products (swath width ≈ 100 km). Note that the azimuthal resolution can be significantly better than the real aperture resolution of the radar antenna on the order of
5 km, since the Doppler centroid analysis can be combined with some amount of
SAR processing. Like in ATI data processing, there is a tradeoff between the effective spatial resolution and the relative rms error (compared to the local expectation
value) of retrieved radial velocity maps. From ASAR Wave Mode data, one velocity
estimate is obtained every 100 km along the track.
As interpreted (Chapron et al., 2005; Johannessen et al., 2008), the Doppler
anomaly is associated with an overall bulk velocity including the mean velocity of the radar detected surface scatters and the desired ocean surface current.
The Doppler anomaly has been found to depend on radar frequency, incidence
angle, polarisation, and environmental conditions, mostly wind speed and direction
(Mouche et al., 2008). The mean velocity of the radar detected scatterers is generally larger for HH than VV polarisation and decreases with the radar wavelength
and for incidence angles greater than 30 ◦ . Although no direct comparison has been
performed so far, the physical mechanisms that lead to differences between actual
surface currents and uncorrected Doppler velocities obtained from Doppler centroid
anomalies seem to be the same as the ones that affect ATI data (Thompson and
Jensen, 1993; Romeiser and Thompson, 2000; Romeiser, 2005).
The partitioning of the Doppler anomaly to the different contributions is a challenging problem, but the technique is robust and has the potential to meet high
spatial resolution requisites. In all cases, the measurement of Doppler anomalies
can complement the generation of conventional SAR images, so that geometrical
and dynamical properties of the ocean scene can be derived together. Under
favourable and well known environmental conditions, it is highly feasible to
clearly identify mesoscale and submesoscale features and to infer absolute surface velocities along the radar line-of-sight direction, as pointed out by Johannessen
et al. (2008).