5 Direct Surface Current Field Imaging from Space
79
parallel to the river bed, it was possible to construct a fully 2-D surface current field
(Fig. 5.4b) from the ATI-derived component (Fig. 5.4a). Again, the SRTM-derived
currents were found to be consistent with a numerical model or the river, UnTRIM
(Casulli and Walters, 2000).
The divided antenna of TerraSAR-X has an even shorter ATI time lag than
SRTM, and the instrument noise level is higher. However, according to Romeiser
and Runge (2007), the smaller pixel size of TerraSAR-X permits more averaging
of original pixel values at the same efffective spatial resolution (see also Fig. 5.1),
which actually overcompensates the phase sensitivity and instrument noise handicap. The effective spatial resolution of current fields from TerraSAR-X in stripmap
mode (swath width = 30 km, nominal pixel resolution = 3 m) was expected to be
better than 1 km at an rms error of current estimates of 0.1 m/s. A major advantage
of TerraSAR-X compared to SRTM is the pure ATI geometry of the divided antenna,
which facilitates absolute current measurements. With a pure ATI system, a phase
difference of 0 corresponds to a line-of sight velocity of 0, while phase differences
from combined ATI/XTI systems include a topographic contribution that is often
not well known.
ATI data acquisitions with TerraSAR-X are possible in various modes of operation, all of which are still in an experimental stage of development. In spring and
summer 2008, a first series of ATI images was acquired in the so-called Aperture
Switching (AS) mode, which uses a single receiver for both antenna halves in
an alternating way at a doubled pulse repetition frequency. This is less desirable,
but easier to implement than the full Dual Receive Antenna (DRA) mode, which
uses two receivers in parallel. In AS mode, the swath width of stripmap images
is reduced to about 16 km, the noise level is a little higher than in DRA mode,
and ambiguities in the SAR processing can produce ghost images of bright targets
on land over water. Nevertheless, Romeiser et al. (2010) were able to process and
analyse six AS-mode images of the Elbe river quite successfully. Again, UnTRIM
model results were used as reference. Example results for three of the six cases
are shown in Fig. 5.5. The data quality of TerraSAR-X AS-mode data seems to
be consistent with the theoretical predictions, and the retrieval of absolute currents
from the pure ATI data of TerraSAR-X (in contrast to relative current variations
within the scene from SRTM data) seems to work, but a final quantitative evaluation
has not yet been done due to the preliminary state of the existing data processing
routines.
5.3.2 Doppler Anomaly Analysis Results
At first order, the Doppler anomaly is mostly wind dependent, as revealed when collocated monthly wind fields from the European Centre for Medium Range Weather
Forecasts (ECMWF) were projected along the radial direction of ENVISAT ASAR
Wave Mode data. Based on the collocated data set obtained this way, a neural network model, called CDOP, was created for an incidence angles of 23 ◦ and 33 ◦ ,
where inputs are wind speed and relative wind direction with respect to azimuth;
79
parallel to the river bed, it was possible to construct a fully 2-D surface current field
(Fig. 5.4b) from the ATI-derived component (Fig. 5.4a). Again, the SRTM-derived
currents were found to be consistent with a numerical model or the river, UnTRIM
(Casulli and Walters, 2000).
The divided antenna of TerraSAR-X has an even shorter ATI time lag than
SRTM, and the instrument noise level is higher. However, according to Romeiser
and Runge (2007), the smaller pixel size of TerraSAR-X permits more averaging
of original pixel values at the same efffective spatial resolution (see also Fig. 5.1),
which actually overcompensates the phase sensitivity and instrument noise handicap. The effective spatial resolution of current fields from TerraSAR-X in stripmap
mode (swath width = 30 km, nominal pixel resolution = 3 m) was expected to be
better than 1 km at an rms error of current estimates of 0.1 m/s. A major advantage
of TerraSAR-X compared to SRTM is the pure ATI geometry of the divided antenna,
which facilitates absolute current measurements. With a pure ATI system, a phase
difference of 0 corresponds to a line-of sight velocity of 0, while phase differences
from combined ATI/XTI systems include a topographic contribution that is often
not well known.
ATI data acquisitions with TerraSAR-X are possible in various modes of operation, all of which are still in an experimental stage of development. In spring and
summer 2008, a first series of ATI images was acquired in the so-called Aperture
Switching (AS) mode, which uses a single receiver for both antenna halves in
an alternating way at a doubled pulse repetition frequency. This is less desirable,
but easier to implement than the full Dual Receive Antenna (DRA) mode, which
uses two receivers in parallel. In AS mode, the swath width of stripmap images
is reduced to about 16 km, the noise level is a little higher than in DRA mode,
and ambiguities in the SAR processing can produce ghost images of bright targets
on land over water. Nevertheless, Romeiser et al. (2010) were able to process and
analyse six AS-mode images of the Elbe river quite successfully. Again, UnTRIM
model results were used as reference. Example results for three of the six cases
are shown in Fig. 5.5. The data quality of TerraSAR-X AS-mode data seems to
be consistent with the theoretical predictions, and the retrieval of absolute currents
from the pure ATI data of TerraSAR-X (in contrast to relative current variations
within the scene from SRTM data) seems to work, but a final quantitative evaluation
has not yet been done due to the preliminary state of the existing data processing
routines.
5.3.2 Doppler Anomaly Analysis Results
At first order, the Doppler anomaly is mostly wind dependent, as revealed when collocated monthly wind fields from the European Centre for Medium Range Weather
Forecasts (ECMWF) were projected along the radial direction of ENVISAT ASAR
Wave Mode data. Based on the collocated data set obtained this way, a neural network model, called CDOP, was created for an incidence angles of 23 ◦ and 33 ◦ ,
where inputs are wind speed and relative wind direction with respect to azimuth;