IEEE Geoscience and Remote Sensing Magazine - March 2013 - 23

with a nominal bandwidth of B r =100 MHz. Choosing
a baseline which is just 10% of the critical baseline, the
corresponding height of ambiguity is about 10 m for an
incident angle of i i = 45°. Such a height of ambiguity is
quite low and it may cause irresolvable height errors in
areas with rough terrain. It is therefore in general not possible to take full advantage of the opportunity for large
baseline acquisitions provided by the new generation of
high-bandwidth radar systems.
A possible solution to this dilemma is an interferometric system with flexible baseline lengths. This enables an
adaptation of the baseline to the actual terrain and it offers
furthermore the possibility to image one and the same area
with multiple baselines of different length. The latter can
be used for an unambiguous reconstruction of the terrain
height [80], [81]. Such a strategy is also employed in the
TanDEM-X mission where two or more interferograms with
a typical baseline ratio of 0.7 are combined to resolve height
ambiguities. The results obtained with this technique are
rather promising [82]. In the future, one may also use systems that allow for the acquisition of multiple baselines
in a single-pass. An example are satellite formations with
multiple spacecraft that allow for adjustable baselines ranging from less than 100 meters up to 10 kilometers and
more [83]. The interferogram in Fig. 12 was obtained with
a height of ambiguity of 47 m and the terrain allowed for
successful phase unwrapping without the need for a second
acquisition with a different baseline.
The last steps in the DEM generation process are
phase-to-height conversion and geocoding. The phase-toheight conversion uses the imaging geometry of Fig. 11 in
combination with the interferometrically derived range
difference to determine by trigonometry for each image
point its position relative to the sensor. This requires
both a sufficiently accurate knowledge of the interferometric baseline between the radar antennas and a precise
knowledge of the range difference for each scatterer. Since
the latter is derived from the unwrapped interferometric
phase, it is unfortunately only known up to a global offset which is a multiple of the wavelength. The remaining uncertainty is typically resolved by using at least one
external reference. An alternative is the employment of
radargrammetric techniques, which measure the mutual
range shift between corresponding SAR image pixels to
derive a coarse DEM. The radargrammetric measurement
is unambiguous but its accuracy is typically two orders of
magnitude worse than that of the interferometric technique. Nevertheless, a sufficient accuracy can be obtained
by averaging over a large scene as it is done in the operational TanDEM-X processor [84]. The final step is geocoding, which involves a transformation from the radar
geometry to the coordinates of a selected geodetic reference system. Fig. 14 shows as an example the geocoded
digital elevation model (DEM) that has been derived by
the experimental TAXI processor [85] from the Atacama
interferogram of Fig. 12.
march 2013

ieee Geoscience and remote sensing magazine

FIGURE 14. Geocoded digital elevation model (DEM) derived from
the unwrapped interferometric phase of Fig. 12.

B. Differential SAR Interferometry
Differential SAR interferometry (DInSAR) is a further clear
example of a well-established interferometric technique.
Similar as with InSAR, the high sensitivity of a SAR instrument to measure the LOS propagation distance is exploited
in order to detect displacements of the Earth surface at a
wavelength scale.
Consider two SAR images acquired with a certain temporal separation that are combined to generate an interferogram. Ideally, a zero-baseline configuration would result
in an interferogram whose phase information would only
be related to the LOS displacement in the scene. In practice
though, a certain baseline is always present, which makes
the interferogram also sensitive to the topography of the
scene. By using an external DEM, the topographic information can be subtracted from the interferogram, leading
to a differential SAR interferometric measurement where
subtle changes of the range distance between the two
acquisitions (e.g., due to subsidence) can be detected. After
the topographic phase removal, the phase of the interferogram becomes

{ disp =

4r
Trdisp, (34)
m

where Trdisp is the desired LOS displacement. As it can
be noted in (34), DInSAR can achieve an accuracy in the
order of a fraction of the wavelength in the measurement
of the LOS displacement, hence becoming a powerful
tool for deformation monitoring of large areas. Similar to
InSAR, the differential phase needs to be unwrapped and
calibrated in order to obtain absolute displacement results
(see example in Fig. 15(a)). But several aspects impair the
performance of this approach. First, the accuracy of the
external DEM needs to be in the same order of magnitude
or better than the phase sensitivity of the interferometric baseline. But more important, one needs to consider
phase noise due to temporal decorrelation and variations
in the propagation medium. The latter is a consequence of
the atmospheric delay, which is mainly due to the water
23

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2013