Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
ERS – ENVISAT TANDEM DATA OVER SEA AND SHELF ICE
Urs Wegmüller, Charles Werner, Maurizio Santoro, Tazio Strozzi and Andreas Wiesmann
Gamma Remote Sensing AG, Worbstrasse 225, CH-3073 Gümligen, Switzerland,
http://www.gamma-rs.ch, [email protected]
ABSTRACT
ERS – ENVISAT Tandem (EET) data are SAR data
pairs acquired by the ERS-2 and ENVISAT ASAR
instruments from approximately the same orbits within
28 minutes. During two dedicated campaigns in
2007/2008 and 2008/2009 ESA specifically acquired
EET pairs with baselines suited for cross-interferometry
(CInSAR). At perpendicular baselines of about 2km the
frequency difference and baseline effects on the
reflectivity spectrum compensate and so coherent
interferograms can be obtained. Sea and shelf ice has
usually a relatively flat surface and so there is little topographic phase observed, even for 2km
perpendicular baselines. Centimeter scale motion
occurring during the 28 minute interval results in
deformation phase. Motion at this rate is often observed
for sea and shelf ice and is of interest to understand
dynamics and stress occurring. More open sea ice
moves at even much higher rates. Here EET coherence
is typically lost but offset tracking and split-beam
interferometry may be used to retrieve motion fields.
The sensitivities of these techniques are in the order of
1/20 of a SAR image pixel. So in azimuth direction this
translates to a sensitivity of about 20cm per 28 minutes interval and in cross-track direction to about 1m per 28
minutes interval. On the other hand maximum rates
which can reliably be retrieved correspond to offsets of
several pixels between the two acquisitions
corresponding to rates up to the order of 100m per 28
minutes interval.
Repeat-pass interferometry, offset tracking, and split-
beam interferometry were already applied in the past
using data at different wavelengths and time intervals.
So far the shortest interval available was 1 day from the
ERS-1/2 Tandem mission. Now, EET data offers data at
a much shorter 28 minutes interval. Consequently, faster
motion rates can be investigated.
1. INTRODUCTION
In 2002 ESA launched the ENVISAT satellite with the
Advanced SAR (ASAR). ENVISAT is operated in the
same orbits as the ERS-2, preceding ERS-2 by
approximately 28 minutes. One of the ASAR modes,
namely IS2 at VV-polarization corresponds closely to
the ERS SAR mode, except for the slightly different
sensor frequency used. A unique opportunity offered by
these two similar SAR instruments operated in the same
orbital configuration is ERS – ENVISAT cross-
interferometry (CInSAR). At perpendicular baselines of
approximately 2 kilometers the look-angle effect on the reflectivity spectrum compensates for the carrier
frequency difference effect. As was shown with
examples over Germany, the Netherlands, Italy, and
Switzerland [1] CInSAR has a good potential to
generate accurate DEMs over relatively flat terrain.
In the case of fast moving targets interferometric data
acquired with a short 28 minute interval may also be of
interest for displacement mapping. Similarly, tracking
and split-beam interferometry techniques may be
applied to these short repeat interval pairs to map even
faster displacements.
After a review of the CInSAR, offset tracking and split-
beam interferometry methodologies the main focus of
this work is on the assessment of the displacement rate
ranges within which EET CInSAR, offset tracking and
split-beam interferometry can be used to derive motion
information.
2 METHODOLOGIES
2.1 EET CInSAR
A SAR observes a surface area under a specific
incidence angle. Given this incidence angle, the radar
carrier or center frequency, and the chirp bandwidth a
certain part of the reflectivity spectrum of the surface is
measured. Changes to the incidence angle or to the radar
frequency cause changes of the reflectivity spectrum. To
obtain coherence between two acquisitions, the
reflectivity spectra of the two signals being interfered
have to overlap. For two acquisitions acquired at the same frequency this means that the incidence angle
difference needs to be small. At larger incidence angle
differences, corresponding to baselines larger than the
so-called critical baseline the reflectivity spectra do not
overlap anymore, resulting in complete decorrelation of
the signal over distributed targets. In the case of
acquisitions at different frequencies the effect of the
frequency on the reflectivity spectrum has also to be
considered. In fact, there exists a condition under which
_____________________________________________________ Proc. ‘Fringe 2009 Workshop’, Frascati, Italy, 30 November – 4 December 2009 (ESA SP-677, March 2010)
the incidence angle effect is exactly compensated by the
radar frequency effect. For a given frequency difference
(f2 – f1) the perpendicular baseline component ⊥B for
which the two effects compensate each other is
approximately [2]
1
112 )tan()(
f
rffB
αθ −⋅−≅⊥ (1)
In the case of the 31 MHz carrier frequency difference
between ERS2 and ENVISAT ASAR perpendicular
baselines of approximately 2 kilometer compensate the
look-angle effect on the reflectivity spectrum
compensate for a flat surface. The condition is however
sensitive to the local incidence angle, (θ - α). Figure 1
illustrates the orbital configuration and the acquisition
geometry with a 2km baseline that compensates the
frequency difference effect. The sign of the baseline is
such that ENVISAT sees the area under a steeper
incidence angle. ERS2 and ENVISAT carry right
looking sensors so seen in flight direction the ENVISAT
orbit has to be further to the right than the ERS2 orbit.
The effect of having a perpendicular baseline just close
to the optimal value is only partial overlap of the
reflectivity spectra. As long as the overlap is significant
high coherence can be maintained by common band
filtering, i.e. by only considering the overlapping part of the reflectivity spectra. Because of the spatial variation
of the slope angle α this common band filtering should
be done spatially adaptive.
Figure 1 The image to the left shows the orbital
geometry with large baselines at higher northern
latitudes. The image to the right shows the
interferometric imaging geometry. ENVISAT ASAR
has a 31MHz higher carrier frequency than ERS2.
Another constraint to finding coherence in an ERS -
ENVISAT interferometric pair is that overlapping
Doppler spectra of the images acquired by the two
sensors are required. Again, in the case of only partial
overlap the coherence is optimized by common band
filtering, this time of the azimuth spectra, is required.
Apart from the range common band filtering that needs
to be adapted considering the center frequency
difference the interferogram generation for a cross-
sensor interferogram is identical to that of a single
sensor interferogram.
In a complex-valued focused SAR image the phase corresponds to a sum of a term related to the return path
of the signal and a term related to the arrangement of
the scatterers in the resolution cell. For cells with many
distributed scatterers this second part results in speckle
with a spatially random phase in the interval [-π,π]. However, images acquired from almost the identical
aspect angle have almost identical speckle. Under such
conditions the phase difference φ can be expressed as
)
44( 1
12
2 rc
fr
c
f ππφ −−= (2)
where f1 and f2 are the radar center frequencies, r1 and r2
are the slant ranges of the two scenes. Notice, that there
is a significant range phase ramp even in the case of a
zero baseline related to the different center frequencies
m
radERSASAR
c
ff
dr
d3.1
)(4=
−=
πφ (3)
In the ERS – ENVISAT case this results in a phase
ramp of 1.3 radian per meter slant range, or more than a
full phase cycle per range pixel spacing.
For f1 = f2 well known equations describe the
dependence of the interferometric phase on orbit
geometry, scene topography, line-of-sight surface
displacements, and atmospheric path delay
noisepathdisporb rh
r
Bφφ
λ
π
θλ
πφφ +++
⋅+= ⊥ 4
sin
4 (4)
with an orbital phase term orbφ , the wavelength, λ , and
the baseline component perpendicular to the look
vector, ⊥B , the incidence angle, θ , the slant range, r ,
the path delay term pathφ , and a phase noise (or
decorrelation) term noiseφ . In the case of slightly
different frequencies f1 ≠ f2 equation (4) can still be used as a good approximation with only minor
adaptations. The orbital phase, this is the interferometric
phase observed of the flat ellipsoid, is calculated using
(2) with r1 and r2 being the slant ranges to the flat
ellipsoid. This term also contains the strong range phase
ramp of (3). Apart from this the only difference is that
an average wavelength is used
21
2
ff
c
+=λ (5)
The phase to height sensitivity of ERS – ENVISAT
interferograms at mid-swath is
⊥
°⋅⋅= B
mmdh
d
23sin000'852056439.0
4πφ
⊥⋅= Bm
rad
dh
d2
0006688.0φ
(6)
For a 2.0 km baseline this corresponds to 1.34 rad/m
corresponding to a height ambiguity, defined as the
height per topographic phase cycle, of 4.70m.
The phase to line-of-sight displacement sensitivity can
directly be found in (4)
noisepathdisporb rhr
Bφφ
λ
π
θλ
πφφ +++
⋅+= ⊥ 4
sin
4 (7)
The relation between phase noise σφ the number of
looks used, NL, and the coherence,γ can be
approximated by [3]
γ
γσ φ
21
2
1 −=
LN (8)
As we see from (7) the phase noise can be reduced by
taking more interferometric looks. For a coherence of 0.5,
for example, it is realistic to reduce the phase noise to 0.2
radian.
Introducing the phase noise into (6) and (7) tells us the
precision of EET CInSAR estimates for elevation
differences and slant range displacements. Using the 0.2 radian phase noise results in an elevation precision of
0.15m and a LOS displacement precision of 0.9mm and
for the average displacement rate over the 28’
observation period the precision is 1.9mm/hour,
respectively 4.5cm/day or 16.7m/year (see Table 1).
An example of an EET CInSAR example over the
Kugmallit Bay, Canada, is shown in Figure 2.
Interferometric phase that most likely relates to
displacement in the line-of-sight direction is observed
over the sea ice. Considering that phase unwrapping is
necessary for a quantitative interpretation we can
determine a very rough upper limit for displacements
which may be retrieved. Assuming that we can rarely
unwrap more than about 10 phase cycles results in a
maximum displacement of 28cm or, converted to an average displacement rate of 60cm/hour, respectively
14.4m/day or 5260m/year (see Table 1).
2.2 EET offset tracking
Since the late nineteen nineties SAR offset-tracking
procedures are a welcome alternative to SAR
interferometry for the estimation displacement rates [4-
8]) Intensity and coherence tracking can be used. With
both techniques the registration offsets of two SAR
images in both slant-range (i.e. in the line-of-sight of the
satellite) and azimuth (i.e. along the orbit of the
satellite) directions are generated and used to estimate
the displacements. The estimated offsets are
unambiguous values which means that there is no need
for phase unwrapping, one of the most critical step in
SAR interferometry. Typical slant-range and azimuth
offset estimation errors are of the order of 1/20th
of an
SLC pixel [9]. This value is rather a conservative
estimate. Using advanced matching techniques
accuracies up to about 1/100 pixel are possible [11]. For
EET pairs it is important to consider the effect of the
terrain height on the range offset. An ambiguity height of 4.7m (for a 2km baseline) means that 4.7m elevation
difference will also cause a 2.8cm slant range
registration offset. For EET pairs the accuracies
resulting from the more conservative 1/20 pixel offset
estimation accuracy in slant-range and azimuth direction
are 40cm and 20cm, respectively. Converting this to
average displacement rate accuracies results for the
azimuth displacements in 43cm/hour and for the slant
range displacement in 86cm/hour (see Table 1). Offsets
of several pixels can reliably be estimated. As an upper
limit to the applicability of the technique (Table 1) we assume that offsets up to 10 pixels in slant range and 50
pixels in azimuth can be estimated.
An EET offset tracking result over sea ice is shown in
Figure 3. Displacements of tens of meters over the 28’
interval are observed.
2.3 EET split-beam interferometry
“Split-beam” interferograms were used in the past to
estimate displacements along the azimuth direction [12].
The processing sequence used is as follows. The
available azimuth bandwidth (about 50% in the case of
the EET pair on 7-Dec-2007) is divided in two non-
overlapping azimuth bands of equal width. This is done
for both SLCs of the EET pair by band-pass filtering.
For both pairs of sub-band SLCs one-pixel
interferograms are calculated. The combined interferogram is then calculated from the one-pixel
interferograms by multiplication of one complex
interferogram with the complex conjugate of the other
interferogram. The resulting phase corresponds to the
phase of the “split-window” interferogram. An initial
multi-looking is then applied to move from the one-
pixel geometry to a one-look geometry.
The phase φ of the split-beam interferogram can be
approximated for small squint angles by [12]
l
dazπφ
2= (9)
With daz the displacement in azimuth direction and l the
antenna size (about 10m for ERS-2 and ENVISAT
ASAR). As in the normal interferometry case the phase
noise can be estimated from the coherence and number
of looks (8).
Figure 2 EET CInSAR (10-Mar-2009, dt= 28minutes, B⊥= 2247m) differential interferometric phase (left, a color cycle corresponds to a phase cycle) relative to a constant height and RGB composite of the coherence (red) backscattering
(green) and backscatter change (blue, right) over Kugmallit Bay, Canada.
Table 1 Summery of ERS-ENVISAT Tandem DInSAR, split beam interferometry and offset tracking sensitivities,
precisions, and approximate maximum applicable displacement rates. The EET repeat interval used is 28 minutes.
Technique Parameter Sensitivity Precision Maximum
LOS
displacement
rate
Along-track
displacement
rate
LOS
displacement
rate
Along-track
displacement
rate
InSAR Unwrapped
phase π
λ
φ 4=
d
drdisp
0.9mm/28´ 4.5cm/day
16.7m/y
NA 28.0cm/28´ 14.4m/day
5260m/y
NA
Split-
beam
InSAR
Unwrapped
phase πφ 2
l
d
dazdisp=
NA 0.14m/28´
7.2m/day
2.6km/y
50m/28´
2.6km/day
940km/year
Offset
tracking
Range &
azimuth
offsets az
daz
dazr
dr
dr
off
disp
off
disp∆=∆= ,
0.40m/28´
20.6m/day
7.5km/y
0.20m/28´
10.3m/day
3.8km/y
80m/28´
4.1km/day
1500km/y
40m/28´
2.1km/day
750km/y
Figure 3 Geocoded sea ice displacement map derived from a 28’ ERS-ENVISAT pair acquired on 7-Dec-2007 over
Franz Josef Land (area size 53km x 56km). The image brightness corresponds to the backscattering of the ASAR
image. For more discussion on this result and the methodology used see [10].
In the case of fast coherent displacements along the
azimuth direction a non-zero phase is observed. In the
data over Franz Josef Land this is the case over some
fast moving sea-ice. Figure 4 shows the split-beam
interferogram phase for a small section of the 7-Dec-
2007 EET pair over Franz Josef Land. The observed
phase in the North-Western corner corresponds to fast
motion as is confirmed by the offset tracking result
shown in Figure 3.
For a coherence of 0.5 we estimated a phase noise of 0.2
radian when using 20 interferometric looks. Now in the
case of the slowly varying split-beam interferogram phase
significantly more than 20 looks can be used. Assuming
100 looks are used we get a phase estimation accuracy of
0.09 radian resulting in an azimuth displacement
accuracy of 0.14m which is lower than the estimated
offset tracking accuracy in azimuth direction. This
estimate conforms to [12] reporting for split-beam
interferometry a higher sensitivity than for offset
tracking.
Figure 4 Geocoded EET phase of split-beam
interferogram on 7-Dec-2007 over Franz Josef Land.
3. DISCUSSION
Concerning its potential for displacement mapping the
main difference of EET pairs is the sorter time interval
covered. As compared to one day interval of an ERS-1/2
Tandem pair, which was in the past the shortest time
interval available, the 28 minutes are about 50 times
shorter. Consequently, interferometry, offset tracking
and split-beam interferometry are applicable for
displacement rates which are about 50 times higher than
for ERS-1/2 Tandem pairs. EET CInSAR is of interest for very fast glaciers and for slowly moving sea ice in
with displacements between a few millimeters and
about 30cm for the 28 minute interval. Tracking and
split-beam interferometry are of interest for faster
moving sea ice with displacements between decimeters
and tens of meters for the 28 minute interval.
InSAR only permits estimation of displacements in the
line-of-sight slant range direction. Split-beam
interferometry only permits estimation of along-track
displacements. Using offset tracking both components
can be estimated. So for smaller displacements in range
direction InSAR is the most accurate technique. The
most accurate technique in the along-track direction is
split-beam interferometry.
InSAR and split-beam interferometry require coherence
between the two scenes. Consequently only pairs with
suited perpendicular baselines near 2km and
significantly overlapping Doppler spectra can be used.
Furthermore, phase unwrapping is required. Tracking is always applicable and no phase unwrapping is required.
It is also applicable, for example, over Antarctic regions
where no EET pairs suited for CInSAR were acquired
so far.
4. CONCLUSIONS
In this paper techniques to map surface displacements
using EET pairs acquired with a 28 minute interval were
assessed. Because of the short interval EET CInSAR,
offset tracking and split-beam interferometry permit
mapping relatively fast movements. As shown in this contribution such movements are observed over sea ice.
Slower movements, e.g. in coastal areas can be mapped
using EET CInSAR. For faster moving ice such a
drifting ice, offset tracking and split-beam
interferometry are well suited.
5. ACKNOWLEDGMENTS
This work was supported by ESA under contract
22526/09/I-LG on ERS-ENVISAT Tandem Cross-
Interferometry Campaigns: Case Studies. ERS and
ASAR data copyright ESA (CAT 6744). SRTM DEM
copyright USGS.
6. REFERENCES
[1] Wegmüller U., M. Santoro, C. Werner, T. Strozzi, A.
Wiesmann, and W. Lengert, “DEM generation using
ERS–ENVISAT interferometry“, Journal of Applied Geophysics Vol. 69, pp 51–58, 2009,
doi:10.1016/j.jappgeo.2009.04.002.
[2] Santoro M., J. I. H. Askne, U. Wegmüller, and C. L. Werner, "Observations, modeling, and applications of ERS-ENVISAT coherence over land surfaces," IEEE
Trans. Geosci. Remote Sensing, vol. 45, pp. 2600-2611, 2007.
[3] E. Rodriguez and J. M. Martin, “Theory and design of
interferometric synthetic aperture radars“, IEE Proc. Part F Radar Signal Process, 139(2), 149-159, 1992.
[4] Rott H., M. Stuefer, A. Siegel, P. Skvarca, and A. Eckstaller, “Mass fluxes and dynamics of Moreno Glacier,
Southern Patagonia Icefield,” Geophysical Research
Letters, Vol. 25, No. 9, pp. 1407-1410, 1998.
[5] Gray L., K. Mattar, and P Vachon, “InSAR results from
the RADARSAT Antarctic mapping mission: estimation of glacier motion using a simple registration procedure,” Proceedings of IGARSS’98, Seattle, USA, 1998.
[6] Michel R., and E. Rignot, “Flow of Glaciar Moreno,
Argentina, from repeat-pass Shuttle Imaging Radar images: comparison of the phase correlation method with radar interferometry,” Journal of Glaciology, Vol. 45, No.
149, pp. 93-100, 1999.
[7] Gray L., K. Mattar, and P Vachon, “InSAR results from the RADARSAT Antarctic mapping mission: estimation
of glacier motion using a simple registration procedure,” Proceedings of IGARSS’98, Seattle, USA, 1998.
[8] Derauw D., “DInSAR and coherence tracking applied to glaciology: the example of Shirase Glacier,” Proceedings
of FRINGE’99, Liège, Belgium, 1999.
[9] Strozzi T., A. Luckman, T. Murray, U. Wegmüller, C. Werner, Glacier motion estimation using SAR offset-
tracking procedures, IEEE Trans. Geosci. Remote
Sensing, Vol. 40, No. 11, pp. 2384-2391, 2002.
10] Santoro M., U. Wegmüller, T. Strozzi, C. Werner, A. Wiesmann and W. Lengert, “Thematic applications of
ERS-ENVISAT cross-interferometry“, Proc. IGARSS'08, Boston, 6-11 Jul. 2008.
[11] Werner C., U. Wegmüller, T. Strozzi, and A. Wiesmann, “Precision estimation of local offsets between SAR SLCs
and detected SAR images”, Proc. IGARSS 2005, Seoul, Korea, 25-29 Jul. 2005.
[12] N. B. D. Bechor and H. Zebker, “Measuring two-
dimensional movements using a single InSAR pair”, Geophysical Research Letters, Vol. 33, L16311, doi:10.1029/2006GL026883, 2006.