Jökull - 01.01.2009, Blaðsíða 92

Sigurjón Jónsson
glacier motion. However, after 1997 ERS-1 was only
used as a backup for ERS-2 until its operation was
stopped in 2000. In addition, the precise pointing ca-
pabilities of ERS-2 failed in early 2001, as mentioned
above, and after that time ERS-2 data are of limited
use for interferometry.
I searched the ESA data archives for ERS-1/2 data
of East Iceland from three parallel ascending tracks
(tracks: 187, 416, and 144) and three parallel descend-
ing tracks (tracks: 195, 424, and 152, see Figure 1).
These tracks cover the Eastern Fjords as there is about
65% overlap between adjacent tracks at this latitude.
The amount of existing ascending data is limited and
was not ordered for this study. The amount of de-
scending data is much greater, especially from track
424, so data from this track were ordered from 1995,
1997–1999, and a few additional scenes from 1993,
1996, 2002–2003. The data were selected based on
acquisition date (summer) and the possibility to com-
bine multi-month and multi-year scenes with a rela-
tively small perpendicular baseline (<200 m) due to
the steep topography in East Iceland (Jónsson, 2007).
In addition, I submitted requests to ESA for both as-
cending and descending Envisat acquisitions above
East Iceland during summers 2004 and 2005. A total
of 44 interferograms were processed in this project of
which 23 are from descending ERS-1/2 data acquired
in 1993–1999 and 21 from Envisat data acquired in
2004–2005 (12 from descending orbits and 9 from as-
cending orbits). The time span of the interferograms
varies from one day to almost four years and the per-
pendicular baselines vary from 0 m to over 700 m.
The data in this study were processed using the
ROI_PAC radar interferometric software developed
by the Jet Propulsion Laboratory (JPL) in Pasadena,
California (Rosen et al., 2004). In my data processing
I followed a typical 2-pass processing procedure using
a simulated interferogram to remove the effects of to-
pography (e.g. Massonnet and Feigl, 1998; Hanssen,
2001). The simulation was formed using a Digital El-
evation Model (DEM) with about 25 m× 25 m resolu-
tion and precise Delft orbit information (Scharroo and
Visser, 1998). Many of the processed interferograms
are of excellent quality with nearly a constant interfer-
ometric phase in non-deforming areas while exhibit-
ing details about ground movements in several places.
Other interferograms proved to be not usable due to
interferometric decorrelation and topographical arti-
facts in the data.
Degradation in interferometric coherence or inter-
ferometric correlation, usually simply referred to as
decorrelation (Zebker and Villasenor, 1992), is one of
the main limitation of using InSAR to measure ground
deformation. Interferometric coherence is a measure
of the consistency of neighboring phase values and is
calculated for a small moving window (often 7 × 7
pixels in size) across the image and is bounded within
the interval [0,1]. There are many factors that cause
a loss of coherence. The most important is tempo-
ral decorrelation which results from changes in the
surface scattering characteristics during the time be-
tween the two radar acquisitions. Such changes can
be caused by many different processes, including veg-
etation growth, erosion by water and wind, agricul-
tural activities, and snow. Many of the processed im-
ages show poor coherence due to high-elevation snow
or due to vegetation growth. The conclusion about
coherence is that snow-free interferograms that span
less than six months can be used for detailed analysis
of small sites. Longer time-spans of up to one year
or even several years can be used for measuring and
monitoring some sites and large deposits, which do
not require detailed pixel-to-pixel analysis.
Another limitation is the DEM that has a resolu-
tion of 25 m × 25 m and was generated by interpolat-
ing digitized 20 m contours of 1:50000 maps from the
National Land Survey of Iceland. Differential inter-
ferometric analysis revealed significant topographic
residuals in the interferograms when the baselines
were longer than about 300 m. A 30 m DEM error will
result in a 1-fringe error in a 300 m baseline interfer-
ogram. Although better accuracy from interpolating
20-m contour lines is expected one needs to bear in
mind that the contour lines themselves also contain er-
rors. Therefore, I concluded that interferograms with
baselines exceeding 200 m include too many topo-
graphical artifacts to be reliable for deformation mea-
surements. Unfortunately, this excludes many of the
processed interferograms from the analysis.
92 JÖKULL No. 59