Subsidence through space and time in the lake Mead area : Insights from cross-platform ERS/Envisat interferometry.
Marie-Pierre Doin(1) , Olivier Cavalie(1)
, and Cecile Lasserre(1)
Ecole Normale Supérieure,
24 rue Lhomond,
75231 Paris Cedex 05,
Interferograms from the archived SAR data set of ERS-1 and ERS-2 satellites
have been calculated and inverted to retrieve the temporal and spatial subsidence
around lake Mead between 1992 and 2001 (see Cavalie et al., this session). It was shown that the
ground motion evolution can be closely associated with the load/unload of
lake Mead water level fluctuations on a elastic or viscoelastic surface.
From 2000 to 2004, the lake level fell drastically, leading
a regional uplift of a few centimers. In order to
record this uplift, we perform a large number of cross-platform ERS/Envisat interferograms
using the JPL/Caltech Roipac software.
We discuss here the additional steps included in the data treatment,
that allow to retrieve a good spatial distribution of the exploitable
phase in the scene. The resulting ERS/Envisat interferograms, and
Envisat/Envisat interferograms are then included in the inversion to
obtain the spatial ground subsidence from 1992 to 2005.
Perpendicular baselines chosen for ERS/Envisat interferograms stand between 1300 m and
2200 m, according to the spectral shift principle (Gatelli et al., 1994).
To increase the interferogram phase coherence, the SAR synthesis
is performed with varying range frequency filtering and an optimal azimuth
frequency filtering (Colesanti et al., 2004). The best common band filter in range
that improves the local interferogram coherence varies in sign and amplitude
across the scene, depending on the local look angle.
We therefore combine the results of the various filters to obtain an interferogram
with a coherency as good as possible.
The rate of fringes in the obtained ERS/Envisat interferograms is extremely high due to the
small elevation ambiguities (large baselines) and because orbital fringes do not everywhere
compensate frequency difference fringes. We thus remove a first model of
orbital fringes, frequency difference fringes, and topographic fringes, using the SRTM DEM.
This process strongly reduces the fringe pattern. However, the SRTM DEM correction also introduces noise,
that must be filtered out in a next step by strong spatial filtering of the
differential interferogram. In mountainous areas, the rate of topographic fringes
is too large in comparison to the pixel size, and the DEM accuracy is not sufficient to
correct the interferogram. Finally, we retrieve the differential interferogram
phase only in flat or moderately steeping areas.
To continue improving the spatial
coverage of retrieved interferometric phase, we perform linear combinations of
wrapped Envisat/ERS differential interferograms and unwrapped ERS/ERS interferograms, and use
the data redundancy to select the phase with the best coherence. The residual large scale pattern
of fringes can then be removed to flatten the interferogram on the scene sides, and spatial
unwrapping is performed by bridging manually non adjacent patches.
Unwrapped and flattened differential interferograms are then included in the inversion
process described by Cavalie et al., this session.