Especially if we are interested in small
deformation signals, the atmospheric
signal can be orders of magnitude larger and
therefore completely mask the deformation signal.
Time series approaches in radar interferometric
processing try to reduce this problem by estimating
the APS or, equivalently, by averaging out its
effect. Procedures can be more or less advanced,
but are all based on the fact that the APS is
uncorrelated between acquisitions with a temporal
baseline of 1 day or more, and that it is spatially
correlated, following a power-law function (Hanssen,
2001).
All estimation or averaging procedures have
in common that a larger amount of available
images results in a better mitigation of the
atmospheric signal. In fact,
the atmospheric signal is the main driver for
the necessity to use large quantities of data.
Consequently, for large stacks of data, say,
more than 80 images, one could argue that there
is no need for additional attempts to obtain
quantitative information on the APS. However,
there are still many areas in the world where
the size of the data archive is far less. Moreover,
if a new satellite such as Envisat
is launched, it may be that the phase history
of persistent scatterers cannot be continued.
In these cases, a new time series needs to be
established. As a result, it will take 2-3 years
until sufficient acquisitions have been acquired
before an independent time series analysis can
be performed. For these reasons, it is interesting
to evaluate alternative techniques
for (i) APS estimation and correction or (ii)
APS variability estimation.
The Medium Resolution (~300 m) Imaging Spectrometer
(MERIS) on board Envisat is a passive spectrometer,
observing in a number of spectral bands, and
only at daytime. Using a differential
absorption method, the ratio between
the measurements in the WV–transparent
window at 885 nm and the nearby WV-absorption
window at 900 nm is applied to estimate water
vapor measures. Without cloud cover, these cover
the full vertical column. In the case of cloud
cover, the measurements reflect the amount of
water vapor in the column between the Top of
Cloud (TOC) and the Top of Atmosphere (TOA),
since cloud droplets obscure information from
below the TOC.
It is a unique and spectacular combination
to have a medium resolution spectrometer and
a microwave instrument onboard. The
potential of synergetic use of both instruments
is very high, for example in ocean applications
where wind can be observed by the SAR and water
vapor and cloud systems by the spectrometer.
Most important here is that both observations
are acquired at exactly the same time and that
the area is imaged from the same imaging viewpoint.
Here we will investigate such a synergistic
application for atmospheric phase screen (APS)
estimation in radar interferograms.
................................
Objective
The major objectives of using MERIS-derived
water vapor products are, in order of priority:
1. to automatically derive
an APS estimate in the SPN processing,
2. to perform an independent
APS correction to the radar interferograms,
and, if possible,
3. to construct a stochastic
approximation of the significance of the APS
related to a specific
acquisition date.
In the framework of this project we focus
on the first objective; derivation and validation
of APS estimates from MERIS observations to
be used in the SPN processing. The main challenge
is therefore to determine whether it is possible
to retrieve valuable APS estimates from MERIS.
Two test sites are chosen for the combined analysis
of ASAR and MERIS. The area of the Netherlands
was chosen since (i) the influence of topography
on the residual phase signal is minimal and
(ii) weather conditions vary significantly between
winter and summer. Moreover (iii) frequent ASAR
acquisitions are planned over the Netherlands
due to the calibration and validation facilities.
However, since cloud cover is an important
limitation in the MERIS total columnar water
vapor retrieval, it appeared that many datasets
were fully covered with clouds, making quantitative
comparisons between ASAR and MERIS impossible.
For this reason, another area of interest within
this project, the Barcelona region, was chosen
as an additional test site. Here, full cloud
cover is expected to be less frequent.
Figure 25 to Figure 27 show
the results of the MERIS WV channel observations
over the Netherlands and the simulated interferogram
that would be produced using the conversion
algorithm. It is clear to see that the interferogram
APS simulations look similar to the atmospheric
contamination in real interferograms. For example,
Figure 25 shows a simulated interferogram with
a rather large phase trend of about 3 cm from
west to east. Such gradients are usually interpreted
as orbit errors. On the other hand, the
local variability in the subset looks
similar to a real APS. The following MERIS products
are more and more contaminated by clouds and
the cloud mask is applied to distinguish clouded
from cloud-free areas. Interpretation of such
areas becomes more and more difficult. Even
though a direct comparison is impossible
in the case of significant cloud cover,
it should be noted that the MERIS data give
a reasonable qualitative indication of the amount
of atmospheric disturbance that could be expected
in the interferograms.
Figure 25: MERIS image acquired
23 April 2003. Left image shows the water vapor
content in g/cm2. Right image shows subset of
image of approximately 100x100 km, the interferogram
size. Water vapor content is translated into
delay in mm. The image is completely cloud-free.
Figure 26: MERIS acquisition
28 April 2003. Explanation similar as in the
previous figure. Cloud masked area is visible,
as well as cloud mask shadow in subset.
Figure 27: MERIS acquisition
25 August 2003.
................................
Over the Barcelona area a first attempt
to come to correction of the interferogram
using MERIS data is shown in Figure
28. The upper left figure shows the
original interferogram, with still some topography
left. The upper right image is the MERIS derived
APS, to be compared with the interferogram.
Finally, by subtracting the MERIS-derived APS
from the interferogram we obtain the residual
signal shown in the figure at the lower left.
The images show that there is a strong
correlation between the MERIS and radar APS,
although the MERIS data contain some clouds
and are partially masked. Nevertheless, the
MERIS-APS seems to capture the long wavelength
component of the signal. Of course more test
sites need to be analyzed to make further conclusions,
but this was not possible within the framework
of this project.
Figure 28: First estimate of
APS derived by MERIS subtracted from original
interferogram. Upper left, original interferogram.
Upper right MERIS APS, lower left the APS-corrected
interferogram.
