GAMMA Questions

Hello,

multi_def_pt divides the interferogram into patches on a rectangular grid. The
reason for this is that the relative atmospheric phase ad deformation increases
the farther you are from a local reference point. Ideally, you have one local
reference point for the entire image, but this requires subtraction of an
estimated atmospheric phase screen from each interferogram. multi_def_pt selects
a set of local reference points and connects the reference points into a
network. There are small errors that accumulate in  the unwrapped phase in the
reference point network such that once the entire stack of interferograms have
been unwrapped using the patching, the atmospheric phase screens can be
estimated and subtracted. The size of the patches to select depends on the point
density and the atmospheric variability. Too large a patch implies that the
atmospheric phase varies beyond where reliable phase unwrapping is possible. Too
small a patch and there may be loss of coverage due to lack of suitable
reference points, and significant error accumulation in the reference point
network. The size of the patch is on the order of 1 km, so the number of pixels
will vary based on the sensor resolution. In the the case of TSX data the larger
patch size in your case has better coverage of the scene. I would however look
at the individual unwrapped phase images for each interferogram and determine
the unwrapping quality by inspection.

Best regards,
Charles

Hi, Charles!

Point density is variable (data contains hi- and low-coherence areas,
many incoherent water and forests, inside forests - coherent areas, and
difficult choose a length spatial filter to estimate atmospheric component),
unwrapping errors also present.
So, I can expand <rpatch> up to 1 km and do not be afraid to made mistakes.
One more moment - many rainy days when data was sensing.

Thanks!

Hello,

Typically the size of the patch is between 500m and 2km in ground-range and
azimuth, mostly as a function of radar frequency.
Generally higher frequency -> smaller patches. The number of pixels/patch side
depends on the resolution. The patch should be roughly square in ground-range,
azimuth.

The first time the program is run, we are working to get an estimate of the
unwrapped phase, from the unwrapped phase we estimate the atmosphere including
the height dependent component. The initial atmospheric phase estimate is
spatially interpolated on an interferogram by interferogram basis to cover the
entire scene. It is also smoothed, then subtracted from the entire scene. It
should then be possible to work without using patches. The goal of the first
steps in processing in the IPTA paradigm is to unwrap the phase using a
combination of temporal and spatial phase unwrapping.  Once the phase is
unwrapped in all the layers, different approaches are possible including using
multi-baseline time-series analysis (using mb or mb_pt), or single-reference
approach.

Best regards,
Charles