Palsar misaligned georeferencing

I’m trying to gecode some Palsar images but the result is shifted a couple of hundred meters.
In the terrain correction step I’ve always chosen WGS84(DD).

The steps I’m taking are:
Calibrate -> Multilook -> Speckle Filter -> Deskewing -> Terrain Correction

I tried without deskewing but I get a different shift in the resulting image.
What am I missing? I tried opening the result in different software but the same shift is there so I guess it is the terrain correction that I’m setting up wrong or something. Do you have any hints?

The first image is only a multilooked image with a shape file created from where the sea level is 0. It is based on SRTM DEM 1arc sec. It is created in ArcMap with GCS_WGS 84

The second one is with deskewing

The third one is with deskewing and terrain correcting

The fourth one is with terrain correction and no deskewing

I had similar problems, please see here:

This resolved it in my case.

Changing only when I deskew didn’t change much. But following the steps you posted in the other thread helped.
This is with Calibration - Deskewing -Terrain Flattening - Terrain Correction

I’ll tinker around some more to see if I can get it better because it is still a bit off.

How sure are you that the polygon is correct? This could theoretically also be a source of error.

Also if the slopes are steep at the coast, incidence angle could make a difference.

I created it in Arcmap based on the sea level from the SRTM 1arcsec DEM which is the same that the geocoding uses, it fitted perfectly on the non-transformed images but havn’t checked in other software after processing yet. I will do that soon.

It is definetely better than before but we are still talking about around 100m N/S shift.

The E/W accuracy seems good though.

The red line is the polygon based on the sea level from the DEM.
The basemap is just a standard Arcmap basemap, I guess it is quite recent.
And the SAR image is HV from 2009.
You can clearly see the shift if you follow the river in the bottom part of the image.

Although since the DEM is from 2000 and the SAR data from 2009 i guess the shift might be explained somewhat geomorphologically, the river might just have shifted a bit. But 100meters I think is too much in that time. Looking at google imagery from 2004 it had only changed a couple of meters.

I was thinking that I might just skip the georeferencing step as of now. It is actually not necessary for the analysis.
But if so, I still need to exactly overlay my small study areas over the image for instance in ENVI or ArcMap.

It works when I import my shape file AOIs into SNAP, but I can’t seem to figure out how to force ArcMAP or ENVI to think that the exported image has the “correct” coordinate system which my AOIs then can be reprojected to…
I hope you understand my problem.
Do you have any ideas of how to approach the problem?

If the quality of the orbits is bad it’s best to use SAR Simulation Terrain Correction as it ensures that things are aligned to the DEM.