Maps generated from SNAP have geometric shift when shown in ArcGIS


I used GRD sentinel 1A data, use thermal noise removal - apply orbit file - calibrate - speckle filtering and terrain correction on two images on the same area. Both images are in WGS84 reference.

I output the data in .img format, and input both maps in ArcGIS. However, i found that there is a shift between the two images. Hundreds of meters i guess. In arcgis, they are shown to share the same reference, WGS84.
When output the third one in the same area, it fits with one of images.
Does anybody know what is the cause. Thank you very much!


Do you see the shift also in SNAP? You can link the cursor in two views to see if there is shift.

Hi marpet,

Thank you for suggestion. I did as you said and found out that the images align with each other in SNAP.
So it should be the problem in ArcGIS. Do you have any suggestions on how to fix that problem in arcgis?


Today, I checked in the ArcMap, and the two images, which had a shift, are perfectly aligned.
I can’t explain this … and im not sure whether the shift will occur again in the future.

I think you have .hdr files with your .img files. You can have a look at it (it’s text) if there is something obviously wrong.
Or try to export to GeoTiff and use this in ArcGIS.

Thank you marpet!

Hi everyone,

I’m encountering the same problem. But in contrast to PaulZhang, I also see the shift between the two SAR images in SNAP.

I did the same processing for both GRDH images:

  1. Apply orbit file (Sentinel precise, polynomial degree: 4)
  2. Multilook (to derive a pixel spacing of 30m)
  3. Calibration (to Beta0)
  4. Terrain flattening (using SRTM 1sec HGT)
  5. Terrain correction on 30m pixel spacing (using SRTM 1sec HGT)
  6. Linear to dB conversion

The shift is only a few pixels, so probably between 30 and 90 m, but still annoying. Does anybody have an idea what could be the issue?

Thanks, marir

maybe you should skip multi-looking and just enter the desired pixel size of 30 m at the Terrain Correction step.

Thanks for your help!
I need the multi-looking in order to get a lower spatial resolution for the terrain-flattening. Is there another possibility to decrease the spatial resolution to 30m?

if you un-check “re-grid” method you can also use lower resolution DEMs for the terrain flattening.

Thanks very much!
It’s better now. The shift seems gone.
Do I see it correctly that it automatically generates the final product using the spatial resolution of the DEM (1 sec)?

the RC terrain correction proposes the smallest resolution possible based on the input data and the available DEM. But you can also select an own resolution if you want. But increasing the value higher than the proposed won’t bring you any better quality.
Also, the resampling type manages how this resolution is achieved, bilinear is mostly the best choice.