Issues in topographic phase removal using external DEM


I am doing interferometry on a glacier in Svalbard, and therefore I have to use an external DEM that covers this region. I am currently using the arctic DEM, but I am having issues in the topographic phase removal step. The topography removed from the interferogram does not cover the same region as the interferogram (see the figure below).

I have seen a lot of threads discussing similar problems, but none that provide a solution that solves this problem. Any help is much appreciated.


in the TPR operator you can select to output the computed topographic phase

Please try if this looks different from the elevation of the ArcticDEM.
Furthermore, it might be worth a try to use the Copernicus 30m DEM.

The elevation band in the figure is the elevation band from TPR. I should have specified this.

Yes, this looks different from the ArcticDEM. The marker in the elevation band is covering land, while this region in the ArcticDEM is ocean. The elevation band seems to be shifted in position from both the DEM used and the SAR images.

can you please create an RGB image from the phase, the topographic phase and the intensity?

Yes, here is the RGB image. Red = phase, Green = topo phase, Blue = intensity

Which DEM was used here and does it look the same with a different DEM?
The contrast of blue is a bit low, maybe you can convert it to dB before the RGB so we can see if there is a shift between the there bands

Here the ArcticDEM was used, converted to WGS84. I have tried with a DEM based on the ArcticDEM, and this gave similar results.

RGB with intensity on dB scale.

thank you - this does not seem right, I agree. Can you now please test how the Copernicus DEM looks like so that we can understand if the problem is related to the external DEM or generally affecting the TPR?

Yes, here are the results with the Copernicus DEM. It looks similar to the results from the ArcticDEM

@jun_lu could you please check if there is something wrong here?

I should mention that I am using ALOS-2 data.

@trukar Why do you used ALOS data ? It is my opinion that L band is better with vegetated areas, but
in your case it is only ice, I guess, so S1 should fine.

I use ALOS-2 data because the longer wavelength penetrates deeper into the snow/ice which increases the temporal correlation between signals.

It seems if the ArcticDEM and the Copernicus GLO-30 DEM give the same results, that the problem must be that the ALOS-2 data is not correctly registered to the DEM. What mode and processing level of ALOS-2 data are you using?


I’m using fine mode dual polarization and high-sensitive mode quad polarization. Both modes are SLC data.

Hi, I was wondering if you have found out anything more about this issue?

I noticed that the “geo_ref_system” is set to GRS80 in the metadata for ALOS-2. Does anyone know if this could be the reason why the topographic phase seems to cover a different area?

The GRS80 and WGS84 reference systems are almost exactly the same, within less than a meter.

The ALOS-2 fine mode is SM3 and high-sensitive mode is SM2, and those have different range and azimuth sampling. I don’t think that the SNAP software can handle a mixed interferogram between SM2 and SM3 modes. The in ISCE2 can do the range and azimuth resampling to form interferograms between different ALOS-2 modes.

If you are mixing different modes, this likely explains why the topographic correction is in the wrong location.