Difference ‘source sence width’ and ‘source scene height’ after making interferograms

Dear all,

I made two interferograms (figure1), one for 24th March 2015 and another one for 24th March 2015. Both pairs are bistatic mode in TanDEM-X.


Then as both interferograms cover my area of interest, then I used ‘spatial subset from view’ and subset areas with same ‘geo cordinates’ but when I look at images then I understood ‘source sence width’ and ‘source scene height’ have very difference.

Figure2. properties for 24th March

Figure3. properties for 29th March

What can I do now to make them equal?


were these the ones with largely different perpendicular baselines?
I guess, the slant geometry is different because of different incidence angles as well. Coregistration of both datasets will probably fail. Is it an option to terrain correct both interferograms (or the information derived from them) and stack them afterwards?

Dear @ABraun
Baselines are in below:

Effective Baseline : 566.3078533580203
Along Track Baseline : 137.8206848948233

Effective Baseline : 566.8990533000549
Along Track Baseline : 122.2526791230573

I did terrain correction for both and then used collocate (figure1). Numbers are correct (figure2,3) but result looks strange!!! (figure1)


figure2. 24th March (height map after terrain correction and collocate them)

figure3. 29th March (height map after terrain correction and collocate them)

I see no error here. The first data simply has a different footprint, thereby covering a larger area. But as long as the images are not distorted towards each other (I don’t think this is the case), everything should be fine.

If first data has bigger coverage, then how both are in same geo cordinates?

The collocation tool brings both into one product of common extent, which is necessarily the maximum coverage of both inputs (=union).
But while some pixels contain data in one image, they are left empty (nodata) in the other.

You might get some regions at the bottom of the scene which are covered by one image only, but I see nothing wrong with this. Or do I misunderstand your problem?

Actually,there is something weird in these two pairs.

After making iInterfrogram, then I tried to do multilooking and I ticked ‘GR square pixel’ for both days but:

For 24th March:

It was atomically ;

Number of range look:1

Number of azimuth look:1

While for 29th March:

It was atomically ;

Number of range look:3

Number of azimuth look:2

So I did not choose any of them and I ticked ‘independent looks’ and chose for both days 3 by 3.

I do not know I did right or not?

And I do not know why when I used ‘GR square pixel’, the automatically chosing for two pairs was different?

the number of looks strongly depends from the incidence angle and the ratio between azimuth and range resolution. It is probable that the products are different regarding this.

But incase you applied independent looks, you created rectangular pixels which also might explain the difference between your images.

Try to find a resolution that works well for both while still contains square pixels. It’s even sufficient if they are close (5 and 7 meters) because the collocation resamples both to the resolution of the master image.

But I still see no need for multi-looking or collocation at an early stage if you simply want to derive height changes. Produce unwrapped interferograms at the native resolution in slant geometry and then convert the unwrapped phase to elevations. Terrain correct both (uncheck to mask out areas without elevation) and then use collocation as the last step. This brings both elevations into one product and you can subtract them in the band maths to highlight changes.

1 Like