Improve the coherence outcome

Dear all,

I am trying to extract an interferogram and compute the displacements due to an earthquake using SNAP and a pair of Sentinel-1 acquisitions.
I am facing with some problems with the coherence.
Is the window size the only parameter I can set to try to improve it?
If I make this window smaller, the coherence is very noisy, while the final result looks better if I make this window size larger.
Can you please explain how does this step work?

Many thanks in advance!
Greetings from Italy and Rome,


A small window creates a noisy result and a biased coherence estimate at low coherence values - therefore a larger window size is recommended. It is a trade-off between resolution and better coherence estimation. We will introduce an improved coherence estimation for SNAP7.0 that uses a so-called non-local (NL) coherence estimator.

1 Like

Hi mengdhal,

thank you so much!
I tried enlarging the window size and, effectively, the coherence looks less noisy.
But the final displacements are totally not reliable: the error is around 20 or 15 cm… (after a comparison with a validated map)

What do you suggest to refine it?
I leave each parameter as default. The area of interest is Anchorage, Alaska.
Thanks again!


Hello. How do I determine the most appropriate window size?

Sir @mengdahl, when I tried changing the coherence range window to 1, I got better coherence values. But do you think it is biased?

Changing the window size leads to different coherence values and it helps to make optimal use of the coherence raster for the identification of surface characteristics or dynamics.
However, it has no impact on the quality of the phase information.