TopSAR Processing

I’d like to start from the reference point step, I think this is different to what you are looking for, Or what is the goal behind the selection of reference point.

In here our colleague has selected a high coherent pixel and then according to the value of coherence result, he considered as a reference threshold of the coherence and hence of the deformation.

This is easy to apply, once you have created the coherence product, you could open the band math, and apply the following statement condition to create the mask, for example,

IF coherence >= 0.35 Then 1 ELSE 0

The result of aforementioned statement is a mask of only the coherence = 0.35 and higher , the other will be 0, this mask named coherence multiply by the original coherence map, will result only an interferogram of the selected coherence,

Please keep in mind, there is a difference between what is mentioned in this post and the reality behind the selection of the reference point.

An important thing it is worth mentioning here to give you an entire answer concerning the reference point, Please take a look at the following post,

Hi, falahfakhri!

Thank you for the explanation, but I am still confused. Is there a way I can communicate with you ‘verbally’?

If it’s not possible, I will make a flowchart to show you the steps I am following and please comment if where do I do/insert the additional steps you’ve mentioned.

Thank you, and my apologies for my ignorance.

@ryeramirez i was also confused about this step, what i understand from reading posts realted to deformation is that , we have to select a point which u think is not chnaged throughout the time span of your study. you can also get this value if someone has GNSS or leveling data of your study area. By comparing the data you can validate your results.


No problem, we are learning from each other, Please don’t hesitate to add your flowchart, in this case the benefits comeback to all.

Hi, falahfakhri!

I’ve uploaded here the following figures:

  1. SNAP Workflow
  2. Screenshot of my Product Explorer Panel
  3. and some clarifications on the post of @vasilisgeo88

I hope you can enlighten me on these things.

Thank you!

Figure 1. SNAP Workflow

Figure 2. Product Explorer Panel Screenshot

Figure 3. Clarifications on what to do with the post of vasilisgeo88

Please take a look at this steps in this post,

However, in your case step 9 is as below,

And step 10 is phase to displacement instead of what is mentioned in here,

Concerning the reference point, I think it well discussed in this post,

@falahfakhri, thank you for sharing those posts. I will go through the threads. Have you considered checking Figure 3 which I uploaded yesterday? For the displacement product, how do I mask those pixels with incoherent values? Could you please check Figure 3 and tell me if I understood properly the instructions by @vasilisgeo88. Thank you.

Thank you for the idea. What I did after obtaining a displacement product after applying terrain corrections was this:

  1. In my coherence band where I also applied terrain correction, I masked the pixels with coherence less than 0.90. I right-clicked on the band —> Properties —> Valid Pixel Expression (Coherence Band Name > 0.9). After doing so, I noted the information about the pixel with somehow the highest coherence specifically its position.
  2. In my terrain corrected displacement product, I looked for that pixel with high coherence and noted the value of displacement. Then I created a new band following these steps —> Raster —> Band Maths —> (New Band Name) and made the following expression (New Band Name - Pixel displacement value with high coherence).

I do not know if what I did are all correct. I am trying to understand and do what vasilisgeo88 mentioned in one of his posts.

Please take a look of what I am following in my study. I hope we can share ideas. @falahfakhri is also helping me and I am very thankful to him.

Hi! I was looking for the graph showing the effect of temporal baseline on coherence in one of the fora here but unfortunately, I couldn’t find it anymore. Can you please provide me the graph and some relevant papers discussing the effects of temporal baseline and perpendicular baseline in coherence estimation? Thank you.

In general the long temporal baseline in vegetated area, or unstable objects within same area affects the coherence correlation between two or more passes,

This matter is discussed here,

DEM creation

Here also is good explanation

Sentinel 1 per. & temp. baseline

Some example of articles,

Impact of Perpendicular and Temporal Baseline Characteristics on InSAR Coherence Maps

The effect of the perpendicular baseline on coherence for SAR images in InSAR

Also it is explained well in here


1 Like

Thank you, so much! I will go through it.

By the way, have you seen already my comments about vasilisgeo88 post? I am still waiting on your response regarding that.

I also would like to ask if it is okay to stack 2 SAR images from different frames but from the same path and with the same polarizations. Thank you.

this one? Subsidence map in 3d view

@ABraun, yes! Thank you so much! So basically, with regards to time and distance, the shorter the temporal baseline (days) and perpendicular baseline (meters) the better the coherence. Am I correct? I’ve read some papers while waiting for this graph and the findings are like that.

This is true for the temporal baseline.
The perpendicular baseline is more indifferent. For example, for DEM generation, 150-300 meters are stated most suitable.


1 Like

Thank you, again! So the effect of perpendicular baseline is indifferent! For my preliminary study, the longest and shortest perpendicular baselines were 97m and 18m, respectively. These do not fall in the range of optimum perpendicular baseline. Will it still be okay? I see some published papers with the same case I have.

For the mean time, I am more focus on the effect of temporal baseline on coherence. Below is a summary showing the relationship between the two I have obtained. Could you tell if this is reasonable? The images I’ve acquired were between October 2014 and December 2016. I have a total of 8 images creating 7 pairs. My master image is the October 2014 SAR image and the slave images have intervals of approximately 3 months from each other.

I also would like to ask if there is any recommended value for coherence to get reliable displacement estimation? Presently, I’ve set mine to 0.50 and above.

Also, can I stack images from the following situations?

  1. Same path, same polarization, different frame
  2. Same path, different polarization, different frame

Thank you, and my apologies for asking too many questions.

I explained this in here,

S-1 dataset selections

Sentinel-1 doesn’t provide data above 150 m (unfortunately), so this is okay. I’d still say, the longer the better.
Coherence image look reasonable, I think most of the decorrelation (decrease of coherence) happens during the first month.

The SInCohMap project investigates the utility of coherence for mapping landcover:
On ResearchGate:

Let me ask you beforehand, Does the P01-P20 represents the selected points? Or Do they represent the coherence of each the whole processed stacked pair?

In general, yes it is reasonable, it reflects the de-correlation of your AOI, which leads to decrease the coherence with time,

I think the threshold of coherence could be coherence >= 0.4 means the 0.4 is included, it’s also depends on the technique you’d like to implement of InSAR for instance PSI, SBAS,,

For this please take a look at the previous post,

I think this is not supported by SNAP at the moment if you mean VV-VH in case this is what you’re looking for. …

P01-P20 were selected points in my AOI. I’ve selected these points or pixels that showed high coherence (0.50 and above) from my previous set of images (from another path with same frames and polarization which is VV). But I was surprised when I used another set of images and analysis schedule. I did coherence mask and those points (almost all) were gone for temporal baseline greater than 96 days. But of course, I’ve managed to get the coherence values less than 0.50 as well as the corresponding displacement of these selected pixels. Below is my displacement vs temporal baseline summary. Zero (0) baseline is my single master image. I just do not know yet if this graph is ‘interpretable’.