About InSAR atmospheric effects


I am working with an interferometric pair and I am wondering which options do I have to reduce the effects of atmospheric artifacts. From other posts in the forum I see that either I use external weather data (not a prefered option in my case) or I use time-series of interferograms for the same area.

My question is, what is the idea behind using multiple interferograms? Should I derive for example 5-6 interferograms from other pairs, stack them and do the average? I see this as a rather simple approach. I have seen other methods (TRAINS, StaMPS) but not sure if they refer to the same.

Thanks for the feedback

@ABraun @falahfakhri looking for previous posts on the forum I found:

I have a similar situation. Want to reduce atmospheric effects for dem generation. I see from papers and discussions here that several paths could be followed (correct me if I am wrong)

  • Go for PSI techniques. StaMPS mainly.
  • Use external weather models
  • Average interferograms to reduce atmospheric artificats (more basic approach)

I am interested in averaging my interferograms but from your post above I got that you did not managed fully. Could you give an update on this?

Moreover, you mention that Bperp and Btemp should be considered when averaging interferograms. Does that mean that only interferograms with similar Bperp should be included as input to create the averaged? Is that to avoid combining interferograms with different acquisition geometries, thus avoid introducing errors when averaging?


Dear colleague,

TRAIN, is the best solution as free available atmospheric correction toolbox, please see

StaMPS: Persistent Scatterer Interferometry Processing Case Study: Mexico City, Nov. 2019 - Nov. 2020

The idea simply more inteferograms return more date expansion in order to detect the displacement in any AOI, this is the first thing, on the other hand more infgams, give the facility to the user to exclude and reduce the high Bp which are more sensitive to height and ultimately get more topographic residual, and also more infgam. give more coherent determined object.,

Please have a look at this post regarding to pyrate

Hope the foregoing answers your query.

Dear @marpet , @mengdahl , I think still pyrate hasn’t been included within SNAP 8!

1 Like

@falahfakhri thanks for your input. In my case I want to reduce atmospheric artifacts for DEM generation. Would this technique still be applicable. I am not interested in removing the topographic contributions obviously

Dear Colleague,

I only referred in my previous answer the goal of multiple infgms., and one reason as I mentioned above is,

Coming back to your query,

I think yes you could do so, please read it carefully,

Please have a look at the TRAIN MANUAL

@falahfakhri thanks. Are you aware if a matlab license is required to use TRAIN? I see on the manual that software is mentined. I guess my options are either go for TRAIN or average several interferograms to reduce atmospheric contributions?

Please have a look at ,


1 Like

@falahfakhri @ABraun, reading the Interferogram averaging for DEM generation post, I see you went through the process. Some tquestiosn that come to my mind:

  • Does the images in the temporal series must be connected? Meaning, the interferograms should always share a date (T1-T2, T2-T3, T3-T4). Why is that? What would be the effect of having a disconnected timeline (T1-T2, T3-T4)
  • The temporal baseline between acquisitions is identical, but what it the time span between interferometric pairs is not the same.
  • Checking the post, can you confirm the order of steps. I see you discussed firest the option of averaging the intf BEFORE unwrapping but did not work out. Can you elaborate on this? The alternative is then to stack the unwrapped phases and then proceed with DEM generation?

Dear Colleague,

Please find below a few links within step forum are related to Sentinel 1, and DEM generation, more other you can find answer your queries as well,

Coming back to your demands,

In general and far a way from whether S1, is the best solution for DEM creation Or not!; basically creating DEM means to get an absolute elevation of the bare earth for a specific area, I’m not talking now about the difference between DTM, and DSM, and DEM,

For getting good result of the absolute elevation Tandem is the best for so, however Tandem is not available via S1, therefore less time expansion in case of S1 is the best in order to avoid any unwanted ground movement, of course it also depends on the study area whether has been subjected to any natural or an anthropogenic effect or not corresponding to time expansion between the two S1 acquisitions, that’s why the average connecting time might be is the best solution in this case in order to decrease the DEM quality error.

Let’s take this example, T = S1 image Ok,

T1 - T2 = Interferogram1,

T2 - T3 = Interferogram2,

So the time interval of the first infgm1. let’s suppose 6 days, and the infgm2 is also 6 days, ultimately the interval time between the both infgms, is only 6 days,

From your question here, I think it’s recommended for you to get back to the principles of SAR, and InSAR, order to get the basic experience of processing steps,

I hope you can use the step forum search to find out all the answers I’m sure already exists fulfill all your queries and demands.

1 Like

Thanks to @ABraun we have good paper for DEM generation - Retrieval of digital elevation models from Sentinel-1 radar data – open applications, techniques, and limitations. Just take a look.


1 Like