Dem-elevation

[falahfakhri]

Dear All,

Would you please to take a look at this elevation band, and tell me what is the reason of this shift, I’m using an external DEM Resolution is 10 m

Furthermore, the following is the kmz of coherence, there is no shift, but take a look how it looks like, (processing two burst, however, when processing one burst with the same DEM the kmz has a big shipt).

[djamainajib]

Dear All,

I’m using SNAP and Sentinel-1 data to extract DEM, I want check the XY cordinate system as well as Z cordinate system of the generated elevation model.

Thank you

[ABraun]

Can you please describe a bit more where is your problem?

[falahfakhri]

Would you please to check out this topic,

Phase to elevation vs phase to height

phase to height converts the unwrraped phase to the heights in the radar coded system so you need ground points with known altitude to converts your results into real ground values. phase to elevation uses a digital elevation model in reference height calculation. This operator converts interferometric phase to digital elevation map. Both operators are correct.

[Ammar]

Dear Falah… I think this unwrapping error and I tried more time to overcome this problem, try to repeat the unwrapping process with different tile and if your images size with about 5000 pixels you can unwrap with 1 tile in rows and columns.
I have a question to you, I tried to generate DEM from Sentinel 1 different images, but my best images with coherence more than 0.7 but all my results with elevation less than my study area elevation about 100 to more than 200 m. As shows in followed picture one result

You can see also unwrapping error. I know Sentinel 1 not suitable to generate DEM with high temporal baseline and the short pier. Baseline.

[falahfakhri]

@Ammar

I think this unwrapping error and I tried more time to overcome this problem, try to repeat the unwrapping process with different tile and if your images size with about 5000 pixels you can unwrap with 1 tile in rows and columns.

Dear Amman, thanks for giving your advice, this issue is solved at the time, actually, it was part of lecture in the Agricultural Dep. Un. of Helsinki.

But coming back to your questions,

@Ammar

I tried to generate DEM from Sentinel 1 different images, but my best images with coherence more than 0.7 but all my results with elevation less than my study area elevation about 100 to more than 200 m.

I have explained the differences of elevations in created DEM Compering to SRTM and ground truth, in the following post, please take a look,

Source of the post

@Ammar

You can see also unwrapping error.

This could be phase jump,

Did you apply Enhanced Spectral Diversity (ESD)?

Please take a look at the following post,

Source of the post

Also check up the corr. accuracy, because the low corr. accuracy could be a reason of phase jump,

Source of the post

To check the corr. error or accuracy in SNAP, use InSAR Stack tool

It’s also mentioned in this post, coregistration error,

Source of the post

It’s worth looking at this presentation as well, Precise coregistration of Sentinel-1A TOPS data
Source: https://files.scec.org/s3fs-public/0129_1400_1530_Fattahi.pdf

[Ammar]

Dear Falah, thanks so much for your reply… I read all your source mention before and also I follow most your post here and download your PHD thesis, it is a great job…

@falahfakhri

This could be phase jump,

Did you apply Enhanced Spectral Diversity (ESD)?

Please take a look at the following post,

Source of the post

Also check up the corr. accuracy, because the low corr. accuracy could be a reason of phase jump,

Source of the post

I think it is not phase jump because it does not appear in phase image and I apply ESD but I told you it was wrapped error for with change parameter in unwrapping process I can overcome this error.

@falahfakhri

I have explained the differences of elevations in created DEM Compering to SRTM and ground truth, in the following post, please take a look,

For SRTM, I don’t compare resulted in DEM with SRTM only, but I compared with most free available DEM such as ASTER v.1, v.2 and v.3, SRTM v.3 with 30 and 90 m resolution and also ALOS 12.5 m All have closed average elevation only DEM generated from S1 with very different elevation. I think there are losses rings in my processes or in SNAP software at issue of DEM generated while not the problem in deformation processes.
Finally, I will more and more try to find the problem, thank you, Dr. Falah, for all your advice to me, or to all the people here and sorry for bad English written.

[Ammar]

@falahfakhri

To check the corr. error or accuracy in SNAP, use InSAR Stack tool

Dear Falah, InSAR stacks tool don’t work with coregistration of Sentinel-1 SLC-IW product why ? While it works with coregistration of ALOS-PALSAR product? It work with sentinel-1 only after I split and deburst my two images and then coregistration with GCP point , that right ?

[falahfakhri]

@Ammar

While it works with coregistration of ALOS-PALSAR product?

These are the right steps of ALOS-PALSAR coregistration and creating interferogram,

1- Select the HH or VH, or both polarization
2- Coregistration (one pair and one stack with multiple images)
3- ALOS Deskewing (for the ASF/SSARA data)
4- Interferogram Formation
5- Coherence Estimation
6-Goldstein Phase Filter (on Ifg)
7-Topographic Phase Removal (SRTM 1sec)

@Ammar

It work with sentinel-1 only after I split and deburst my two images and then coregistration with GCP point , that right ?

For S-1 Apply S1 TOPS coregistration

[Ammar]

Thank you so much dear Falah,

@falahfakhri

Yes, this right you should firstly Corr. your data, and then apply InSAR Stack tool,

But here corr. with GCP not corr. for S1TOPS? what mean stack tool?
anthor question please, are there a way to know corr. accuracy for azmiuth and for range?

[falahfakhri]

@Ammar

anthor question please, are there a way to know corr. accuracy for azmiuth and for range?

I edited the previous post, might be I have misinterpreted your question,

Concerning the corr. accuracy I mentioned within the above post,

Source of the post

[Ammar]

Thank you, dear Falah, for all your answer …and I read most post here about this matter and also your P.HD thesis.

@falahfakhri

I edited the previous post, might be I have misinterpreted your question,

I know about ALOS-PLASER and don’t have a problem, but I want to ask you how I get co-registration accuracy for Sentinel -1, as you see here I apply coregistration for TOPS with ESD but you go to InSAR stack tool in SNAP v.7 I can’t found co-registration accuracy values

But I can found co-registration accuracy values if follow steps:
1- Sentinel-1 split
2- S-1 apply orbit
3- S-1 deburst
4- coregistration with GCPS as shows in below pic.

My question, this right way to get coregistration accuracy for Sentinel-1 SLC product? or there are anther ways to get it ?

[falahfakhri]

The Corr. residual should be produced as an indicator for all corr. accuracy.

[ryeramirez]

Hi, @ABraun and @falahfakhri!

I just want to ask how to get rid of the ‘discontinuities’ marked by the black boxes from the Sentinel-1 derived DEM? Please see image below.

Below are the coherence and phase bands also.

The image pair has a temporal baseline of 12 days and perpendicular baseline of 170 m. Unfortunately, that’s the only pair I have that satisfies the 150-300 m perpendicular baseline requirement for DEM generation. There is no 6-day pair available also. My AOI is within 2 subswaths (IW2 and IW3). I’ve merged the 2 subswaths after the ‘coregistration and interferogram formation’ process (i.e., BG–>ESD–>IFG–>DEB).

Here are some additional information after the coreg and ifg process.
Goldstein Phase Filtering
-Adaptive Filter Exponent: 1.0
-FFT Size: 64
-Windw Size: 3
Snaphu Export
-Statitical-cost mode: TOPO
-Initial method: MCF
-Number of Tile Rows and Columns: 20
-Number of Processors: 4
Row and Column Overlaps: 200
Tile Cost Threshold: 500

Thank you in advance for your help.

[ABraun]

You can try two things:

1. Lower the exponent to 0.4 during the Goldstein phase filter and check if less discontinuities are left after filtering.
2. Test different snaphu settings (MST or MCF, SMOOTH instead of TOPO, different tiling…)

Unwrapping can have problems for strong phase discontinuities or in shadow areas. There’s no ultimate solution to this. So you can only try different settings and compare.

[ryeramirez]

Yup. That’s what I am currently doing now. Playing with the Goldstein phase filtering and SnaphuExport parameters.

Thank you.

[ABraun]

Yes, I think it is not a processing error, but simply a tricky topographic situation with respect to the looking direction.

[ryeramirez]

I see. Unfortunately, I do not have any ascending images for the AOI, only descending images with one data pair satisfying the 150-300m perp baseline requirement.

FYI, the upper middle portion (with white to yellow colors) is actually a forested area. I hope to have the appropriate set of parameters to at least produce a reasonable S1 derived DEM.

[ryeramirez]

I tried to lower the number of tile rows and columns (i.e., 5 and 1) few days ago while waiting for your suggestion but it always returns an error. I am stuck with 10 for both. I am relying now on the statistical-cost mode and initial method for the improvement if there is.

[tarik44]

I’m trying to generate DEM in Snap software with Cosmo SKYmed images.

My datas have three difeerent temporal baseline (short, middle and long time).

I have made all processing step for DEM generation (coregistration-interferogram-goldstein phase filter-snaphu export-snaphu import-phase to elevation-range doppler terrain correction).

When I produced DEM, image have 2 bands (elevation and elevatin_VV).

1- Which is real InSAR DEM? I think “elevation_VV” isnt it?
2- When generate DEMs for all data, short baseline data have avarage 27 m error from my real DEM , in addition, other datas elevation_VV bands have so irrelevant values. Do you think what is the problem ?

Elevation_VV bands for three data

Colour manipulations;

Short temporal baseline;

Middle temporal baseline;

Long temporal baseline;

Short baseline’s values approximately near (avarage 25m difference) to my real data but others so bad . I cannot understand why this is?

Thank you in advance for your help