DEM generation

[AkshayGore]

@FTP Can you please send me graph builder .xml file for DEM generation.
akshaygore@live.com
Akshay Gore

[AkshayGore]

@Sara.Aparicio
Can you please send me graph builder .xml file for DEM generation.
akshaygore@live.com
Akshay Gore

[ABraun]

instead of using graphs, try to follow these steps and check if the output of each operator matches your needs.
Just applying a graph will probably end in bad results. Futhermore, the unwrapping can not be included in a SNAP graph and has to be performed externally.

1. Corregistration by using the function “S-1 TOPS Corregistration” (Sentinel 1 Toolbox)
2. “Interfarogram Generation”. In this step I’m considering the corregistered IW Swath;
3. deburst
4. “Goldstein Phase Filtering”;
5. “Snaphu Export”. I’m not applying the topographic phase removal, since I’m interest in the resulting DEM and not DinSAR,
6. “Snaphu Import” of the unwrapped phase;

Accordining to the Nest turtorial for Etna, I used:
7. Interferometrix-Products- PHASE to height
and Range-doppler creation.

[lveci]

Good advice from @ABraun.
If you still would like a graph there are some sample graph in the Graphs menu along the top of the Graph Builder.

[Ailin]

Hi all,

I have been trying to generate a DEM with two Sentinel 1A images:
S1A_IW_SLC__1SSV_20150215T095140_20150215T095208_004632_005B63_33CB
S1A_IW_SLC__1SSV_20150227T095140_20150227T095208_004807_005FA1_24DF

I made 3 tries withouth success, these are the steps:

1. Split
2. Apply orbit file
3. Coregistration
4. IFG
5. Goldstein filter
6. Deburst
7. Subset
8. Unwrap
9. Phase to elevation
10. TC

I changed some of the steps and applied ESD in other attempts but withouth good results. This is one of the DEMs I obtained.

Could someone please tell me where the mistake is?
Thanks a lot!

Ailin

[falahfakhri]

Please take a look at this

DEM creation

Do an interferogram and include coherence Apply goldstein filtering to clean it up or multilook it (uncheck output intensity) Don't apply topo phase removal if you want a DEM. Export to Snaphu Select TOPO mode for DEM generation and method MCF From Linux, run snaphu with the command in the snaphu.conf file Import from snaphu into SNAP Run Dem Generation http://step.esa.int/main/doc/tutorials/sentinel-1-toolbox-tutorials/

[Ailin]

Thanks Falah,

Those are the steps I followed, exactly. I also followed the tutorials TOPSAR interferometry and others like 2012_InSAR_Tutorial_ERS-Etna (from the web: https://earth.esa.int/web/nest/tutorials) but having in mind that I am working with Sentinel´s TOPSAR images.
I also read many of the forum conversations but could have skipped sth… That is why I ask for help, I can´t realise what is wrong.

Thanks again, regards,
Ailin

[falahfakhri]

Could you please skip the seventh’s step and try out once more.

[ABraun]

does your data look alright after debursting? I guess the deburst module left extreme phase jumps in the interferogram which then could not be unwrapped properly.

[Ailin]

Alright, will try that and let you know then, thanks

[Ailin]

Hi ABraun, this is how it looks like:

[ABraun]

is this the case after the debursting?
Did you then have a look at the unwrapped phase?
Did you chose TOPO for the SNAPHU export?

[Ailin]

Hi!
@ABraun yes, that is after debursting. I chose TOPO and MCF for export.
The unwrapped phase (right after unwrapping) looks like a nice painting :

@falahfakhri I tried to do all the process without subsetting but run out of memory twice (8GB RAM is not much). Is there a problem in subsetting the image?

[Igor]

Hi Ailin,

You can’t generate a normal DEM, when you have low values of Bperp (slide 13 from https://earth.esa.int/documents/507513/0/InSAR_Tutorial_ERS-Etna/41625d92-93d0-41e4-8c7e-c96450d82fdb – 66.39 m). In fact, the document InSAR_Tutorial_ERS-Etna presents only the order of the steps! But the scene is like there (1 day in example) to get a DEM is not used for the case Sentinel-1! Bperp should be as large as possible. Usually to build a DEM choose scenes obtained with a low Btemp and perhaps a large Bperp.

For example, on page 132 ( http://www.smu.edu/-/media/Site/Dedman/Academics/Departments/EarthSciences/PDF/Lu/098_Lu_etal_InSAR_DEM_2012.ashx?la=en ):
“Assuming we have more than about 20 repeat-pass SAR images, we can generate a stack of N coregistered multitemporal interferograms. Because our goal is to generate a new DEM or update an existing low-resolution or low-accuracy DEM, interferograms with short time separations and large baselines are preferred. This is critical to derive a high-accuracy DEM.”

So, to correctly build a DEM, you should have a lot of different scenes. And already from it it is necessary to choose a pair that best meets the criterion of building DEM.

Best regards,
Igor.

[ABraun]

“Assuming we have more than about 20 repeat-pass SAR images, we can generate a stack of N coregistered multitemporal interferograms. Because our goal is to generate a new DEM or update an existing low-resolution or low-accuracy DEM, interferograms with short time separations and large baselines are preferred. This is critical to derive a high-accuracy DEM.”

So, to correctly build a DEM, you should have a lot of different scenes. And already from it it is necessary to choose a pair that best meets the criterion of building DEM.

thanks for the input, Igor. Multi-temporal data can surely increase the quality if interferometric analyses.

I was trying to do exactly what you describe but I am no longer sure if this is possible in SNAP:

Interferogram averaging for DEM generation

I wondered if it's possible (and reasonable) to calculate multiple interferograms and average them for later phase to elevation calculation. I have Sentinel-1 A/B interferograms (6 days temporal baseline each) in an arid region with high coherence. I would like to exclude atmospheric effect by a combination of interferograms. But I struggle to do it in SNAP. When would be the best step to do the averaging? Directly on the filtered interferograms (I doubt that this makes sense) After unwrap…

Based on the publication provided by mengdahl here, interferograms should be combined after unwrapping:

Interferogram averaging for DEM generation

A quick search found this:

Does anyone have an idea how they could be co-registered in SNAP?

[aminevsaziz]

@Ailin you can try 2 higher order consecutive Goldstein filtering but with lower values typically in logarithmic scale to ensure smoothing the results as much as possible while keeping fine details but with 8 GB of RAM & large scene it would be kind of difficult also I’m not so sure if this work in stbx1 but give it a try …

[aminevsaziz]

@ABraun I’m digging a similar subject this days about averaging stacks …I would say it’s possible on paper to implement it on snap as separate enhancement but it’s not easy as it sounds.

[falahfakhri]

the main idea of subset is to reduce the scene size, in case of S1-A/B, the better is to split and choose only the bursts of your study area, still the question is raised about the coherence of your study area, take a look at this

The quality of your dem is strongly affected by coherence and atmospheric disturbances.

Take a look at this

Problem in Snaphu Import section

For DEM-generation high coherence is needed, say 0.75 and above. Even with perfect coherence the DEM will still suffer from distortion caused by atmospheric artefacts. These artefacts can be mitigated by averaging many generated DEMs.

[Ailin]

Thanks to all for your replies.

You can’t generate a normal DEM, when you have low values of Bperp

You are right Igor, I read that the optimum Bperp should be between 150 and 300 m (mine is 24m)
So I think it is not possible to build a DEM with Sentinel images. Also, the area of my work is mountainous and with low coherence values in some parts. And SNAP does not allow to work with different subswaths (eg: IW1 and IW3) so I can not combine Sentinel 1 and 2 images…

@aminevsaziz I used consecutive Goldstein filters with a subsetted part of the image but had the same results as the image I posted.

in case of S1-A/B, the better is to split and choose only the bursts of your study area,

Yes, I used one burst.

So, should I try doing interferograms averaging, or is it worthless?

[falahfakhri]

Would you please to try out the whole image if that possible following the same processing chains, and let see what is the results? In this case there is no need to subset