DEM generation methods for radar images

in my opinion, the limiting factor is the data and the technique, not the software. You cannot avoid temporal decorrelation over vegetation.

I think, if there was a technique or approach to retrieve reliable DEMs from Sentinel-1, data someone would already have applied it to the entire earth. The amount of available data is huge, but they are just not designed for it. Please someone correct me if I’m wrong.

By today, I have not seen a single peer-reviewed publication which produced a DEM from Sentinel-1 data which has a better quality than the freely available DEMs (SRTM, ASTER, AW3D30,…)

I am really interested what exactly makes people believe that this is possible. We discuss this topic at a weekly base and I don’t understand where these expectations come from. :smiley:

1 Like

I think the tutorials that start with a dem generation play a role in this mis-perception. For more than a year I tried to obtain a usable DEM from S1A/B from time to time and thought the bad results was only because I did not find the right tool chain or missed something.

In this regards, it is maybe helpful to highlight the suitable applications and Not ToDos of S1 somewhere in this forum to save sometime for new comers like me.

1 Like

I think that is a good idea - perhaps the FAQs could contain this information.

thank you for your suggestion. Can you please specify what kind of information (which is not contained in the tutorial) would have helped you to prevent your misunderstandings?

Perhaps it would be enough to mention that certain applications like S-1 DEMs or Soil Moisture are “very challenging” so that people so not expect that they can achieve meaningful results just by processing some imagery.

we included both in the FAQs some time ago already to emphasize the many constraints and challenges. But I am of course open to add aspects which make it clearer.
FAQ 10 How can I derive a DEM from SAR data?
FAQ 14 I want to extract soil moisture from SAR data

My bad, I did not read FAQ10 carefully as it already mentioned this issue.

however, at the end of this FAQ10, the first link with title " DEM generation with Sentinel-1 - Workflow and challenges" may give a false signal that generating DEM is possible and with good results.

it is possible!
But not for forest areas or image pairs with short perpendicular baselines. The tutorial already contains several “warnings” that not every image pair will result in great DEMs. But I will add some more remarks on this.
I have collected, compared and discussed all of these aspects in a paper: Retrieval of digital elevation models from Sentinel-1 radar data – open applications, techniques, and limitations

thank you for your helpful work!

and yes, now I know. My study areas do not meet those conditions to get a good dem, they are heavily vegetated and with very high moisture.

then photogrammetric approaches based on VHR optical data might be more suitable.

thanks for your suggestion! I am trying the optical approach now.

hello. When following the tutorials for DEM generation, is it a must to do the coregistration with ESD step? I keep on getting java head space .

Please check these FAQs: I’m getting the error “Cannot construct DataBuffer”, “GC overhead limit exceeded” or “Java Heap Space”. What can I do?

Thank you so much

Is it okay if on the Snaphu-unrapping step no display shows even if you checked the Display execution output box?

yes, that just refers to the print output in the tool.

Hello everyone.

can i ask if is it a must to use a S1A and S1B images to create a DEM or they can both be S1A?

If its a must, what’s the solution if all the S1B that slightly fits you S1A are offline?

S1A and S1B are for most practical purposs identical and you can use and combine their data any way you’d like.

you can order offline products as discussed here: ESA Copernicus data access - Long Term Archive and its drawbacks

Thanks. I was able to request the images i wanted.