Hi Snap users, long time reader first time poster.
With the launch of Sentinel-1c and return to global coverage I wanted to share what I have been working on in terms of DEM creation and change detection to see if it could benefit researchers the community.
My experience is that DEMs can be using Sentinel-1 even with it’s small perpendicular baselines typically in the range of 50-100 meters. I have processed quite a few areas now and the results look pretty promising.
Below is a comparison of the DEM created at 14m resolution as compared to COP30 (30m resolution) and Scoop-18 (0.5m resolution).
My experience using Sentinel-1 can be summarized as follows:
Man made structures and exposed ground (dirt, roads) can be measured very well if you look at enough image pairs.
Vegetation does not have really any easily extracted information in forested areas.
The main sources of error seem to be atmospheric and low signal to noise in the interferograms created from small baselines if you are trying to use the full resolution of the sensor.
Dr. @ABraun, I have read your paper and many posts. I would really appreciate your thoughts.
I would be happy to share some samples or process an area for active research where a currently solution is needed.
Thank you for sharing. I am impressed by the quality of the S1 results, especially that so many artificial structures are contained with sharp edges. What kind of area is shown in your example?
Did you retrieve these results based on traditional InSAR methods?
I didn’t look at the area when taking the screenshot but looking back it is about 40 sq-km in Toronto Canada. I processed all three IWs which covers south western Ontario.
Happy to share a link with some COGs as I would welcome the review from an expert.
@grahamgarvey Thank you for the reply. Really nice results that should be shared in publication, if possible. The most remarkable is that the perp. baseline is small, but the results are nice.
Enjoy!
Thank you very much for the encouragement. A larger baseline would definitely help reduce thermal noise but I was very surprised how well it processed.
I don’t really publish but am happy to provide data for those that do if it can advance their field of study.
Snap was used for coregistration and geocoding and I created custom processor for unwrapping.
It took quite a while to come to a satisfactory result but it seems to work quite well for non-vegetated areas.
I have processed urban, rural and mining areas and am able to resolve changes beyond what can be decerned from the image alone. Happy to share with you or look at other areas you may be familiar with.
The Benin city uses a very rudimentary analysis to validate census estimates to a surprisingly precise level.
The DRC data shows the scary level of mining expansion. I have my doubts about the vertical accuracy for mines but you can definitely identify changes. It is great for finding dirt piles even in small scale mining operations.
