Confusing ambiguity in the "Generating DEM" Chapter

[AlexeyPechnikov]

The "Generating DEM" chapter appears to be misleading. It does not cover the generation of Digital Elevation Models (DEMs) using Sentinel-1 radar images as expected. Instead, the content primarily focuses on downloading pre-existing DEMs for use in InSAR processing. This discrepancy can cause confusion for readers seeking information on DEM generation from Sentinel-1 data specifically.

[dsandwell]

Hi Alexey,

Would the term Assembly of DEM rather than Generating a DEM be reasonable?
One could make a custom DEM from Sentinel-1 but it wouldnot be very accurate since the baselines are short and the atmosphere error may dominate.

Thanks,

david

On Mar 18, 2024, at 9:46 AM, Alexey Pechnikov @.***> wrote:

The "Generating DEM" chapter appears to be misleading. It does not cover the generation of Digital Elevation Models (DEMs) using Sentinel-1 radar images as expected. Instead, the content primarily focuses on downloading pre-existing DEMs for use in InSAR processing. This discrepancy can cause confusion for readers seeking information on DEM generation from Sentinel-1 data specifically.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.

David T. Sandwell IGPP 0225 Scripps Inst. of Oceanography La Jolla, CA 92093-0225

Ph. 858 534-7109 Cell 858 663-9426 http://topex.ucsd.edu

Express Mail Address: 8795 Biological Grade Room 1102 La Jolla, CA 92037

[AlexeyPechnikov]

Hi David,

Yes, I think "Assembly of DEM" is a more suitable title.

One could make a custom DEM from Sentinel-1 but it wouldnot be very accurate since the baselines are short and the atmosphere error may dominate.

It works well even for a single interferogram. You can click the 'Open In Colab' button below to open the interactive example on Google Colab:

PyGMTSAR Elevation Map: Erzincan, Türkiye.

We also have the capability to detect stratified and turbulent atmospheric delays, as illustrated in this case where SBAS analysis is used to detect the atmospheric phase, and PS analysis utilizes the detected phases to exclude them, along with the topographic phase. If I remember correctly, the same 'model' phase option exists in GMTSAR sources as well too. Note that the PS results here are free from phase bias without any selection of reference points, so the approach works optimally.

PyGMTSAR SBAS and PSI Analyses: Golden Valley, CA.

And, of course, the best DEM can be produced using multiple interferograms analysis, like detecting topographic phase residuals in GMTSAR SBAS by applying linear regression for pixel phases versus baseline.

Moreover, outside of the USA, where we have access to 1m LiDAR DEMs with a vertical accuracy of 0.1m, we usually only have access to different DEM variants with a resolution of 30m and a vertical accuracy of 5-15 meters. In comparison, the atmospheric delay from Sentinel-1, which ranges from 100-600 mm (I don't recall it being larger in real cases), appears relatively small, even without any filtering.