Almost identical results are obtained when using External DEMs with different vertical reference (WGS84, EGM2008, EGM96). However, products produced with/without “Apply Earth Gravitational Model” in SAR-Simulation are misregistered by one 20m pixel east/west. We believe that externalDEMApplyEGM should not have this effect.
Use of the same Copernicus GLO30 DEMs via auto download results in images shifted one 20m line south relative to the externalDEMApplyEGM=false images, and one 20m pixel east and south of externalDEMApplyEGM=true images.
What is the correct terrain correction? We obtained a 30m product from ASF’s RTC_GAMMA which also uses Copernicus GLO30 DEMs. The ASF product matches S1TBX product using auto download but only after shifting the S1TBX product one 20m pixel east.
Also: The DEMs saved from SARSim-Terrain_Correction are shifted one 20m pixel east and south relative to Copernicus GLO30 DEMs.
Problem Summary:
- With/without EGM introduces a one pixel shift.
- Use of External DEM and Auto Download of the same DEM result with different Terrain Corrected results.
- None of the 3 cases of S1TBX Terrain Corrected results ~match GAMMA geometry.
- Saved DEMs have a one pixel and line shift relative to the original DEMs.
SARSimTCGraph.xml (4.8 KB)
FYI: Range Doppler Terrain Correction results and ASF product all match geometrically. The RDTC products are produced with: Autodownload Copernicus 30m, External WGS84 Copernicus 30m DEM without externalDEMApplyEGM, and External EGM96 Copernicus DEM with externalDEMApplyEGM. The WGS84 and EGM96 DEMs both require prior conversion since Copernicus is EGM2008 vertical reference. I used StereoPipeline 2.7 dem_geoid for the conversion.