Sentinel 1 Preprocessing, Problem with terrain correction

Microwave Toolbox
1

Hi all,

We are working using Sentinel 1 imagery for classification purposes. The problem comes at the Terrain correction stage of this graph:

We used our DEM of the AOI projection set to EPSG27700 (British National Grid)

The tool works but the result is shifted several meters. The parameters selected are:

image.png1416×922 51.3 KB

The results (we can see how those lines representing field boundaries are displaced several meters):

image.png1033×1008 127 KB

Could anyone help us out?
Thanks a lot in advance!

2

The product used is:

S1A_IW_SLC__1SDV_20160420T062231_20160420T062258_010901_01055B_7E13

Regards,
Manuel

3

I too have noticed this shift when trying to project directly to OSGB within SNAP. It doesn’t seem to be product dependant.

Projecting to Auto UTM or WGS84 and then re-projecting to OSGB in GDAL produces the expected result.

Perhaps the OSGB in SNAP is not defined correctly?

4

I’m not sure if I understood you right, but the DEM must be in WGS84 for Terrain Correction, regardless of the chosen output projection.

5

I have tried using the SRTM Auto-Download (both 3-sec and 1-sec) and an external DEM in WGS84 to do the Terrain Correction.

When setting the output Map Projection to WGS84 or AutoUTM, the S1 scene is projected as expected. I can even reproject this data to OSGB using GDAL and it remains in the correct location. However, when Terrain Correcting directly to OSGB (EPSG:27700) in SNAP, there seems to be an offset that @msanabria is also experiencing.

6

Could this have something to do with the Datum?

7

Thanks a million,

Projecting to Auto UTM or WGS84 and then re-projecting to OSGB in GDAL made it work as expected.

Regarding the software, I still do not understand what could be the reason.

Thanks again.

8

I wonder if “Apply the Earth Gravitational Model” has something to do with the issue.

9

Digging this thread out, as this issue with EPSG 27700 still occurs in the wild: