Error in terrain-correction results around 180deg longitude

Microwave Toolbox
#1

Hi SNAP developers,

I am processing S1-A/B data covering the Ross Sea area in Antarctica. My processing graph is simple (see attached graph file) - the workflow is as follows:

Apply-Orbit-File -> Terrain-Correction -> Convert-Datatype -> Write

I am using a standard polar stereographic projection. The resulting images have an obvious fault along the 180deg longitude line - a ribbon of missing (zero) data. For example

S1A_EW_GRDM_1SDH_20190316T112545_20190316T112654_026362_02F2E4_62F9.zip

image
image861Ă—828 387 KB

The raw S1 data does not have this fault. So it appears to be the terrain-correction process that is introducing the error. I have tested this in snap versions 7.0 and 8.0.

Any help in resolving this problem would be gratefully appreciated.

Regards,
Glenn.

sen2img_single.xml (4.2 KB)

#2

Which DEM is used for the terrain correction? Maybe the gap is introduced by its resampling (you could try nearest neighbor)?
Is the “mask areas without elevation” option unchecked?

#3

It’s an old issue with terrain correction and has not been fixed yet. The issue is tracked by ticket https://senbox.atlassian.net/browse/SITBX-647

#4

Thank you for your reply @jun_lu and the link to the issue tracker.

I am somewhat dismayed that this is an old problem that has yet to be fixed.

#5

I raised the priority of this issue to the highest category as it is indeed time to fix it.

#6

The problem has been fixed. Please give it a try and let us know if the issue still persists.

#7

Thank you @jun_lu for promptly attending to this issue. How do I test your fix? Do I have to wait for the next release?

#8

You have to wait for the next release. Or if your SNAP is built from source, you can update your source code and build SNAP again. The fix should be there.

#9

The latest release contains the fix - s1tbx v8.0.1 - and produces the correct result …

image
image626Ă—603 162 KB

Many thanks for your help. A good start to the new year.

2 Likes
#10

Dear SNAP developers,

Unfortunately, it seems the bug over the Ross Sea appeared again in SNAP Version 9.0.6.
I am processing Sentinel-1A/B data over the Ross ice shelf but the Terrain Correction with the Copernicus DEM 30 m to polar stereographic projection (EPSG 3031) fails with the error: java.lang.ArrayIndexOutOfBoundsExceptionFailed to read the data: level=0, region=[x=0, y=512, width=512, height=512]. due to java.io.IOException: closed. In contrast, it does work when projecting to Geographic Lat/Lon WGS84.

I tested on single and dual polarized data on Windows (16 GB RAM) and Linux (125 GB RAM):
S1B_EW_GRDM_1SSH_20170516T110027_20170516T110132_005622_009D90_52A4.zip
S1A_EW_GRDM_1SDH_20230927T112618_20230927T112722_050512_061574_A08A.zip

For simplicity, I only used Read → Apply Orbit File → Terrain-Correction → Write
TC.xml (4.4 KB)

Any help to solve this problem is highly appreciated.

#11

JIRA ticket SNAP-3561 created