Biophysical creates 7km NULL-border

Optical Toolbox Sentinel 2 Problem Reports
#1

Hello,
I’ve got an massive problem with biophysical processor to calculate LAI. Usage of S2-L2A sen2cor data delivers a 7km thick black NULL border around the scene.

For 10m-biophysical processor it needs to be resampled to 10m before. I’ve done this with ‘resampling’ processor in ‘raster->geometric->resampling’. The bicubic 10m-resampling works and delivers a correct scene, but if i calculate the LAI on this, i got a black 7km border.

The only way that biophysical processor supplies correct and full spatial scene calculation of the LAI, was the ‘s2-resampling processor’ in ‘optical->geometric->s2_resampling’. But i cannot build a graph for batch processing with the s2-resampling processor. It gives the info, that s2 source product (downloaded L2A ZIP-file) is ‘invalid’.

The LAI output based on ‘resampling’ processor is shown in the screenshot.
It shows the direct result of resampling->biophysical in the left side AND a reprojected scene based on resampling processor in the right side. So i think here is a big problem in ‘Resampling’ or ‘Biophysical’ processor.

Hope there is a solution to fix em.
SNAP ver 8.0.9 WinX64

SNAP_Resampling_LAI
SNAP_Resampling_LAI1753×1166 405 KB

#2

@FlorianD Can you take a look?

#3

Hi,

The invalid source issue will corrected in the SNAP 9.0.0. It will released soon.
I haven’t reproduced the black border issue on S2 L2A product from the scihub.
Please, could you share your input product?
Thanks in advance.
Best regards

#4

Hi,
The source tile was downloaded from finder.code-de.org
the filename was S2A_MSIL2A_20160604T103032_N0213_R108_T32UPE_20200708T095609 as u can see in the screenshot. It was a L2A-Sen2Cor processed file.

#5

Thanks for the feedback.
I got no black border on SNAP 9 RC. So, I assume the bug is corrected in the next major update.

image
image1724×937 245 KB

I will try on SNAP 8.
I let you know.

#6

I was wrong. The issue is also on SNAP 9 RC.
It is dependant to the interpolation method. The nearest method works, but the nearest and bicubic methods introduce the NULL-border.
I will create a JIRA ticket issue when I will identify the origin of the issue: resampler or biophysical processor.

#7

Hi,

The issue tracking ticket is here: [SIITBX-488] - JIRA
For waiting the fix, the workaround consists to use advanced method interpolation, select Nearest method for the sun and view bands and the bi cubic method for the other bands.

1 Like
#8

Thx for reply.
We need bicubic interpolation for the 20m-bands. So (bi)linear interpolating is not suitable for us.