Major errors in L2 DST calculation

[olegzhelezn]

Hi David,

I noticed that some tiles in my datacube are missing substantial proportions of data in the corners. Upon closer investigation I discovered that this is due to malfunctioning DST calculation in L2 processing. The issue affected 7 thousand out of ca. half a million images.

When using FORCE v. 3.10.04-dev:::2025-12-11_08:43:39, what happens in some cases is that parts of the image closer to the corners have invalid, below-zero values of DST, which means that these image parts are then excluded from higher-level processing.

Image:

DST:

Using an earlier version of FORCE (I think v. 3.8.01, but not 100% sure anymore, since I was processing last autumn) the "corner" part was missing (NA) altogether.

I am attaching an image and a code example: https://syncandshare.lrz.de/getlink/fiSo3MPhDz3WZjgwSSfpkS/DST%20issue

[davidfrantz]

Hi Oleg,

I tested your data and could not reproduce this (with the latest dev). However, I used a minimal example as I don't have your exact data dependencies here:

• no reprojection
• no tiling
• no DEM, and no atmospheric/topographic correction
• no water vapor database
• no co-registration

Please test this by yourself, and iteratively add more options until you find the one that results in this behaviour.