Harvest detection: Different results on subsequent runs


We did 2 different L4C RUNS at the end of the season 2023 aiming to detect different agricultural practices.
For you information, we did those RUNS after the defined harvest end period.
We have noticed that for some parcels , we get different results depending on the run ( for example: there is a Harvest detected on one of the tehm but not on the other one).

Knowing that both runs were performed after the defined harvest period and that the system reports the first Harvest detected date within the configured period, we are surprised to get different output on the same parcel.
Neither the geometry nor the L4C parameters changed between both runs.

What could have happened?

For your information , we are using SEN4CAP 3.2.

Thank you!

Normally, the results are always the same if the inputs do not changed.
This means that if you have different results at different moment of times, probably some additional input products were pre-processed between the two executions (maybe some L2A were pre-processed by MAJA or some S1 were pre-processed by the services). I assume the parcel geometries did not change between the two executions.
You can check the “product” table for new L2As or S1 AMP or COHE products that have insertion dates between the two executions. You can also check the files in /mnt/archive/marker_database_files/mdb1 for your site and season/year to see also changes in the timestamp of the marker files.

Hope this helps.

Best regards,