I need to use the L4C processor to detect the harvest week.
Unfortunately, I cannot run this processor as no S1 coherence images are produced by the L2-S1 preprocessor (I created a topic about this issue). I think that I’ve found the reason for this: coherence is only generated for 6-day interval periods but my study area is only covered by S1A, not S1B. Hence, the revisit frequency on my study site is 12 days.
Is there a way to change the parameters of Sen4CAP so that coherence images are derived for 12-day interval periods?
There is a configuration key, found in the database [config] table, named
processor.l2s1.temporal.offset
It defaults to 6 (meaning 6 days), as coherence should be computed between less-than-a-week acquisitions.
You can set it to 12:
psql -U admin sen4cap -c “update config set value = ‘12’ where key = ‘processor.l2s1.temporal.offset’;”
but the coherence, in this case, will be less relevant and may affect the output of the L4C processor.
Also, you need to update the status of S1 products (just to make sure they will be taken into account):
psql -U admin sen4cap -c “update downloader_history set status_id = 2 where satellite_id = 3;”
After setting these, either you restart the services:
sudo systemctl restart sen2agri-services
or wait for the processing job to kick again (it is scheduled each 1 hour).
Thank you for your reply, this is very helpful.
I changed the temporal offset to 12 days. Now there is no coherence flag anymore. However the processor still doesn’t generate the coherence because the acquisitions are from the same sensor (no problem with amplitude).
Here’s an extract from the sen2agri-services journal:
sept. 16 11:58:00 ns61108343 start.sh[26077]: 2020-sept.-16 11:58:00.502 INFO [DefaultQuartzScheduler_Worker-1] o.e.s.p.s.Sentinel1Level2Job - Processing acquisition S1A_IW_SLC__1SDV_20190824T003729_20190824T003756_028704_033FD2_5A85.SAFE
sept. 16 11:58:00 ns61108343 start.sh[26077]: 2020-sept.-16 11:58:00.733 DEBUG [DefaultQuartzScheduler_Worker-1] o.e.s.p.Sentinel1L2Processor - Sentinel-1 Level2 processor received flags: AMPLITUDE=true, COHERENCE=true, OVERWRITE=false
sept. 16 11:58:00 ns61108343 start.sh[26077]: 2020-sept.-16 11:58:00.736 WARN [DefaultQuartzScheduler_Worker-1] o.e.s.p.Sentinel1L2Processor - Product S1A_IW_SLC__1SDV_20190824T003729_20190824T003756_028704_033FD2_5A85.SAFE was previously processed, but the expected results count is not good [found:2 {ids: 1499,1498}; expected: 10]
sept. 16 11:58:00 ns61108343 start.sh[26077]: 2020-sept.-16 11:58:00.736 WARN [DefaultQuartzScheduler_Worker-1] o.e.s.p.Sentinel1L2Processor - Products must come from different sensors [found S1A_IW_SLC__1SDV_20190824T003729_20190824T003756_028704_033FD2_5A85.SAFE, S1A_IW_SLC__1SDV_20190812T003729_20190812T003756_028529_0339C7_2786.SAFE]
The way to generate this was already described above (12 days is the interval between two products of the same sensor). Unfortunately, in this moment it is working only if the sensors are both S1B. Please note that this mechanism was not fully implemented/tested/used as the Coherence at 12 days did not presented any interest for any of the L4x products.
If you prefer, we could provide you a modified version of the services that removes the S1B limitation but using these products will limit any further assistance concerning the L4x results obtained with them as inputs.
Thank you for your reply.
I understand that the coherence at 12 days has limited interests for weekly analyses, but in my case I have no other choices because S1B doesn’t cover my site.
I would be very grateful if you could provide the version of the services without the S1B limitation.
Many thanks in advance!
Just a quick note to the 6 vs. 12 days coherence. The L4C processor is designed to work with 6 days coherence. Consequently all the assessment is based on weekly analysis of relevant markers. So you may use the system to generate the 12 days coherence datasets (if the S1B limitation is removed) but the further harvest analysis shall be done outside the system using your own procedures that take into account all the impacts of longer coherence interval.
Thank you for your message.
We are indeed aware of this and are thus developping our own harvest analysis script. However, I’d like to let you know that the L4C in its current state works with 12 days intervals S1 data. It detects the harvest for a large amount of parcels – not all of them but I haven’t tested a lot of parameters yet – and very consistent harvest dates.