satpy
satpy copied to clipboard
Add true color RGB for FCI
Add a true color RGB for FCI. Refactor the AHI green corrector compositor into its own module, because FCI needs to use the same. The green band misses the chlorophyl peak, so instead the true color RGB uses a combination between VIS 0.5 and NIR 0.8.
- [ ] Closes #xxxx
- [ ] Tests added
- [ ] Fully documented
true_color_raw
(was already supported):
The true_color
with this PR looks like:
Without the green correction, the Congo Basin looks burnt down:
For the correction, I used a ratio of 0.93 VIS 0.5 and 0.07 NIR 0.8, such as suggested by @pnuu and reported here on Slack.
These images were generated from the FCI test data, version May 2022, simulating 2017-09-20. Repeat cycle 75, covering 12:20-12:29.
Codecov Report
Merging #2181 (718a27a) into main (7a30eb8) will increase coverage by
0.00%
. The diff coverage is100.00%
.
@@ Coverage Diff @@
## main #2181 +/- ##
=======================================
Coverage 94.13% 94.14%
=======================================
Files 293 294 +1
Lines 45079 45103 +24
=======================================
+ Hits 42437 42461 +24
Misses 2642 2642
Flag | Coverage Δ | |
---|---|---|
behaviourtests | 4.68% <0.00%> (-0.01%) |
:arrow_down: |
unittests | 94.79% <100.00%> (+<0.01%) |
:arrow_up: |
Flags with carried forward coverage won't be shown. Click here to find out more.
Impacted Files | Coverage Δ | |
---|---|---|
satpy/composites/ahi.py | 100.00% <100.00%> (ø) |
|
satpy/composites/spectral.py | 100.00% <100.00%> (ø) |
|
satpy/tests/reader_tests/test_abi_l2_nc.py | 100.00% <0.00%> (ø) |
|
satpy/readers/abi_base.py | 94.52% <0.00%> (+0.07%) |
:arrow_up: |
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
Coverage increased (+0.003%) to 94.742% when pulling 718a27a1958e963e4d0f5f7b24eb70ba4bed784c on gerritholl:fci-true-color into 7a30eb871237b6e3bb997998c65f2c9c1d4e395e on pytroll:main.
the vis_08 channel is too bright, so that a factor of 0.8 should be applied to get correct radiances/reflectances
Although I don't think tuning based on buggy test data is the most productive, I've tried to replace the 0.07
by 0.056
all else being equal, I get:
which is very slightly darker and not an improvement IMHO, so I think further tuning should be done when we have real data.
LGTM!
This fails only for the experimental run now, which, as I understand it, is allowed. So the PR is ready as far as I'm concerned.