Implement FCI "Flames" fire visualisation

[ameraner]

This PR implements an FCI composite blending the True Color with a fire visualisation on top. To select the fire pixels to be colorised, a simple fire mask compositor is implemented.

WARNING: this is just a first attempt at this kidn of visualisation. The fire detector employed here is handcrafted, extremely simple and is prone to false alarms and missed detection. It is designed just for PR-like visualisations of larger fires and not for any other (operational fire detection) usage.

The implementation of the compositor is generic so that it can be used by other instruments after tuning the according parameters, although it has only been tested with FCI.
The parameters are tuned to be used with FDHSI+HRFI inputs (500m or 1km composite output)

• Tests added
• Fully documented

Example code to generate an FCI image:

from satpy.scene import Scene
from glob import glob

filenames = glob('/yourdatafolder/*20240916*_O_0080_*')  # remember to pass both FDHSI and HRFI files to the Scene for best results!
scn = Scene(filenames=filenames, reader='fci_l1c_nc')
scn.load(['true_color_flames_with_night_ir105'], upper_right_corner='NE')
scn_r = scn.resample(scn.finest_area(), resampler='native', reduce_data=False)
scn_r.save_datasets(writer="simple_image", base_dir="/outdir/")

Some example images:
Portugal September 2024 wildfires:

Wildfires in southern Africa:

Attica Fire on 11.08.2024:

Etna Eruption on 04.08.2024:

[coveralls]

Pull Request Test Coverage Report for Build 15437805336

Details

• 35 of 35 (100.0%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.002%) to 96.358%

---

Totals
Change from base Build 15437601707:	0.002%
Covered Lines:	55930
Relevant Lines:	58044

---

💛 - Coveralls

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 96.25%. Comparing base (a966efe) to head (0bb743a).
Report is 4 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3129   +/-   ##
=======================================
  Coverage   96.25%   96.25%           
=======================================
  Files         429      429           
  Lines       57773    57808   +35     
=======================================
+ Hits        55608    55643   +35     
  Misses       2165     2165           

Flag	Coverage Δ
behaviourtests	3.80% <5.71%> (+<0.01%)	⬆️
unittests	96.34% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[djhoese]

Neat! I'm not an expert on stuff like this, but the images look cool and the code seems simple enough.

[pnuu]

Very nice (and surpisingly simple!) composite!

But is this a "spectral" composite comparable to SpectralBlender and HybridGreen composites? In #3123 (comment) I went through the currently existing compositors in composites.__init__.py and tried to categorize them. From those (and the existing modules) I think this would go in the "other" category.

[pnuu]

Wasn't there a recommendation from the RGB workshop to name all day-only products with a day prefix?

[ameraner]

the fire overlay works day&night, so also the background composite is day&night, so we can probably leave it like this? Naming of complex composites gets ugly quite quickly...

[ameraner]

But is this a "spectral" composite comparable to SpectralBlender and HybridGreen composites? In #3123 (comment) I went through the currently existing compositors in composites.__init__.py and tried to categorize them. From those (and the existing modules) I think this would go in the "other" category.

I put it in spectral as it is using spectral tests, but yes indeed it doesn't fit well with the others. So I can move it to __init__ and I leave the new recategorisation to you. It could also be the first of a "Feature Detection" compositor class maybe...

[mraspaud]

Here is the VIIRS version

[ameraner]

I moved the compositor to the init as suggested by Panu

[mraspaud]

Alternative blending and colorscale (will be another PR)

[mraspaud]

Just a suggestion to fixe typos