Feature: add support for HRFI imagery in the FCI L1c reader

[ameraner]

This PR adds support for high-resolution full-disc HRFI imagery in the FCI L1c reader.

It was decided to keep the same channel names as in the FDHSI case, ignoring the added _hr inside the HRFI file channel names. This has the advantage that HRFI and FDHSI files can be used interchangeably: when HRFI data is provided together with FDHSI files, Satpy will automatically select the finest resolution version of a channel (similarly e.g. to the MODIS L1b reader). If needed, a user can manually define the wished resolution by defining the resolution kwarg in the dataset load call (e.g. scn.load(['vis_06'], resolution=1000))

The PR also refactors the FCI L1c and GEOSegmentVariableYAMLReader tests, to improve the usage of fixtures, to have clearer class names and to be able to use parametrised tests for double FDHSI and HRFI checks.

• Closes Add support for HRFI imagery in the FCI L1c reader #2273
• Tests added
• Fully documented
• Add your name to AUTHORS.md if not there already

[gerritholl]

Thanks for the good work. I left some inline comments.

I think we also need another area in areas.yaml, mtg_fci_fdss_500m. Or is this part of another PR?

[codecov]

Codecov Report

Merging #2287 (d73bdde) into main (4962f2a) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main    #2287   +/-   ##
=======================================
  Coverage   94.35%   94.35%           
=======================================
  Files         310      310           
  Lines       46554    46575   +21     
=======================================
+ Hits        43926    43947   +21     
  Misses       2628     2628           

Flag	Coverage Δ
behaviourtests	4.59% <3.45%> (+<0.01%)	⬆️
unittests	94.99% <100.00%> (+<0.01%)	⬆️

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

Impacted Files	Coverage Δ
satpy/_compat.py	69.38% <100.00%> (+2.72%)	⬆️
satpy/readers/fci_l1c_nc.py	98.23% <100.00%> (+0.04%)	⬆️
satpy/readers/yaml_reader.py	97.49% <100.00%> (-0.01%)	⬇️
satpy/tests/reader_tests/test_fci_l1c_nc.py	100.00% <100.00%> (ø)
satpy/tests/test_yaml_reader.py	99.61% <100.00%> (+<0.01%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[coveralls]

Coverage increased (+0.002%) to 94.948% when pulling d73bdde on ameraner:feature_add_support_for_hrfi into 4962f2a on pytroll:main.

[ameraner]

@gerritholl thanks for your review, I also added the new AreaDefinition for the 500m grid to the areas.yaml already in this PR now

[ameraner]

In the process of handling the different grid types coming from the FCI L1c reader depending on the file type, I also now optimised/refactored a bit the handling of the missing segments computation in the GEOVariableSegmentYAMLReader. In particular, the collection of the position of the available segments is now postponed to the get_dataset call, instead of being done early during the Scene initialisation. As this was the only part of code that was requiring information from the file during the Scene init, this opens the door now for a new optimisation PR that avoids the file opening alltogether in the FCIL1cNCFileHandler/NetCDF4FileHandler init, that would significantly reduce the time needed to initialise a Scene.

[ameraner]

The result of

from satpy import Scene
from glob import glob

scn = Scene(filenames=glob(path_to_testdata + "*BODY*.nc"), reader='fci_l1c_nc')
scn.load(['ir_105'], upper_right_corner='NE')
scn.save_datasets(overlay={"coast_dir": '/coastlines_shapefiles'})

is:

note that:

• the path_to_testdata folder contains both FHDI and HRFI data, so Satpy has automatically decided to load the ir_105 channel from the HRFI files, outputting an image with 11136x11136 px, as expected from the 1km grid (instead of the 2km grid in the FDHSI files).
• the test data from the EUM ground processors contains a proxy SEVIRI image, with some high-res LEO data put on top, in order to excercise the navigation processors on data with realistic resolutions.
• some chunks have deliberately been removed from the folder, in order to trigger and test the padding process.

From a longer geolocation test code, here is an ROI over the Canaries, plotted for both the FDHSI and HRFI case (with padding activated). Note how the geolocation is the same, with the HRFI image being at a better resolution:

[ameraner]

From my side this is ready for a second round of reviews :) One test run is failing due to the mitiff issues discussed on Slack (not concerned by this PR).

[pnuu]

Just few comments/suggestions, otherwise LGTM.