Gradient search and bilinear resampling not working as expected.

[simonrp84]

Describe the bug The gradient_search resampling method accessed through satpy is incorrectly giving a Source and destination areas are identical. error when resampling GOES-16 ABI data. If I use the bilinear method instead then the resulting image has a significant amount of speckle (black, missing, pixels) even if I set the radius_of_influence to an exceptionally high number.

To Reproduce

from satpy import Scene
from glob import glob

scn = Scene(glob(f'{some_goes_dir}/*.nc'), reader='abi_l1b')
scn.load(['true_color_nocorr'])

scn2 = scn.resample(scn.coarsest_area(), resampler='gradient_search')
# The above will produce an error.

scn3 = scn.resample(scn.coarsest_area(), resampler='bilinear', radius_of_influence=100000, fill_value=0)
scn3.save_datasets()
# The resulting image will have many black missing pixels.

Expected behavior I expect gradient_search to recognise that the ABI channels are at different resolutions and hence require resampling, as used to happen. I expect bilinear to produce a smooth output image with no obvious missing pixels.

Environment Info:

• OS: Windows and Linux
• Satpy Version: Latest main
• PyResample Version: Latest conda (1.25.1)

[lobsiger]

@simonrp84 maybe some problems are related to what I see when testing 'bilinear' and 'gradient_search' for LEOs:

https://groups.google.com/g/pytroll/c/8Fpo9QszZyc

?