The Trouble with MODIS

[lobsiger...@gmail.com]

Dear developers,

after "The Trouble with Harry, 1955" here is another Hitchcock:

Using LEO Aqua, Terra data with "generate = False" gives ugly
big pink stains for 'true_color_thin' that are absent when using
the slower "generate = True". Other MODIS composites seem to work.

I made a minimum script including Terra data that reproduces
the problem. You can download this from wetransfer with link:

https://we.tl/t-cIufngDaCJ

(issue0731.zip, 168MB)

The link is valid one week. I will open a satpy issue shortly.
The script produces the four UK images that I attach reduced.

Best regards,
Ernst

P.S.
satpy          0.37.1      pyhd8ed1ab_0       conda-forge
pyresample     1.25.1      py39h1832856_0     conda-forge

I might be offline for ?? days starting August 24th as our
provider replaces my old CATV home access with glass fibre.

[lobsiger...@gmail.com]

Dear developers,

The harder I try the more weird results I get from EUMETCast thin MODIS Aqua/Terra data. I tried different resamplers and see this:

Resampler 'bilinear' works for 'true_color_thin' with both generate=True and generate=False while 'nearest' gives only acceptable results for generate=TRUE. Interestingly I also get different results from 'bilinear' and 'nearest' if I only image/resample MODIS channel '12' (the filler channel in 'true_color_thin'). Resampler 'gradient_search' does not work with any MODIS composite or even single instrument channel. I have no problems with this resampler processing Metop-B avhrr data though.

I updated my script to easily use different resamplers with automatic naming of the results. You can download it with the respective Terra segment files from wetransfer:

https://we.tl/t-exoyS9kLiM

issue2188.zip, 250MB, link valid one week from now

Best regards,
Ernst

P.S. I also tested MODIS multiscene with different resamplers.
Basically only 'nearest' works. No problem with MetopB avhrr.
I attach three outputs of an Aqua multipass using 'green_snow'.