Inconsistent units of geostationary radiances

[sfinkens]

Describe the bug
Units and standard names of GOES-16, MSG-4 and Himawari-8 radiances seem to be inconsistent with the YAML definitions. The YAML definitions indicate units W m-2 um-1 sr-1 and standard name toa_outgoing_radiance_per_unit_wavelength for all imagers. However, GOES-16 radiances are in mW m-2 sr-1 (cm-1)-1 (standard name toa_outgoing_radiance_per_unit_wavenumber). MSG-4 radiances seem to be in mW m-2 cm-1 sr-1 (factor 10^7). Himawari-8 radiances are smaller than GOES/MSG by a factor of 10, not sure what unit that is.

For GOES-16 scene['C14'].attrs['units'] is correct though. Probably the attribute is overwritten with the value from the netCDF file.

To Reproduce
I compared ~10um IR images at ~09:00 local time (00 UTC for Himawari-8, 09 UTC for MSG-4, 14 UTC for GOES-16).

from satpy import Scene
import glob
import matplotlib.pyplot as plt


# GOES 16
filenames = ['OR_ABI-L1b-RadF-M3C14_G16_s20182481400355_e20182481411122_c20182481411188.nc']
g16_channel = 'C14'
g16_scene = Scene(filenames=filenames, reader='abi_l1b')
g16_scene.load([g16_channel], calibration='radiance')

# Himawari 8
filenames = glob.glob('HS_H08_20180905_0000*')
h8_channel = 'B13'
h8_scene = Scene(filenames=filenames, reader='ahi_hsd')
h8_scene.load([h8_channel], calibration='radiance')

# MSG 4
filenames = glob.glob('H*0900-__')
msg4_channel = 'IR_108'
msg4_scene = Scene(filenames=filenames, reader='hrit_msg')
msg4_scene.load([msg4_channel], calibration='radiance')

print 'GOES-16', g16_scene[g16_channel].max().values, g16_scene[g16_channel].attrs['units']
print 'Himawari-8', h8_scene[h8_channel].max().values, h8_scene[h8_channel].attrs['units']
print 'MSG-4', msg4_scene[msg4_channel].max().values, msg4_scene[msg4_channel].attrs['units']

plt.figure()
plt.imshow(g16_scene[g16_channel])
plt.colorbar()
plt.title('GOES-16')

plt.figure()
plt.imshow(h8_scene[h8_channel])
plt.colorbar()
plt.title('Himawari-8')

plt.figure()
plt.imshow(msg4_scene[msg4_channel])
plt.colorbar()
plt.title('MSG-4')

plt.show()

Output:

GOES-16 150.17288146540523 mW m-2 sr-1 (cm-1)-1
Himawari-8 11.933104 W m-2 um-1 sr-1
MSG-4 159.7228 W m-2 um-1 sr-1

Environment Info:

• OS: Linux
• SatPy Version: 0.10.0a0.dev0
• PyResample Version: 1.10.2a0.dev0

[sfinkens]

Wait, if GOES-16 radiances are given per unit wavenumber, shouldn't they be a factor of 2*pi larger than those per unit wavelength?

[djhoese]

I don't usually deal with radiances directly and not sure I understand what the units physically mean so I'm not sure I can help that much. However, if a reader has units specified in the YAML file and the resulting band has different units then that is a bug. As @mraspaud mentioned on slack we tend to use the units provided in the raw file.

In the past we tried harder to standardize the units between readers, but realized that the better answer may be to have the arrays convert to other units more easily by the person/code using it. This is something that the xarray/metpy/unyt communities are working on. Right now satpy only really prefers that reflectance data be in % (0-120ish) and brightness temperature data be in K.

[mraspaud]

So what is the status on this issue ?

[sfinkens]

I'm going to lookup the radiance units in the documentation and then update the YAML files if necessary.