AIRS AMSUA observation converter ... Zeeman coefficients and channels

[timhoar]

The observations/obs_converter/AIRS/convert_amsu_L1.f90 ultimately requires two pieces of microwave metadata: mag_field and cosbk. The amsua_netCDF_support_mod.f90 has a section that is supposed to fill in the microwave metadata for the mag_field and cosbk variables ... It is only executed if the obs_def_rttov_nml:use_zeeman = .true.. I have no idea which channels to support for this option, nor do I know the proper values to supply.

I am really confused by the contraditions in the the RTTOV v12 User Guide Section 8.12 Zeeman effect for SSMIS and AMSU-A and the information about the AMSU-A mission from Remote Sensing Systems.

In one case, only Channel 14 is affected, in the other, Channels 4-14 are affected.

http://www.remss.com/missions/amsu/ states that:

Molecular Oxygen has a complex of relatively strong absorption lines near 60 GHz. ... The AMSU instruments are composed of 2 sub units, AMSU-A and AMSU-B. AMSU-B is a humidity sounder (not discussed further), and AMSU-A is a 15-channel temperature sounder similar to MSU. Of the 15 channels, 11 (Channels 4 through 14) are located in the 60 GHz absorption complex and thus are most closely related to atmospheric temepratures at various heights above the surface.

Which is true ... Channels 4-14 are 52.8, 53.596, 54.4, 54.94, 55.5, 57.29034, 57.29034, 57.29034, 57.29034, 57.29034, 57.29034 GHz, according to the center_freq variable in the data files from the NASA Earthdata Data Access Services website.

The RTTOV v12 user guide states that:

For microwave sensors that have high peaking weighting functions in the mesosphere such as SSMIS, channels close to lines of molecular oxygen may be significantly affected by the redistribution of line intensity through Zeeman splitting as described in the RTTOV v10 Science and Validation Report. The absorption for the affected channels will depend on the strength and orientation of the magnetic field. You must specify two input variables for the geomagnetic field in the rttov_profiles structure, these being the magnitude, Be, of the field and the cosine, cosbk, of the angle between the field vector and the viewing path considered. For SSMIS, values will be available with the satellite data stream, and will therefore already match the geographical location and orientation of the viewing path. For AMSU-A, this is not the case, but the values may be obtained from a pre-computed look-up table.

[snip paragraph about SSMIS]

For AMSU-A, only channel 14 is affected. This channel, while dominated by oxygen absorption, sounds lower down in the atmosphere than the Zeeman channels of SSMIS, and it is also located further from the oxygen line centres. The impact is therefore much smaller (~0.5K). If you run with a non-Zeeman coefficient file, all channels will use the usual set of mixed gas predictors and the Zeeman effect will not be represented in channel 14. If a Zeeman coefficient file is used, then a small set of additional predictors will be included. These will contribute for channel 14 but will be nullified for the other channels by zero coefficients.

Any help from the community would be appreciated.

[lgstarn]

The oxygen absorption line changes in the presence of a magnetic field, a phenomenon called the Zeeman effect. This really only matters for channels that peak in the mesosphere as stated. All of those AMSU/A channels are "close" to the oxygen absorption line, but precisely how close determines where they peak. A channel very, very close to the line will see saturation as soon as few oxygen molecules are encountered, i.e. very high in the atmosphere; a channel farther away needs to see a lot more oxygen molecules before saturation, so will peak lower.

I recommend to leave in support for the Zeeman effect in general but comment it out for AMSU/A and wait for a user who cares about the approximately 0.5 K impact on channel 14. I believe NWP users are unlikely to use channel 14 for data assimilation.

-JS

[lgstarn]

https://www.researchgate.net/profile/Viju-John/publication/33689746/figure/fig9/AS:669384376864769@1536605080113/Atmospheric-zenith-opacity-due-to-oxygen-and-water-vapor-for-the-microwave-frequency.ppm

Here's a graphic for a visualization. Note the logarithmic scale

[lgstarn]

And here is what that corresponds to in terms of weighting function: https://www.drroyspencer.com/wp-content/uploads/AMSU-weighting-functions.gif

[timhoar]

I don't understand how the weighting function of channel 14 is any different than channels 9-14, which all have an identical center frequency: 57.29034 GHz.

I will certainly comment out support for the Zeeman effect in general, and issue a statement to reference this issue.

[lgstarn]

While the centers are the same, they differ by the number and location of bands that are +/- averaged together.

https://www.star.nesdis.noaa.gov/mirs/amsua.php