You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Quickbeam gases(): O2 line-overlap interference term mis-transcribed from Liebe (1985); W-band dry-air attenuation ~4-5x too high #159
While building a JAX translation of COSP (jax-cosp) and validating routine-by-routine against compiled COSP, we found that the O2 resonance-line sum in the Quickbeam gaseous-attenuation routine appears to mis-transcribe the Liebe (1985) line-overlap interference term.
Location:subsample_and_optics_example/optics/quickbeam_optics/quickbeam_optics.F90, line 1195 (master 5eb05e5):
with aux3 = v0(i)-f, aux4 = v0(i)+f, x = aux3**2+gm2, y = aux4**2+gm2.
Problem: the interference bracket uses the near-resonance denominator x for both terms and subtracts the mirror term. The Liebe (1985) MPM / Rosenkranz line shape is
i.e. the mirror-resonance term should be aux4/(y) and added inside the bracket (each resonance carries its own denominator; cf. Rosenkranz's o2abs: SF1=(WIDTH+(FREQ-FR)*Y)/((FREQ-FR)**2+WB2), SF2=(WIDTH-(FREQ+FR)*Y)/((FREQ+FR)**2+WB2)).
Impact (94 GHz, CloudSat): the interference correction from the 50–70 GHz band is large at W-band, so the error is significant for the gas correction:
conditions
as coded (one-way)
corrected
literature (ITU-R P.676 / MPM)
sea level, 288 K, dry
0.232 dB/km
0.054 dB/km
~0.045–0.06 dB/km
200 hPa, 220 K, dry
0.030 dB/km
0.006 dB/km
—
sea level, 288 K, q=10 g/kg
0.62 dB/km
0.45 dB/km
~0.5 dB/km
So dry-air (O2) attenuation is overestimated ~4–5x; total two-way gas attenuation in moist columns is ~15% high, and more in cold/dry air (upper troposphere, high latitudes), which biases simulated attenuated dBZe low along long dry paths.
(The same commented-out expression appears in the H2O sum at line 1218, but it is inactive there.)
Happy to open a PR if useful. Found during systematic validation of the JAX port against the compiled Fortran; the port's corrected form is anchored to the W-band dry-air literature value.
Following up on this line-shape function issue, the gaseous attenuation routine appears to return one-way attenuation, as defined in Eq. (5a) of Liebe (1985) (Eq. (18a) also suggested that the expression is for one-way attenuation), but the comment describes it as “two-way”.
Activity
cflsepyt commentedon Jul 27, 2026
Following up on this line-shape function issue, the gaseous attenuation routine appears to return one-way attenuation, as defined in Eq. (5a) of Liebe (1985) (Eq. (18a) also suggested that the expression is for one-way attenuation), but the comment describes it as “two-way”.