Separate from #1130 (which fixed the hot-band Y sign) — this one survives that fix.
Running ARTS 2.6.x built from source (pyarts 2.6.19-dev, includes #1130/#1133),
I'm seeing a large BT artifact near the CO2 ν2 Q-branch when HITRAN line mixing
is on. It traces to isotopologue 4 (627); its ν2 line mixing pushes the
inter-line absorption negative across roughly 664.7–666.0 cm-1, so the BT floor
jumps from ~218 K to ~250 K.
Setup: AFGL US Standard, nadir, blackbody surface, CO2 only (no H2O/continuum),
HITRAN 2020 .par, 25 cm-1 cutoff, mode="VP_Y", BT at 0.002 cm-1.
It's the line-mixing step, not the catalog. At 665.00 cm-1:
|
BT (K) |
| LM off |
219.58 |
| LM on, all iso |
252.98 |
LM-off agrees with an independent LBL code to 0.08 K here, so the base line
shape is fine.
Filtering the relmat basedir by isotopologue pins it to iso-4:
| LM config |
BT(665.00) |
| iso-1 only |
217.62 |
| iso-1 + 2 |
219.12 |
| iso-1 + 3 |
217.91 |
| iso-1 + 4 |
252.96 |
| all iso |
252.98 |
iso-4 on its own reproduces the whole thing; iso 1–3 are fine.
At 0.002 cm-1 it's a comb, not one line: iso-4 lifts the floor between the lines
while the cores still absorb. The band intensity clearly isn't being conserved.
The lines involved are in the 627 band files S40110100001I/S40110100001II
(split into I/II), near 664.97/665.00/665.03 cm-1.
Reproducer: drop the S4*.dat band files (and their rows in BandInfo.dat /
Bandinfo_ini.dat) from the relmat basedir and BT(665.00) goes back to 217.6 K;
put them back and it's 253.0 K.
Same result in VP_Y and VP_W, so it looks like the per-isotopologue relmat
assembly/normalization for 627 rather than the full-W code. My guess is the I/II
split-band handling or a population term specific to this isotopologue. Happy to
attach a plot of the comb.
Separate from #1130 (which fixed the hot-band Y sign) — this one survives that fix.
Running ARTS 2.6.x built from source (pyarts 2.6.19-dev, includes #1130/#1133),
I'm seeing a large BT artifact near the CO2 ν2 Q-branch when HITRAN line mixing
is on. It traces to isotopologue 4 (627); its ν2 line mixing pushes the
inter-line absorption negative across roughly 664.7–666.0 cm-1, so the BT floor
jumps from ~218 K to ~250 K.
Setup: AFGL US Standard, nadir, blackbody surface, CO2 only (no H2O/continuum),
HITRAN 2020 .par, 25 cm-1 cutoff,
mode="VP_Y", BT at 0.002 cm-1.It's the line-mixing step, not the catalog. At 665.00 cm-1:
LM-off agrees with an independent LBL code to 0.08 K here, so the base line
shape is fine.
Filtering the relmat basedir by isotopologue pins it to iso-4:
iso-4 on its own reproduces the whole thing; iso 1–3 are fine.
At 0.002 cm-1 it's a comb, not one line: iso-4 lifts the floor between the lines
while the cores still absorb. The band intensity clearly isn't being conserved.
The lines involved are in the 627 band files
S40110100001I/S40110100001II(split into I/II), near 664.97/665.00/665.03 cm-1.
Reproducer: drop the
S4*.datband files (and their rows in BandInfo.dat /Bandinfo_ini.dat) from the relmat basedir and BT(665.00) goes back to 217.6 K;
put them back and it's 253.0 K.
Same result in VP_Y and VP_W, so it looks like the per-isotopologue relmat
assembly/normalization for 627 rather than the full-W code. My guess is the I/II
split-band handling or a population term specific to this isotopologue. Happy to
attach a plot of the comb.