Very low O2 optical depths in UV around 52000 cm-1

[emmacturner]

Hello,

I've been looking at O2 optical depths in the UV region and am seeing some very low values (around 10E-17) just below 52000 cm-1 and also some at the top of the region just below 87000 cm-1 - see my attached figure. I am wondering if you can help me track down the source of these anomalies as I don't know if they are realistic. This is both O2 lines and continuum. I have included broadening for the Schumann-Runge O2 lines by setting EXBRD in the LNFL creation of the TAPE3 and I set IBRD to 1 in TAPE5. I tried setting a minimum optical depth using the DPTMIN which set it to a default of 0.0002. Then when I removed this setting there were less low values. FYI I still see some negative optical depths no matter what is set here. I see similar low values in both the uninterpolated and interpolated (to 1 cm-1) optical depths - see my second figure. It would be really helpful to know why these values are occurring.

Thanks!
Emma

[emmacturner]

Just a gentle reminder we are still waiting for an answer on this. Is this something that can be fixed or should we use a workaround where we apply a floor to the od values? I can provide the TAPE5 that generated this data if required.

[pernak18]

@emmacturner sorry for the delay. we're working with a skeleton crew at this point. i'm trying to bring eli into this issue as well.

[mjiacono]

Hello Emma,
Thank you for reporting this issue; we are looking into it. I am able to reproduce the issue you reported, but only to a lesser degree. It would be helpful at this stage for you to let us know which versions you used of LBLRTM, LNFL, and the AER line file and to provide us with the TAPE5 files you used with LBLRTM and LNFL. With that I can try to duplicate your result here. In the meantime, you might try setting IOD=0 and let us know what you get. In my testing, this setting, which avoids the interpolation to a requested DV and uses an internal default, did not show this anomalous activity near 52000 cm-1.
Regards,
Mike Iacono

[emmacturner]

Hi Mike,
Thanks for your help. I have tried to attach two TAPE5 files for LBLRTM and the O2 TAPE5 file for LNFL:

TAPE5_LNFL_ONLYo2_UV.txt
TAPE5_50000_TMED_1_O2PRESENTT_1_KCORR_CONT_VARRES.txt
TAPE5_50000_TMAX_1_O2PRESENTT_1_KCORR_CONT_VARRES.txt

I am using versions 12.16, 3.2 and 3.8.1 for the line file, respectively.

I had a look at the IOD value. We were using IOD=3 with IMRG=1. This produced an ODexact file which we interpolated subsequently. On looking at the documentation (below) apparently it should produce an ODint file directly? It is strange that it doesn't. When I tested IOD=0 as you suggested I got an ODdeflt file. However, on inspection this file was exactly the same as the ODexact file from when IOD=3? And hence we have not changed the interpolated output. Is there some other part of the TAPE5 that I should be changing?

[emmacturner]

Hi Mike,

I have heard back from the authors of a paper that Eli pointed me towards about errors in the Schumann-Runge O2 lines in HITRAN 2020. I assume that also means the errors are in line file 3.8.1. as I think the lines have not changed since HITRAN 2016?

https://www.sciencedirect.com/science/article/abs/pii/S0022407325003401?via%3Dihub - original paper

https://www.sciencedirect.com/science/article/abs/pii/S0022407325002912?via%3Dihub - more description of self parameter

Rafael Fernandez says and I quote:

'The transitions in these bands are predisssociative, so you only need to consider self-broadening (γself) when calculating the cross sections. You should not include neither "γair" nor "nair" in the calculation.
With this in mind, all you need to do is to get the latest HITRAN data and just multiply by two (x2) the downloaded “γself” variable from HITRAN. In the next public version of HITRAN, this will be corrected ... but until the 2020 version is available you just need to make that correction that arises from un old-format error correction when shifting between HWHM and FWHM.'

Maybe this is already happening within LBLRTM but I thought I would point it out just in case it isn't.

Hope to hear from you,
Emma

[mjiacono]

Emma,
Regarding the TAPE5 files that you sent, they did not work for me at first. I had to change the formatting of the first three values on Record 3.5 from your E15.8 to E10.3 as specified in the documentation. Since this crops the surface pressure boundary to 1013., i also had to change this value in the boundary specification records from 1013.25 to 1013.0. With these changes, I can run with your TAPE 5 files. However, both files (which use IOD=3) correctly generate ODint files for each layer, not the ODexact files you report. I only get the ODexact output files when I set IOD=2 as documented. My modified version of one of your TAPE5 files is attached. Please try a run with that and let me know what happens.

TAPE5_50000_TMAX_1_O2PRESENTT_1_KCORR_CONT_VARRES_aer.txt

[mjiacono]

Emma,
Thanks for the information from Fernandez. I'll take it up with Eli when he's back in the office next week.

[emmacturner]

Hi Mike,

I think I know why you had to change my TAPE5 file, I was running it with LBLRTM version 12.8! Which requires a different format. When you ran the amended version of my TAPE5 did you still obtain the very low optical depth values? Thanks

Emma

[mjiacono]

Hi Emma, Yes I do, but my low anomalous optical depth values are not as low as yours. We may still be doing the interpolation differently. Do you know what the DV is for your output, or can you send me a TAPE6 file from one of the problem runs?
Mike

[emmacturner]

Mike,

TAPE6_50000_TMAX_O2PRESENTT_CONT_PROFILE1_12.16_VARRES.txt

See attached, we are interpolating to 1 cm-1. There may be some differences due to interpolation in the two versions (12.8 vs. 12.16) but I think this issue is not due to the specifics of the set up but is in the core code as we are both seeing the anomalous values, which should not be there.

[mjiacono]

I've resolved why we were seeing different optical depth values. The plotting software I'm using was dropping the negative values while plotting the y-axis on a log scale. After I plotted the optical depths on a linear scale (around zero) the very low (and slightly negative) optical depths are apparent, which is consistent with your result. Two of the plots below show the low-end optical depths at layer 16 (14.8 km) and layer 48 (76.4 km) from your profile in the 51800-52000 cm-1 spectral range. The negative OD values are a bit larger closer to the surface and trend toward zero above the stratosphere. The two panels in each graph only differ in the use of the aer_v_3.6.1 and aer_v_3.8.1 line files, though the results appear the same for each. The line strengths from these line files over the same range, shown in the third plot below, don't show any obvious problem, but the negative OD values do tend to occur when the line strengths are somewhat flat or otherwise linear.

[emmacturner]

Mike.

Thanks. Coming back to the original question as to why these low values are occurring, I don't think it is the line strengths. Could you check that the self broadening parameter is being treated properly, i.e. being multiplied by two as Rafael suggests above, and if not can you provide a fix please.

Emma

[emlawer]

Emma,

I started to look at this today and thought I'd give you a brief update. I'm in the stage of a bug investigation where the confusion increases rather than decreases. First, I have a hard time understanding how any problem with a half width can lead to a negative optical depth in the line calculation. If you can think of a way that could happen, then it would help to hear what you think. If that is the case, then the issue with the HITRAN lines (which we must deal with after we resolve this issue) shouldn't have anything to do with the issue of negative ODs. This is also the case for another issue that I've discovered -- the widths we are using are being adjusted for temperature dependence despite their tdep variable being = 0 in the line file -- somehow the code thinks tdep = 1. So I have to figure why that is the case, but still I don't see how the width being incorrect due to that would cause a negative OD. The consistent negative floor for the ODs suggested to me an issue with the continuum, but a look at the Herzberg continuum module didn't seem to have any possibility of computing a negative value. However, I've so far just read the code but haven't tested the code.

So that's where I am. I plan to pick this up again on Monday.

Eli

[emmacturner]

Thanks Eli!

I only see negative OD's around the 52,000 cm-1 (min -0.5) region and again around 87,000 cm-1 (min -706) , so I wonder if it something not specific to the lines or continuum..

Emma