Page 1 of 1

Support for netCDF I/O?

Posted: Thu Jan 24, 2019 11:55 am
by peterukk
Hello,

this software seems very interesting, thank you for providing access to it. To facilitate broader applications, have you considered adding support for arbitrary atmospheric profiles defined in netCDF files? In my field (meteorology/climatology) this is the standard. This should be easy to implement since the program is in python, which works great with netCDF.

Re: Support for netCDF I/O?

Posted: Fri Jan 25, 2019 9:36 am
by fsteinmetz
Hi Peter, and thank you for your interest in SMART-G.
If your NetCDF atmospheric profiles have the same content as the AFGL files, it should be quite straightforward. However if you want your custom atmospheric profile to define also the aerosols, it may be more complex.
Could you send an example of such atmospheric profile ?

Re: Support for netCDF I/O?

Posted: Thu Apr 25, 2019 3:46 pm
by peterukk
Hi,

I had forgotten about this, sorry! Not sure if SMART-G is suitable for my needs, I'm right now interested in accurate & computationally efficient calculation of optical depths in clear-sky atmosphere. It was the GPU support that caught my interest, as I have access to some. Anyway, my profiles do not have aerosols, so this is not a problem. I was unable to attach a netCDF file, but I'll probably have at least 7 molecular species, e.g. CH4 as well, so more than the AGL files. They are simple vertical profiles with units in mixing ratio (kg/kg).

How are the profiles actually used? I had a look at config.py, and it seems that the optical properties come from some band model in libradtran? Would it be possible to use line-by-line absorption coefficients (from HITRAN files) instead in SMART-G?

Thanks a lot!

Re: Support for netCDF I/O?

Posted: Fri Apr 26, 2019 8:58 am
by fsteinmetz
Hi Peter,
The calculation of optical depth spectral profiles is not executed on GPU, it is part of the preparatory computation (run in atmosphere.py), that are provided to the radiative transfer code. Are you interested only in the optical depth, or in the radiative transfer ?
It is possible to use line-by-line absorption coefficients from HITRAN, and there is an example provided in the demo_notebook.ipynb (High Spectral Resolution gaseous absorption using HITRAN). The example is provided for O2, and applying it to CH4 and other gases would require providing their profile, so it is not completely straightforward, but definitely feasible.

Re: Support for netCDF I/O?

Posted: Fri Apr 26, 2019 9:53 am
by peterukk
I am in the early stages of a PhD project about using machine learning to parameterize atmospheric radiation. My starting point is gas optics and not radiative transfer, and based on your answer SMART-G does not offer advantages for this part of radiative calculations. However, I may well turn to this tool later. Thanks very much for your help!