I assume that you are working in the infrared?  If this is the case, then the AER Line-by-Line Radiative Transfer Model (LBLRTM) is the most flexible and accurate IR RTM model that I am aware of:

http://rtweb.aer.com/lblrtm_frame.html

Rodolfo GuzziThe best way is to use a 3D rt model. I suggest

http://code.google.com/p/i3rc-monte-carlo-model/

for more information see also Radiative trasfer Solution Techniques Encyclopedia of Remote sensing

http://www.springer.com/new+%26+forthcoming+titles+%28default%29/book/978-0-387-36698-2

It would be helpful indeed if you could specify at least the spectral range of your interest. A 3D rtm might be a very good tool but doesn't guarantee a spherical geometry.

I would recommend DISORT, which comes in a plane-parallel and in a pseudo spherical version (gives you an idea about the error you introduce by ignoring spherical shape). Like SBDART it is an one-dimensional model. If it really fits depends on your other requirements.

Hope that helps

Rolle

you can also try the TUV program that was developed at NCAR by Sasha Madronich et al. to calculate photosynthesis rates over the Solar UV and VIS range. However, it also provides the output for the integrated or spectral fluxes. One can choose a simple plane-parallel model and also the DISORT type.

http://cprm.acd.ucar.edu/Models/TUV/Interactive_TUV/

Two models that I know are SCIATRAN and ARTS which can be used for radiance calculations at high solar zenith angle.

http://www.iup.uni-bremen.de/sciatran/

http://www.sat.ltu.se/arts/

I am not sure but libRadtran also has option for pseudo spherical radiative transfer solution and hence can be used at high solar zenith angles. 

http://www.libradtran.org/doku.php?id=basic_usage

Yes, this is true. Actually libRadtran features a number of models (called solvers there) and the pseudo spherical DISORT is part of it.

Maybe this would serve your needs...

@Ronald Scheirer: Hi Rolle, thanks for advertising libRadtran!

The pseudo-spherical disort (called sdisort) in libRadtran includes a correction for the direct solar beam in spherical geometry and can therefore be used for low sun. However, be aware that this is only an approximation of spherical geometry since the diffuse radiation is still calculated in plane-parallel. However, libRadtran also includes a one-dimensional Monte Carlo solver (rte_solver mystic) which includes fully spherical geometry if you switch it on using mc_spherical 1D. This solver does not use any simplifying assumptions and can be used as a benchmark for the sdisort calculation. The only disadvantage of the Monte Carlo solver: It is slower than sdisort and the result is a bit noisy - depending on the number of photons (or computational time). Try it out!

I owe my sincere gratitude to  Nick Nalli, Hugo Abi Karam,  Rodolfo Guzzi, Ronald Scheirer, Irina Petropavlovskikh, Harish Gadhavi and Bernhard Mayer for all your enlightening answers, which are really helpful to me, and I am going to do some in-deep investigation according your suggestions and hope to get some inspiring findings.

You can consider using LIDORT or VLIDORT (http://www.rtslidort.com/mainprod_general.html) for your requirement.

Hope this helps,

Madhavan

Good Answer Yongbo Zhou