Am looking on physical based corrections :)

Dimitris,

This is really quite a challenge. The two main alternatives are: visibility data from nearby airport, if existing, or "guessing" it from visual inspection on the image transformed to apparent reflectance. If you have some images with known visibility from the same area you can compare them and get a reasonable guess for the actual image. I have seen some image based automatic estimates giving weird values, thus I prefer the visual inspection which has given fairly good results. With a reasonable guess the uncertainty of the estimate will probably be within the model sensitivity and model uncertainty to this parameter.

All the best and I hope this helps!

Why are you not using standard atmospheric conditions. Atmospheric biases are not easy to remove. 

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=true&arnumber=964986

http://terpconnect.umd.edu/~sliang/atmocorr1.pdf

http://www.iub.edu/~act/files/curriculum_vitae/dlu_webCV/atmospheric_correction_IJRS_2002.pdf

http://atmcorr.gsfc.nasa.gov/Barsi_IGARSS03.PDF

you may need erdas and envi software.

Usman

Hi Dimitris,

I hope you have found a solution for your atmospheric correction problem by now. 

In case if one don't have the AOT and water vapor data available, they can just use the built-in atmospheric models (i.e. Tropical, Midlatitude Summer, Midlatitude Winter, etc) in physical atmospheric correction methods (i.e. 6S, FLAASH and ATCOR). One can select the atmospheric model from the Table provided in the link (attached), according to the location of the study area. No matter if the image is acquired before or after 2000.

Hope this helps!

http://www.harrisgeospatial.com/docs/FLAASH.html#Select2

Thanks Majid for the answer. Indeed thats what i follow to do.