[GRASS-user] i.atcorr: How to interpret the result?

[Tomáš Brunclík]

Hello there,
Does anybody know how to interpret the i.atcorrresults, i.e. how to get 
actual reflectance or radiance values from the resulting rasters? I mean 
when you process a Landsat band as in the examples in the manual and in 
the wiki, you get a nice atmospherically corrected image with the pixel 
values rescaled to 0-255 or whatever you select by the**oscl parameter. 
But I do not have a clue how to convert these pixel values back to 
actual absolute unitless reflectance or absolute radiancein W/(m² * sr * 
µm) as are the TOA inputs, or how to set the parameters to compute the 
results directly in these units.
In fact the problem is I do not get the meaning of the iscl and oscl 
parameters at all. Their description in the documentation is, in my 
opinion, insufficient - because there is only stated that you can scale 
to a range, but not what the extremes of such range mean in original 
unscaled units. My first bet was, that setting both to 0,1 and using 
unscaled TOA reflectance as input would give me unscaled corrected 
reflectance as output, but this is definitely not the case looking at 
the results... Not setting them at all leads to using defaults (instead 
of not using any scaling)... Setting iscl from actual min-max of the 
input is possible, but we do not know the actual min-max of the output 
in advance, so there seems to be no way to get absolute 
reflectance/radiance as a result... The scaling seems to be arbitrary 
between scenes, means you cannot say if ground reflectance between dates 
changed or not, if it increased by 5% and so on, which is one of the 
points of doing atmospheric correction. The 6S algorithm itself is 
absolute correction method by its nature, so the tool should be also. Or 
I am missing something?
Similar questions have been asked for years (i.e. 
http://lists.osgeo.org/pipermail/grass-user/2010-December/059104.html), 
but usually without a helpful reply.
Most my experiments with i.atcorr were done in GRASS 6.4.3 - if there is 
any difference in 6.5 or 7.0 regarding the issue, please tell me.
I would be grateful for any advice or link describing how to resolve 
absolute corrected radiance/reflectance values from resulting rasters, 
because without this the usability of the tool is very limited.

Regards,
Tomas Brunclik