[GRASS-user] i.atcorr with ToA reflectance

[Nikos Alexandris]

* Micha Silver <micha at arava.co.il> [2015-05-29 16:32:24 +0300]:

> I'm trying to do atmospheric correction on some landsat 8 tiles. I have 
> successfully prepared ToA reflectance (with i.landsat.toar -r ).

i.landsat.toar -r would give Radiances. Would you please confirm that
the above wasn't simply a "typo"?

And what is the type of the output? Uncorrected, or any of the DOS
corrected types?

 
> Now when I run i.atcorr, I first get the range parameters by extracting 
> the actual minimum,maximum values from each band. And I set the rescale 
> parameter to 0,1. However this warning appears:
> 
> 
> r.info -r lc81740392014220_toa_refl2
> min=0.0749609194287782
> max=0.649816591520163
> 
> i.atcorr -r input=lc81740392014220_toa_refl2 output=atcorr_test 
> elevation=ast_dem parameters=6S_lc81740392014220_b2.txt 
> range=0.0749609194287782,0.649816591520163 rescale=0,1
> WARNING: Scale range length should be > 0; Using default values: [0,255]
> Atmospheric correction...

Here, I guess that Reflectance is (correctly) the input imagery.

> 
> 
> and the band gets values much larger the the desired 0,1 range for 
> reflectance. How should this be dealt with? Is it really necessary to 
> set the range to the exact min/max values for each band?


We should set a clear set of Question and Answers, and document it in
the wiki for all of these issues.

> or can I just 
> choose arbitrary range values that cover the min/max for all the bands 
> I'm using?

Not sure.

> (BTW, the above seems to work OK if I use ToA *radiance*, and a specific 
> set of range values for each band, again based on the actual min/max for 
> that band)

I think this is correct, meaning going from DN > Radiance (via
i.landsat.toar), then from Radiance > Reflectance (via i.atcorr).


Can we compare the output of i.landsat.toar and the output of a custom
r.mapcalc formula for conversions from DN > Radiance > Reflectance? I
have too seen sometimes, for Landsat8 data, heavily skewed histograms
for Landsat8 reflectances, ranging in [0, 1] which might or might not be
ok.

Nikos