[GRASS-user] i.landsat.toar comments have only 0 value

[Nikos Alexandris]

Daniel Victoria wrote:

> Niko,

(thanks for paying attention to details, correct, in Greek you have to omit 
the final "s" when you call someone by its name, the vocative case!)

 
> Attached goes a screenshot showing my Grass version (6.4.3) and the
> correction methods available in i.landsat.toar
> Noticed the "corrected" method option.


I think I recall (now) seeing it also in the past (in the 6.4.x series). 
Hemm... ?  Not present now in

g.version -rg
version=6.4.4svn
revision=57976
date=2013
libgis_revision=50937 
libgis_date="2012-02-25 15:14:51 +0200 (Sat, 25 Feb 2012) "

here.


> Also, you can see in the GIS Layer Manager the output of r.info of one
> of the reflectance images. Notice how all values are equal to zero...

This must be (one of) the old buggy version of the code -- otherwise, there is 
no reason to get all-zero except, I guess, if you feed all-zero data.

 
> This is what the manual says about the "corrected" method:

(Below ignore most of my comments, I just repeat things for my own practice.)

So, this is for 6.4.3 
(<http://trac.osgeo.org/grass/browser/grass/tags/release_20130727_grass_6_4_3/imagery/i.landsat.toar/description.html#L63>):

> ----
> Corrected at-sensor values (method=corrected)
> At-sensor reflectance values range from zero to one,

right, it's a unitless ratio (see Wikipedia or else on Reflectance -- there is 
also an old post in which I asked Yann Chemin about it...)


> whereas at-sensor radiance must be greater or equal to zero.

right, Spectral Radiance is  W · sr^−1 · m^−3  or  W · m^-2 · sr^-1 · μm^-1  
and it's normal to have values ranging above 1.


> However, since Lmin can be a negative number then the at-sensor values can
> also be negative. To avoid these possible negative values and set the
> minimum possible values at-sensor to zero, this method corrects the radiance 
> to output a corrected at-sensor values using the equations (not for thermal
> bands):

> radiance = (uncorrected_radiance - Lmin)
 
> reflectance = radiance / sun_radiance

> Note: Other possibility to avoid negative values is set to zero this
> values (radiance and/or reflectance), but this option is ease with
> uncorrected method and r.mapcalc.
> -----

It was removed in r57858:

<http://trac.osgeo.org/grass/browser/grass/branches/releasebranch_6_4/imagery/i.landsat.toar/description.html?rev=57858#L15>

?

Browsing more of the source code, I found a comment: "Deleted in 2013".  Let's 
ask Jorge :-).

Nevertheless and in general, I think, we require Spectral Reflectance values 
to practice remote sensing. So having Radiance values, in a way "corrected 
Radiances" to do away with issues arising from "negative" values, doesn't mean 
you'd want to use them directly. You'd probably need to convert them to 
Reflectances, right?

By the way, a nice overview is the presentation entitled

"Calibrated Landsat Digital Number (DN) to Top of Atmosphere (TOA) Reflectance 
Conversion" by Richard Irish.

Greetings, Nikos

[..]