[GRASS-user] Re: i.atcorr returns all NULL values

[Daniel Victoria]

I'm also struggling with i.atcorr for Landsat. A friend of my
corrected some TM images using the Flash module in ENVI (based on
modtran) and I tried to correct the same images using 6S in grass
(i.atcorr). We both used the same radiance images that had some
problems with a few negative values in water areas. I set the negative
values to 0.

After correcting the TM images in Grass I created 6000 random point in
order to compare the images (flash vs. 6s). The attached graph shows
the density plot (~histogram) for each band (band 6 is actually TM
band7). For all images, value have been scaled from 0 to 1. Black line
is radiance, red is 6S and blue is flash. What we see is that for
bands 4, 5 and 7 there are no big differences but, for bands 1, 2 and
3, the correction done by flash is more pronounced.

Now, what I don't know:
1) Is this type of analysis (looking at the image histogram) OK?
2) Should atmospheric correction change the histogram?
3) For i.atcorr I get a sort of "linear" relation between radiance &
reflectance. That is, for each radiance there is only one value for
reflectance. But the flash correction gives a point cloud when I
compare radiance & reflectance (more than one reflectance value for
the same radiance). What is the expected behavior?

I'm using Grass 7 svn from about 2 weeks ago. The images are from the
Pantanal wetland in Brazil in the dry period (still lots of water
there). For 6S the settings are: Tropical atmosphere, continental
aerosol mode, 20km visibility.

Thanks and cheers
Daniel


On Fri, May 27, 2011 at 7:34 PM, Juan Benavides Duque <jbenavid at siu.edu> wrote:
> I'm sending the code I used to run the i.atcorr module (including the
> i.landsat.toar transformation)
>
> (I also used the code sent by Yann Chemin (May 23) with the same results)
>
> Of two files attached one has the code and other one the  univariate stats
> output of the band 1 TOAR raster  for one of the 1986 scenes
>
> A copy of the original landsat used have been  placed temporarily at the
> following location " assets "
>
> hope somebody can help
>
>
> thanks
>
> juan
>
>
> I'm using a desktop running Ubuntu 10.10 with a 64bit  Intel(R) Core(TM) i7
> CPU 860  @ 2.80GHz
>
>
>
>
> 2011/5/27 Gaspar Reyes Póndigo <gos47 at hotmail.com>
>>
>> I had same problem, i.atcorr not work with landsat 5TM band 1, 2 and
>> sometimes with band 3
>> I tried i.atcorr in different versions of grass, this is my results:
>>
>> 1. GRASS-6.4.1-2 under Mac Snow-Leopard 10.6.7: Not work band 1, 2, and
>> sometimes with band 3 (for years 1992, 1993, 1995 and 1997).
>> 2. GRASS-6.5. revision 46428 compiled in Mac Snow-Leopard 10.6.7: Not work
>> band 1, 2, and sometimes with band 3 (for years 1992, 1993, 1995 and 1997)
>> 3. GRASS-6.5svn45719(2011) under Microsoft Windows Xp sp3: All the bands
>> work,THIS IS VERY RARE
>>
>> My images landsat 5 TM are from Tonameca, Oaxaca, México of years 1985 to
>> 1999, and 2010 to 2011 (one image series for year), i am work with AOD and
>> visibility value,  and convert manually DN to radiance (a).
>>
>> (a) Lλ = [(LMAXλ − LMINλ) / (Qcalmax − Qcalmin)] (Qcal − Qcalmin) + LMINλ
>>
>> Gaspar Reyes Póndigo
>> Universidad del Mar, Oaxaca, México
>> Campus Puerto-Ángel
>> Carretera principal Puerto Ángel Zipolite km 1.5
>> C.P. 70902
>>
>> -------------------------------------------------------------------
>>
>> Message: 1
>>
>> Date: Fri, 27 May 2011 08:53:32 +0530
>>
>> From: "Chemin, Yann (IWMI)" <Y.Chemin at cgiar.org>
>>
>> Subject: RE: [GRASS-user] Re: i.atcorr returns all NULL values
>>
>> To: "Juan Benavides Duque" <jbenavid at siu.edu>,
>>
>> <grass-user at lists.osgeo.org>
>>
>> Cc: Markus Metz <markus.metz.giswork at googlemail.com>, Elena Mezzini
>>
>> <elena.mezzini at gmail.com>
>>
>> Message-ID:
>>
>> <001F233294EA894387CFDAADAB87DE25085063E9 at iwmix.IWMI.CGIARAD.ORG>
>>
>> Content-Type: text/plain; charset="us-ascii"
>>
>> Hi Juan,
>>
>>
>>
>> Can you please give the download link for your images,
>>
>> Could you also check the output of i.landsat.toar for them and tell if
>>
>> reflectance ranges are valid/logic.
>>
>>
>>
>> Thank you,
>>
>> Yann
>>
>>
>>
>> From: Juan Benavides Duque [mailto:jbenavid at siu.edu]
>>
>> Sent: Thursday, May 26, 2011 11:38 PM
>>
>> To: Chemin, Yann (IWMI)
>>
>> Cc: Markus Metz; Elena Mezzini
>>
>> Subject: Re: [GRASS-user] Re: i.atcorr returns all NULL values
>>
>>
>>
>>
>> I tried Yann's code for the i.atcorr on some Landsat TM5 images
>>
>> It worked very nice for images taken after 1995 but before that still
>>
>> returns null values when running i.atcorr for bands 1 and 2
>>
>> I used TM5 scenes from the same location for 1986, 1989, and 1996 and
>>
>> the first two (1986 and 1989) had the same problem with the null
>>
>> values...
>>
>> sorry I can provide a way to fix it but  I hope somebody will
>>
>> juan c
>>
>>
>>
>>
>>
>>
>> On Tue, May 24, 2011 at 3:49 AM, Chemin, Yann (IWMI)
>>
>> <Y.Chemin at cgiar.org> wrote:
>>
>> Each created visibility map holds only one value r.mapcalc
>>
>> expression="visibility=${vis_list[$i]}" --overwrite so why not just use
>>
>> this visibility value as input in the >parameter file instead of
>>
>> creating a map where all cells have the same value?
>>
>> Yes, this is a valid point, the script is designed to (hopefully soon)
>>
>> encompass raster maps of visibility data (maybe interpolated...), so
>>
>> this is a placeholder for the next steps.
>>
>>
>> -----Original Message-----
>>
>> From: Markus Metz [mailto:markus.metz.giswork at googlemail.com]
>>
>> Sent: Tuesday, May 24, 2011 12:52 PM
>>
>> To: Chemin, Yann (IWMI)
>>
>> Cc: grass-user at lists.osgeo.org; Elena Mezzini; Markus Neteler
>>
>> Subject: Re: [GRASS-user] Re: i.atcorr returns all NULL values
>>
>> On Mon, May 23, 2011 at 9:36 PM, Markus Neteler <neteler at osgeo.org>
>>
>> wrote:
>>
>> Hi Yann,
>>
>> On Mon, May 23, 2011 at 5:07 AM, Chemin, Yann (IWMI)
>>
>> <Y.Chemin at cgiar.org> wrote:
>>
>> p192 r030 image of July 2003 of Italy (L5TM) successfully corrected
>>
>> for band 1 and 2.
>>
>> The following script works well in GRASS GIS Trunk SVN.
>>
>> (better send as attachment since a series of line breaks got broken)
>>
>> ...
>>
>> # For i.atcorr scripting
>>
>> #-----------------------------------------------------
>>
>> vis_list=(10 10 8 9.7 15 8 7 10 10 9.7 12 9.7 7 12 12 12 3 15 12 9.7
>>
>> 6 15) vis_len=${#vis_list[*]} echo $vis_len
>>
>> ... this is not clear to me: why so many values and where do they
>>
>> originate from? Since we have only 6 channels to work with...
>>
>> Moreover, how can it be possible that visibility is different for
>>
>> different channels of the same scene? From the documentation it seems
>>
>> that visibility is related to aerosol concentration at 550nm which can
>>
>> not possibility be different for different bands of the same scene.
>>
>> Each created visibility map holds only one value r.mapcalc
>>
>> expression="visibility=${vis_list[$i]}" --overwrite so why not just use
>>
>> this visibility value as input in the parameter file instead of creating
>>
>> a map where all cells have the same value?
>>
>> Markus M
>>
>> _______________________________________________
>>
>> grass-user mailing list
>>
>> grass-user at lists.osgeo.org
>>
>> http://lists.osgeo.org/mailman/listinfo/grass-user
>>
>>
>>
>> -------------- next part --------------
>>
>> An HTML attachment was scrubbed...
>>
>>
>> URL: http://lists.osgeo.org/pipermail/grass-user/attachments/20110527/211bcf66/attachment-0001.html
>>
>> _______________________________________________
>> grass-user mailing list
>> grass-user at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/grass-user
>>
>
>
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/grass-user
>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Screenshot.png
Type: image/png
Size: 75649 bytes
Desc: not available
Url : http://lists.osgeo.org/pipermail/grass-user/attachments/20110529/f96cac96/Screenshot-0001.png