[GRASS-user] Correction of terrain effects for MODIS data?

GRASS GRASS user
1

I got the 2nd edition of the GRASS-GIS book from a friend.

Excellent! It is one of the best introductions in GIS and RS that I have
seen. Special attention is given to the terminology used in the field of GIS
and RS which is very important... as we tend to use language(s), very often,
not correct.

I tried to follow the example for correction of terrain effects in a MODIS
surface reflectance NIR band
(...as far as I know, MODIS surface reflectance imagery is only corrected
for atmospheric effects).

I don't get the desired results... The output is too bright from what one
would expect it to be.

Any recommendations?

--
View this message in context: http://www.nabble.com/Correction-of-terrain-effects-for-MODIS-data--tf4917622.html#a14074638
Sent from the Grass - Users mailing list archive at Nabble.com.

2

Nikos Alexandris wrote:

I tried to follow the example for correction of terrain effects in a MODIS
surface reflectance NIR band
(...as far as I know, MODIS surface reflectance imagery is only corrected
for atmospheric effects).

I don't get the desired results... The output is too bright from what one
would expect it to be.

Any recommendations?

I've never used it for MODIS, but you could try i.landsat.rgb to see if that
helps. (it wants three 0-255 bands, it doesn't care about it being LANDSAT or
not)

Hamish

      ____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page.
http://www.yahoo.com/r/hs

3

Hi!

Will this approach enable a correction for terrain as well? I don't see
how I can incorporate elevation data or anything relevant.

I used the cosine correction approach which is described in GRASS-GIS in
order to correct a MODIS image. But the result is over-corrected (!) as
mentioned also in the book.

I am looking for a practical approach in order to minimize terrain
effects in a MODIS NIR surface reflectance acquisition. The area in
which I am interested to visually identify the land cover, is
mountainous. So any correction would be of great importance.

Thank you in advance.

On Fri, 2007-11-30 at 15:10 -0800, Hamish wrote:

Nikos Alexandris wrote:
> I tried to follow the example for correction of terrain effects in a MODIS
> surface reflectance NIR band
> (...as far as I know, MODIS surface reflectance imagery is only corrected
> for atmospheric effects).
>
> I don't get the desired results... The output is too bright from what one
> would expect it to be.
>
> Any recommendations?

I've never used it for MODIS, but you could try i.landsat.rgb to see if that
helps. (it wants three 0-255 bands, it doesn't care about it being LANDSAT or
not)

Hamish

      ____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page.
http://www.yahoo.com/r/hs

--
Nikos Alexandris
.
Department of Remote Sensing & Landscape Information Systems
Faculty of Forestry & Environmental Sciences, Albert-Ludwigs-University Freiburg
.
Tel. +49 (0) 761 203 3697 / Fax. +49 (0) 761 203 3701 / Skype: Nikos.Alexandris
.
Address: Tennenbacher str. 4, D-79106 Freiburg i. Br., Germany

4

Hi Nikos,

do you use
MODIS/Terra Surface Reflectance Daily L2G Global 250m SIN Grid V005
http://edcdaac.usgs.gov/modis/mod09gqv5.asp
MODIS/Aqua Surface Reflectance Daily L2G Global 250m SIN Grid V005
http://edcdaac.usgs.gov/modis/myd09gqv5.asp
?

Did you check the ATBD08 (Algorithm Technical Background Document)?
http://modis.gsfc.nasa.gov/data/atbd/atbd_mod08.pdf
But indeed - it doesn't seem to mention it.

here maybe some pointers:
http://scholar.google.com/scholar?q=modis+terrain+correction&hl=en&lr=&scoring=r&as_ylo=2002

Markus

Nikos Alexandris wrote:

Hi!

Will this approach enable a correction for terrain as well? I don't see
how I can incorporate elevation data or anything relevant.

I used the cosine correction approach which is described in GRASS-GIS in
order to correct a MODIS image. But the result is over-corrected (!) as
mentioned also in the book.

I am looking for a practical approach in order to minimize terrain
effects in a MODIS NIR surface reflectance acquisition. The area in
which I am interested to visually identify the land cover, is
mountainous. So any correction would be of great importance.

Thank you in advance.

On Fri, 2007-11-30 at 15:10 -0800, Hamish wrote:

Nikos Alexandris wrote:
> I tried to follow the example for correction of terrain effects in a
MODIS
> surface reflectance NIR band
> (...as far as I know, MODIS surface reflectance imagery is only
corrected
> for atmospheric effects).
>
> I don't get the desired results... The output is too bright from what
one
> would expect it to be.
>
> Any recommendations?

I've never used it for MODIS, but you could try i.landsat.rgb to see if
that
helps. (it wants three 0-255 bands, it doesn't care about it being
LANDSAT or
not)

Hamish

____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page.
http://www.yahoo.com/r/hs

--
Nikos Alexandris
.
Department of Remote Sensing & Landscape Information Systems
Faculty of Forestry & Environmental Sciences, Albert-Ludwigs-University
Freiburg
.
Tel. +49 (0) 761 203 3697 / Fax. +49 (0) 761 203 3701 / Skype:
Nikos.Alexandris
.
Address: Tennenbacher str. 4, D-79106 Freiburg i. Br., Germany

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

--
View this message in context: http://www.nabble.com/Correction-of-terrain-effects-for-MODIS-data--tf4924201.html#a14109516
Sent from the Grass - Users mailing list archive at Nabble.com.

5

Markus,

thank you for the interest.

On Sat, 2007-12-01 at 12:54 -0800, Markus Neteler wrote:

Hi Nikos,

do you use
MODIS/Terra Surface Reflectance Daily L2G Global 250m SIN Grid V005
http://edcdaac.usgs.gov/modis/mod09gqv5.asp
MODIS/Aqua Surface Reflectance Daily L2G Global 250m SIN Grid V005
http://edcdaac.usgs.gov/modis/myd09gqv5.asp
?

Yes. And a earlier image (version 4) in 2006.

Did you check the ATBD08 (Algorithm Technical Background Document)?
http://modis.gsfc.nasa.gov/data/atbd/atbd_mod08.pdf
But indeed - it doesn't seem to mention it.

I ve read almost every line from the MODIS algorithm.

There seems to be an attempt to couple BRDF but no DEM is incorporated
in the approach.

I even contacted Kamel Didan working very close with MODIS-people and
knows the authors of ATBD. He pointed out to look for another paper and,
why not, contact the authors of ATBD. I am planning to do so.

FYI, I attach his reply below as a P.S.

here maybe some pointers:
http://scholar.google.com/scholar?q=modis+terrain+correction&hl=en&lr=&scoring=r&as_ylo=2002

I am looking at it now...

Markus

Best regards,

Nikos
---

P.S.

                              From:
kamel <kamel@Ag.arizona.edu>
                                To:
Nikos Alexandris
<nikos.alexandris@grid.unep.ch>
                                Cc:
kamel@cals.arizona.edu
                           Subject:
Re: Correction for terrain effects
in MODIS surface
reflectance products (?)
                              Date:
Fri, 30 Nov 2007 09:30:25 -0700
(17:30 CET)

No problem, but I'm not an expert on atmosphere correction to give you
the most accurate & up to date answer.

Nonetheless, here is some info.

The current MOD09 surf. corr. algorithm does account for terrain relief,
but at a coarser resolution.
you can read their ATBD (it is very long)
http://modis-sr.ltdri.org/MAIN_SURFACE_PRODUCTAND%20USER%
20GUIDE/atbd_mod08.pdf

For a more accurate description of correction over rugged terrain look
at the Dozier paper:
J. Dozier and J. Frew, 1981: Atmospheric corrections to satellite
radiometric data over
rugged terrain. Remote Sens. Environ.,11,191-205.

Obviously it is important to account for the terrain relief (atmosphere
thickness) and aspect.
But it is quite hard in practice especially since you're limited with
your input data to the atmosphere
correction model. At coarse resolution it is even more complex because
of the complexity of the pixels.

For further information please contact Drs Alex and Eric they may have a
better idea and better suggestions.

Alex Lyapustin alyapust@pop900.gsfc.nasa.gov

Eric Vermote eric@ltdri.org

Good luck,

K.Didan
AZ

Dear Kamel,

I have a question... (again ;-| )!

How important do you think is a correction for terrain effects in a
MODIS surface reflectance acquisition?

AFAIK,

only the MODIS algorithm does only atmospheric correction(s).

I did some test with a MODIS image (using GRASS-GIS) but the result
doesn't look like what I want to have. It looks to bright!

Thank you in advance,

Nikos.

Nikos Alexandris wrote:
>
> Hi!
>
> Will this approach enable a correction for terrain as well? I don't see
> how I can incorporate elevation data or anything relevant.
>
> I used the cosine correction approach which is described in GRASS-GIS in
> order to correct a MODIS image. But the result is over-corrected (!) as
> mentioned also in the book.
>
> I am looking for a practical approach in order to minimize terrain
> effects in a MODIS NIR surface reflectance acquisition. The area in
> which I am interested to visually identify the land cover, is
> mountainous. So any correction would be of great importance.
>
> Thank you in advance.
>
> On Fri, 2007-11-30 at 15:10 -0800, Hamish wrote:
>> Nikos Alexandris wrote:
>> > I tried to follow the example for correction of terrain effects in a
>> MODIS
>> > surface reflectance NIR band
>> > (...as far as I know, MODIS surface reflectance imagery is only
>> corrected
>> > for atmospheric effects).
>> >
>> > I don't get the desired results... The output is too bright from what
>> one
>> > would expect it to be.
>> >
>> > Any recommendations?
>>
>> I've never used it for MODIS, but you could try i.landsat.rgb to see if
>> that
>> helps. (it wants three 0-255 bands, it doesn't care about it being
>> LANDSAT or
>> not)
>>
>>
>> Hamish
>>
>>
>>
>>
>> ____________________________________________________________________________________
>> Never miss a thing. Make Yahoo your home page.
>> http://www.yahoo.com/r/hs
> --
> Nikos Alexandris
> .
> Department of Remote Sensing & Landscape Information Systems
> Faculty of Forestry & Environmental Sciences, Albert-Ludwigs-University
> Freiburg
> .
> Tel. +49 (0) 761 203 3697 / Fax. +49 (0) 761 203 3701 / Skype:
> Nikos.Alexandris
> .
> Address: Tennenbacher str. 4, D-79106 Freiburg i. Br., Germany
>
>
> _______________________________________________
> grass-user mailing list
> grass-user@lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/grass-user
>
>

--
Nikos Alexandris
.
Department of Remote Sensing & Landscape Information Systems
Faculty of Forestry & Environmental Sciences, Albert-Ludwigs-University Freiburg
.
Tel. +49 (0) 761 203 3697 / Fax. +49 (0) 761 203 3701 / Skype: Nikos.Alexandris
.
Address: Tennenbacher str. 4, D-79106 Freiburg i. Br., Germany

6

Hi,

On Saturday 01 December 2007 21:11:28 Nikos Alexandris wrote:

> Did you check the ATBD08 (Algorithm Technical Background Document)?
> http://modis.gsfc.nasa.gov/data/atbd/atbd_mod08.pdf
> But indeed - it doesn't seem to mention it.

I ve read almost every line from the MODIS algorithm.

There seems to be an attempt to couple BRDF but no DEM is incorporated
in the approach.

That's two different things altogether. Atmospheric correction tries to
minimise the effects of the... well, atmosphere :slight_smile: For that, it does need an
idea of the angular reflectance properties of the surface (or assume
something). In order to bring this into the algorithm, the BRDF (the
variation of the reflectance as a function of illumination and viewing
geometry) is used. Apart from the fact that the correction needs to take into
account the width of the atmosphere (and hence needs an estimation of
height), terrain is not a (major) issue for this correction.

I think your problem comes to correct the variations in reflectance of the
terrain in different images. This is the expected behaviour: in areas with
high relief, you are looking at an interesting BRDF function, and small
changes in illumination/viewing geometry will be greater than in flat terrain
(which is more Lambertian-like). The good news is that you can use the BRDF
product to ensure a constant illumination/viewing geometry for all your data.
Using this does away with the geometry of acquisition variations, and
depending on your terrain relief, you can choose the most optimal setup (you
just work out $\rho$ from the kernel parameter estimates in MOD43 product).

Cheers,
José

7

Nikos Alexandris wrote:

> > I don't get the desired results... The output is too bright from what one
> > would expect it to be.
> >
> > Any recommendations?

Hamish:

> I've never used it for MODIS, but you could try i.landsat.rgb to see if
> that helps. (it wants three 0-255 bands, it doesn't care about it being
> LANDSAT or not)

Nikos:

Will this approach enable a correction for terrain as well? I don't see
how I can incorporate elevation data or anything relevant.

I used the cosine correction approach which is described in GRASS-GIS in
order to correct a MODIS image. But the result is over-corrected (!) as
mentioned also in the book.

No, i.landsat.rgb does not touch the data, it only normalizes the color tables
while preserving the relative hues (-p). If your image is simply over-bright it
can lower the visual brightness*, but not fix the contrast.

[*] actually it is (mostly) designed for brightening dark images, so you would
need to edit the script and swap the hard-coded MIN=`r.univar percentile=2` to
$((100-BRIGHTNESS)) and MAX=`r.univar percentile=$BRIGHTNESS` to "2" to make it
dimmer. Or get rid of the "2" and cut from both ends with $BRIGHTNESS.

I guess we could add a new flag to the module to do that if there was demand.

The result is dependent on the image, so not much use for quantitatively
comparing a time series of images with different reflectance angles.

Hamish

      ____________________________________________________________________________________
Be a better sports nut! Let your teams follow you
with Yahoo Mobile. Try it now. http://mobile.yahoo.com/sports;_ylt=At9_qDKvtAbMuh1G1SQtBI7ntAcJ

8

Hello Jose!

Thank you for your clarifications.

Jose Gomez-Dans wrote:

Hi,

On Saturday 01 December 2007 21:11:28 Nikos Alexandris wrote:

> Did you check the ATBD08 (Algorithm Technical Background Document)?
> http://modis.gsfc.nasa.gov/data/atbd/atbd_mod08.pdf
> But indeed - it doesn't seem to mention it.

I ve read almost every line from the MODIS algorithm.

There seems to be an attempt to couple BRDF but no DEM is incorporated
in the approach.

That's two different things altogether. Atmospheric correction tries to
minimise the effects of the... well, atmosphere :slight_smile: For that, it does need
an
idea of the angular reflectance properties of the surface (or assume
something). In order to bring this into the algorithm, the BRDF (the
variation of the reflectance as a function of illumination and viewing
geometry) is used. Apart from the fact that the correction needs to take
into
account the width of the atmosphere (and hence needs an estimation of
height), terrain is not a (major) issue for this correction.

"I just mentioned the BRDF to emphasize that there is no link to any
terrain parameters or to a topographic correction within the description
of the algorithm (ATBD). And it is understandable since it is mainly an
atmospheric correction algorithm."

I think your problem comes to correct the variations in reflectance of the
terrain in different images. This is the expected behaviour: in areas with
high relief, you are looking at an interesting BRDF function, and small
changes in illumination/viewing geometry will be greater than in flat
terrain
(which is more Lambertian-like).

"Here it is interesting, indeed!"

The good news is that you can use the BRDF
product to ensure a constant illumination/viewing geometry for all your
data.
Using this does away with the geometry of acquisition variations, and
depending on your terrain relief, you can choose the most optimal setup
(you
just work out $\rho$ from the kernel parameter estimates in MOD43
product).

So I will look for it... MOD43 (?). What is $\rho\$ exactly?

Anyway, don't bother to reply to this question if it is explained on the
web (for MOD43).

Cheers,
José

Thank you for your time.

Nikos.
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

--
View this message in context: http://www.nabble.com/Correction-of-terrain-effects-for-MODIS-data--tf4924201.html#a14112242
Sent from the Grass - Users mailing list archive at Nabble.com.

9

(...I have to stop using nabble for posting -- it doesn't seem to work
very well -- Sorry for reposting)

Hello Jose!

Thank you for your clarifications.

So I will look for it... MOD43 (?). What is $\rho\$ exactly?
        
Anyway, don't bother to reply to this question if it is explained on the
web (for MOD43).

Thank you for your time.
        
Nikos.

On Sat, 2007-12-01 at 21:54 +0000, José Gómez-Dans wrote:

Hi,

On Saturday 01 December 2007 21:11:28 Nikos Alexandris wrote:
> > Did you check the ATBD08 (Algorithm Technical Background Document)?
> > http://modis.gsfc.nasa.gov/data/atbd/atbd_mod08.pdf
> > But indeed - it doesn't seem to mention it.
>
> I ve read almost every line from the MODIS algorithm.
>
> There seems to be an attempt to couple BRDF but no DEM is incorporated
> in the approach.

That's two different things altogether. Atmospheric correction tries to
minimise the effects of the... well, atmosphere :slight_smile: For that, it does need an
idea of the angular reflectance properties of the surface (or assume
something). In order to bring this into the algorithm, the BRDF (the
variation of the reflectance as a function of illumination and viewing
geometry) is used. Apart from the fact that the correction needs to take into
account the width of the atmosphere (and hence needs an estimation of
height), terrain is not a (major) issue for this correction.

I think your problem comes to correct the variations in reflectance of the
terrain in different images. This is the expected behaviour: in areas with
high relief, you are looking at an interesting BRDF function, and small
changes in illumination/viewing geometry will be greater than in flat terrain
(which is more Lambertian-like). The good news is that you can use the BRDF
product to ensure a constant illumination/viewing geometry for all your data.
Using this does away with the geometry of acquisition variations, and
depending on your terrain relief, you can choose the most optimal setup (you
just work out $\rho$ from the kernel parameter estimates in MOD43 product).

Cheers,
José
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

--
Nikos Alexandris
.
Department of Remote Sensing & Landscape Information Systems
Faculty of Forestry & Environmental Sciences, Albert-Ludwigs-University Freiburg
.
Tel. +49 (0) 761 203 3697 / Fax. +49 (0) 761 203 3701 / Skype: Nikos.Alexandris
.
Address: Tennenbacher str. 4, D-79106 Freiburg i. Br., Germany