[GRASS-user] any tips/advice for workign with MODIS data?

Hi,

I've been trying for 2 or 3 days now to get some nice looking images out
of MODIS TERRA/AQUA 500m land products (ie 7 bands). I've seen the screen
shots and see the quality of the images that can be obtained--but at best
I
some very faded/obscured images. I've loaded the data into GRASS
using HEGTool to convert the HDF file to separate bands. I've then loaded
that and tired combining Bands 143 (RGB) using r.composite. I've tried
using r.colors' standard grey scale and grey.eq before making the composite
image. My values for the images ranges from -28672 to approx. 9000 .

I've tried rescaling the data to get it between 0 and 255, and the
resulting image quality is equally bad.

Any suggestions on what to try next?

I've checked the data out in other programs that are in other research
groups and the data makes nice RGB images--so nothing wrong with the
source data.

Best, Mark Hall

mhall@berkeley.edu writes:

> Hi,

> I've been trying for 2 or 3 days now to get some nice looking images
> out of MODIS TERRA/AQUA 500m land products (ie 7 bands). I've seen
> the screen shots and see the quality of the images that can be
> obtained--but at best I some very faded/obscured images. I've loaded
> the data into GRASS using HEGTool to convert the HDF file to separate
> bands. I've then loaded that and tired combining Bands 143 (RGB)
> using r.composite. I've tried using r.colors' standard grey scale
> and grey.eq before making the composite image. My values for the
> images ranges from -28672

        I guess it may be a fill (``null'') value of the source data
        set. Try to eliminate it with, e. g.:

$ r.null map=BAND setnull=-28672

        and check the range afterwards. You may eliminate HEG's own
        fill value as well with $ ... setnull=0.

> to approx. 9000 .

> I've tried rescaling the data to get it between 0 and 255, and the
> resulting image quality is equally bad.

> Any suggestions on what to try next?

        In the ``Tutorial on Producing MODIS True Color Images'' [1],
        authors suggest a table to be used for translating MODIS
        reflectances into RGB values. You may apply this table to each
        of the bands (via $ r.colors ... color=rules) prior to using
        `r.composite'. IIRC, the result was about of the same quality
        as the example in the tutorual when I've last tried that.

> I've checked the data out in other programs that are in other
> research groups and the data makes nice RGB images--so nothing wrong
> with the source data.

PS. May be a wiki [2] page for MODIS-related issues?

[1] ftp://ftp.ssec.wisc.edu/pub/IMAPP/MODIS/TrueColor/
[2] http://grass.gdf-hannover.de/wiki/

mhall@berkeley.edu wrote on 12/06/2007 09:08 AM:

Hi,

I've been trying for 2 or 3 days now to get some nice looking images out
of MODIS TERRA/AQUA 500m land products (ie 7 bands). I've seen the screen
shots and see the quality of the images that can be obtained--but at best
I
some very faded/obscured images. I've loaded the data into GRASS
using HEGTool to convert the HDF file to separate bands. I've then loaded
that and tired combining Bands 143 (RGB) using r.composite.

I did something like this time ago:
http://mpa.itc.it/rs/modis_reflectance/
and didn't have problems.

Maybe you need to reset the color table to grey and start again?

Markus

        I guess it may be a fill (``null'') value of the source data

set. Try to eliminate it with, e. g.:

$ r.null map=BAND setnull=-28672

        and check the range afterwards. You may eliminate HEG's own

fill value as well with $ ... setnull=0.

> to approx. 9000 .

Thanks much! This solved the bulk of my headaches.

> Any suggestions on what to try next?

        In the ``Tutorial on Producing MODIS True Color Images'' [1],

authors suggest a table to be used for translating MODIS
reflectances into RGB values. You may apply this table to each
of the bands (via $ r.colors ... color=rules) prior to using
`r.composite'. IIRC, the result was about of the same quality
as the example in the tutorual when I've last tried that.

I'll have to play around with that.

PS. May be a wiki [2] page for MODIS-related issues?

I guess this will be a project to take a stab at it during the university
holiday furlough...

Best and Thanks, Mark Hall