[FWTools] MODIS processing

[Ned Horning]

Frank,

First the good news. Using -s_srs "WGS84" fixed the problem. 

Now the bad news. The resulting image is poorly projected. After digging
into this a bit it seems as if the initial GCP assignment might not be
accurate. I noticed this when I loaded the HDF image into OpenEV_FW and
checked the coordinates of some points around the image (using the cursor
and overlaying a shapefile) and the coordinates were off 10's of kilometers.
I would expect these coordinates to be quite accurate. The processing also
took about 20 minutes for each image I supposed because there were so many
GCPs. When I ran gdalwarp on some Landsat images it was lighting fast. 

I used ENVI to project the same file and that result looks fine so I'm
pretty sure the data is okay (or at least as ok as HDF data ever gets). In
any case, I uploaded a sample HDF file, the output from gdalwarp (band 1
only), the output from ENVI, and a screen shot with a vector overlay on the
HDF image in OpenEV_FW. These are all on my ftp site:
ftp://ftp.amnh.org/pub/people/horning/downloads/MODIS

Any thoughts before I start looking into other options?

One more question. Is it possible to project two images without doing
interpolation using gdalwarp? In other words can you do a rigorous
transformation using all pixels rather than interpolating between points or
applying a polynomial?

Thanks for your help with this.

Ned 


> Ned,
> 
> Normally the GCPs would be marked as WGS84, and that would
> establish the source coordinate system.  I wonder if something
> might be broken.  Can you make the modis dataset in question
> available for me to grab and try?
> 
> You could add "-s_srs WGS84" to the commandline but I suspect
> there is something more wrong.
> 
> > My input was:
> > C:\Program Files\FWTools1.0.0a7>gdalwarp -t_srs "+proj=utm +zone=38
> > +datum=WGS84"
> >
> HDF4_EOS:EOS_SWATH:\AMNH\ModisCloudImages\MOD35_L2.A2000184.0755
> .005.2005182
> > 222430.hdf:mod35:Cloud_Mask modis_cloud.tif
> >
> > Any thoughts?
> >
> > Also, is it possible to select a single band from the HDF subdataset for
> > gdalwarp? I didn't see a band selection switch.
> 
> It seems that gdalwarp does not offer an option for this.
> You could try and subset the bands before warping, or
> after, using gdal_translate.  It would likely be better to do
> it afterwards, and GeoTIFF format is unlikely to hold the
> number GCPs in the MODIS dataset.
> 
> Best regards,
> --
>
---------------------------------------+------------------------------------
--
> I set the clouds in motion - turn up   | Frank Warmerdam,
warmerdam at pobox.com
> light and sound - activate the windows | http://pobox.com/~warmerdam
> and watch the world go round - Rush    | Geospatial Programmer for Rent