[GRASS-user] MODIS HDF conversion in TIF tool.

[Ivan Shmakov]

>>>>> Hamish  <hamish_nospam at yahoo.com> writes:

 >>>> Reprojecting and subsetting of HDF-EOS grid data is, indeed,
 >>>> straight-forward, and usually could be done with, e. g.,
 >>>> gdalwarp(1).  Not so for the swath data.

 >>> Could the swath data be reprojected using gdalwarp's -tps thin
 >>> plate spline function? ISTR that's what -tps was commisioned
 >>> for. (??)

 >> I've tried gdalinfo(1) on AIRS .hdf:

 >> $ gdalinfo AIRS.2007.05.31.214.L2.RetStd.v4.0.9.102.D07281163221.hdf

 >> and it shows no ground control points.  I suppose, this is because
 >> the GCP's are given in separate HDF arrays (Latitude & Longitude),
 >> and not in the metadata.  Solving this problem may make gdalwarp(1)
 >> an alternative to HEG.

 > you can use gdal_translate to add (per scanline) GCPs to the metadata
 > before running gdalwarp.  gdal_translate -gcp pixel line easting
 > northing [elevation]

	Indeed, it works, at least for the AIRS data.  (Although I've
	had to upgrade GDAL 1.3.2, as supplied with Debian 4.0 r1, to
	1.4.3.)

	For those who are interested, the script I've used looks roughly
	as follows:

--cut--
#!/bin/bash
file=AIRS.2007.07.09.079.L2.RetStd.v4.0.9.102.D07190164053.hdf
file_gcp=airs-2007-07-09-gran-79-l2-retstd-tsurfstd-gcp+0.5.tiff
file_out=airs-2007-07-09-gran-79-l2-retstd-tsurfstd-gcp+0.5-laea-10km.tiff

## FIXME: specify ``no projection'' instead?
source_srs='+proj=latlong'
target_srs='+proj=laea +lat_0=55.0 +lon_0=90'

## FIXME: more clean error handling
set -x -e

long_lat_to_gcp () {
    gawk '$1 != -9999 && $2 != -9999 {
              print "-gcp", (NR - 1) % 30 + .5, int ((NR - 1) / 30) + .5, $1, $2;
          }'
}

gdal_translate \
    $(paste <(hdfdump --text "$file" {Long,Lat}itude) | long_lat_to_gcp) \
    "HDF4_EOS:EOS_SWATH:\"$file\":L2_Standard_atmospheric&surface_product:TSurfStd" \
    "$file_gcp"
gdalwarp \
    -tps -s_srs "$source_srs" -t_srs "$target_srs" \
    -tr 1e4 1e4 \
    "$file_gcp" "$file_out"
--cut--

	Is there any chance that GDAL will obtain ``GCPs'' from the
	HDF-EOS file directly?

[...]

 >> The one more problem with the MODIS L2 data is that it consists of
 >> overlapping scans -- MODIS could ``see'' some parts of the Earth in
 >> consequent scans.

[...]

 > I think you may get a much better answer about this if you ask on the
 > gdal-dev mailing list. (www.gdal.org) Lots of smart experts on this
 > stuff there.

	I'm cross-posting it there.