Does anyone know the appropriate project information to georeference (and hence reproject) Himawari-8 AHI data in the IR? The images are 5500x5500 pixels.

Currently I try these:

gdal_translate -a_srs  "+proj=geos +a=6378169 +b=6356583.8 +lon_0=140.7 +h=35785831" -a_ullr $ulx $uly $lrx $lry HDF5:"TEST"://img tmp.tif

gdalwarp -srcnodata -32767 -s_srs '+proj=geos +lon_0=140.7 +h=35785831 +x_0=0.0' -t_srs '+proj=latlong +datum=WGS84' -tr 0.03 0.03 -te 55.7 -80 225.7 80 -order 3 tmp.tif FINAL.tif

This, however, results in an image that's not quite right. Australia and Indonesia almost match a coastline boundary overlay but as one moves further from the subsatellite point there exists a larger and larger offset. See this comparison image: Reprojected landmask against coastline boundaries

If someone could point me to the correct code to use for reprojection that would be great. I suspect it's the a_ullr values that I have wrong.

  • hi Simon, could you let me know server link for downloading the Hma 8 data? Thanks.
    – Roman
    Commented May 19, 2016 at 9:42
  • The data was received from the satellite operators, I'm not sure if it is publicly available anywhere.
    – os1
    Commented May 19, 2016 at 13:08
  • @jean see the note at data.jma.go.jp/mscweb/en/himawari89/cloud_service/…
    – AndreJ
    Commented May 19, 2016 at 16:02
  • @AndreJ Thanks for the information. I have one more question, the original data seems raw binary format, how OP converted it into tif file at first?
    – Roman
    Commented May 22, 2016 at 12:12
  • I guess they sell netcdf or HDF5 data as well, if you ask for it.
    – AndreJ
    Commented May 22, 2016 at 12:24

1 Answer 1


Your assumption is right, you need the correct values for the corner coordinates.

Since the source image is slightly larger than the Earth, you can not simply reproject the corner coordinates. You may use a different approach:

According to the project site, the 5500x5500pix data is providing a 2km resolution on the center. So the extent from the center is 5500/2*2000m = +/-5500000m.

With that, and the WGS84 ellipsoid data, I used this transformation:

gdal_translate -a_srs "+proj=geos +h=35785863 +a=6378137.0 +b=6356752.3 +lon_0=140.7 +no_defs" -a_ullr -5500000 5500000 5500000 -5500000 PI_H08_20150125_0230_TRC_FLDK_R10_PGPFD.png temp.tif
gdalwarp -overwrite -t_srs "+proj=latlong +ellps=WGS84 +pm=140.7" -wo SOURCE_EXTRA=100 temp.tif Himawari8.tif
  • That did the trick, yes. I hadn't understood exactly how the details of the GEOS projection work. Understood now.
    – os1
    Commented May 19, 2016 at 13:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.