The German Weather Service (DWD = Deutscher Wetterdienst) offers Open Data via their Website.

An example is this product https://opendata.dwd.de/weather/radar/composit/rx/raa01-rx_10000-latest-dwd---bin

It is a binary file which contains a composite of precipitation analysis of all the german radar stations, the file format is described in this file radolan_radvor_op_komposit_format_pdf.pdf

With this data I'm able to generate this image:

enter image description here

So this image is in DWD's RADOLAN projection, which is roughly described in this document: Unterstuetzungsdokumente - Verwendung von RADOLAN-Produkten in GIS.pdf but more precisely in the first PDF I mentioned.

The examples in the second PDF refer to the their files in ASCII format, but those are not available for the product I'm interested in.

They have an example of how to use gdalwarp in order to transform their ASCII files in their RADOLAN-projection to GeoTiff with EPSG:3857 projection.

# Native Projektion ("RADOLAN-Projektion"):
R=6370040 # Erdradius (Kugel)
R="+a=$R +b=$R" # zu 'R' zusammenfassen -> Kugel
nat_proj="+proj=stere +lon_0=10.0 +lat_0=90.0 +lat_ts=60.0 $R +units=m"
gdalwarp -co "TILED=YES" -co compress=lzw \
    -s_srs "$nat_proj" \
    -t_srs "EPSG:3857" \
    -r near \
    -overwrite \
    -of GTiff  "RW_20160114-0750.asc" "RW_20160114-0750.tif"

This is what I want to archive, but by applying the transformation directly to my png file.

I have no clue on how to do this. I've tried the following with a gazillion of variants, for hours:

import os
import glob

files = glob.glob("FXLATEST_000_MF002.png")

ncols         900
nrows         900
xllcorner     -523462
yllcorner     -4658645
cellsize      1000
NODATA_value  -1

# -523462,376462,-4657644,-3758644

sources = files

for source in sources:

  commands = []

  if True:
    command = []
    command.append('  -overwrite')
    command.append('  -s_srs "+proj=stere +lon_0=10.0 +lat_0=90.0 +lat_ts=60.0 +a=6370040 +b=6370040 +units=m"')
    command.append('  -t_srs "EPSG:3857"')
    #command.append('  -to SRC_METHOD=NO_GEOTRANSFORM')
    command.append('  -to DST_METHOD=NO_GEOTRANSFORM') # -te xmin ymin xmax ymax
    command.append('  -te_srs "+proj=stere +lon_0=10.0 +lat_0=90.0 +lat_ts=60.0 +a=6370040 +b=6370040 +units=m"')
    #command.append('  -te -523462 -4657644 376462 -3758644')
    #command.append('  -te -00000000 50000000 30000000 80000000') # [-te xmin ymin xmax ymax]
    command.append('  -te 207716.2755078182 5827593.4587101545 1795826.6626303399 7415703.845832676') # [-te xmin ymin xmax ymax]
    command.append('  -r near')
    command.append('  "'+source+'"')
    command.append('  "_tif_'+source+'.tif"')

  if True:
    command = []
    command.append('  -of png')
    command.append('  "_tif_'+source+'.tif"')
    command.append('  "_png_'+source+'.png"')

  print '#####################################################'

  for command in commands:
    print '==========================================='
    for idx, line in enumerate(command):
      if idx > 0:
        print '  ',
      print line

    os.system(' '.join(command))

files = glob.glob("*.asc.tif")
#for file in files:
#  os.remove(file)
files = glob.glob("*.asc.png.aux.xml")
for file in files:

The target image should look something like the following, focus on the outline instead of the colors and content:

enter image description here

That image I got from here: https://maps.dwd.de/geoserver/dwd/wms

The most "prominent" warning I'm getting is "There is no affine transformation and no GCPs.". I believe that this may be crucial, but I don't know what it means. I believe that I'm somehow not telling gradwarp where in that projection space that image segment is located, and possibly also not telling it where i want that image to be located in the target projection, and that this warning is trying to tell me that.

Can someone lead me in the right direction? I'm doing this in a hobbyistic context, I don't know much about reprojections, but this is a project I've done before, so I do know a little bit.

  • The usual workflow is to apply CRS and extent with gdal_translate -a_srs -a_ullr (with the extent in the map source projection), then use gdalwarp to reproject to something useful. But I can't help you doing that in python.
    – AndreJ
    Jan 23, 2018 at 11:58
  • How did you obtain the PNG files? Does the PNG have auxiliary files that describe either its projection (coordinate system) or its geotransform/ground control (position of the image within the coordinate system)? The latter may have a .pgw extension. The error you mention is leading me to think that the input PNG has no georeferencing information and therefore cannot be warped into another coordinate system. While gdalwarp gives an option (-s_srs) to specify the coordinate system, I don't think it gives an option to supply the geotransform. I could be wrong about that though. Jan 23, 2018 at 15:14
  • @AndreJ I'm calling the shell command via python, I'm only using it to batch-process files. How do I get that CRS? Is this some kind of file the DWD should be offering for download?
    – Daniel F
    Jan 23, 2018 at 15:39
  • @LoganByers The bin file in the first link contains data very similar to the ASCII format. I'm parsing that binary file and creating the PNG via PIL. So I have no additional data. I think I could write that data into an ASCII file and then create the GeoTIFF (preferably directly a PNG), but I'd like to try first applying it to the PGN directly. The second image I'm getting from a map server which the DWD hosts, which let's you specify the target projection in the url (I updated the answer to include the url).
    – Daniel F
    Jan 23, 2018 at 15:45
  • @DanielF the CRS should be the polar stereographic projection you have already used in your code.
    – AndreJ
    Jan 23, 2018 at 17:34

1 Answer 1


After the follow ups in the comments on the original question, this is looking like a fairly simple issue of not having a geotransform associated with the PNG. You can associate a geotransform with an image using an ESRI World File that could be constructed by hand once and copied around for different images. World files have the same base file name as the image but an extension of either .wld (generally) or .pgw (PNG) or .tfw (GTIFF).

To make the world file need to know the raster resolution and bounding box. I believe this information is on page 11 of the first PDF you linked, but my German isn't so great. Maybe it is elsewhere in the documentation as well. You might need a little math or coordinate transforms to get the corner coordinates into a usable form.

With the geotransform associated with the PNG you have created you should be able to continue with the script you are currently running.

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.