[gdal-dev] Warping EUMetsat MSG-4 MPE dataset to webmercator projection

[Andre Joost]

Am 20.04.18 um 22:10 schrieb Even Rouault:
> On jeudi 19 avril 2018 09:17:35 CEST Menno van Scheers - HUSS wrote:
>> Hi,
>>
>> Built version 2.2.4 of gdal with no additional libraries.
>> Currently trying to warp a grib dataset to webmercator projection with the
>> following arguments : gdalwarp  -t_srs "+proj=geos +lon_0=0 +h=35785831
>> +x_0=0 +y_0=0 +a=6378140 +b=6356755 +units=m +no_defs"
>> 201804181115_L-000-MSG4__-MPEF________-MPEG_____-000000___-201804181115-__.
>> grib webm.tiff
>>
>> This task fails with : ERROR 1: Too many points (5625 out of 5625) failed to
>> transform, unable to compute output bounds.
>>
>> Source dataset has a geostationary projection
>> Basic info on the dataset :
>>
>> PROJ.4 string is:
>> '+proj=geos +lon_0=0 +h=35785831 +x_0=0 +y_0=0 +a=6378140 +b=6356755
>> +units=m +no_defs ' Origin =
>> (-5570248.477582973428071,5570248.477582973428071)
>> Pixel Size = (3000.403165948275728,-3000.403165948275728)
>> Corner Coordinates:
>> Upper Left  (-5570248.478, 5570248.478) ERROR 1: tolerance condition error
>>
>> Lower Left  (-5570248.478,-5567248.074) ERROR 1: tolerance condition error
>>
>> Upper Right ( 5567248.074, 5570248.478) ERROR 1: tolerance condition error
>>
>> Lower Right ( 5567248.074,-5567248.074) ERROR 1: tolerance condition error
>>
>>
>> Example of the dataset :
>> https://hussbv-my.sharepoint.com/:u:/g/personal/menno_huss_nl/EQwTxBKwPQlOjP
>> Z_DZcuZWcBgBs0y96_9Vev3tIqnne3KA?e=k6hmYU
>
> Menno,
>
> Yes gdalwarp has a hard-time with such images since a lot of points in them correspond to
> points not on the Earth surface...
> The workaround is to explicitly specify the target extent (in the target SRS), so for full world
> coverage:
> -te -20037508.34 -20037508.34 20037508.34 20037508.34
>


I tried:
gdal_translate -a_srs "+proj=geos +h=35785831 +a=6378169 +b=6356583.8 
+no_defs" -b 1 
201804181115_L-000-MSG4__-MPEF________-MPEG_____-000000___-201804181115-__.grib 
temp.tif
gdalwarp -t_srs EPSG:3857 -te -20037508.34 -20037508.34 20037508.34 
20037508.34 temp.tif output.tif

but still get the same too much points. Even with selecting a smaller 
target extent that should be within the visible sector.

Greetings,
Andre Joost