problem reading netcdf

Question

I am using qgis 2.2 (I have also tried this in qgis 2.0 and arcgis 10 with the same result) (I have some basic understanding of netcdf, but I am probably missing a lot of details...)

I have manged to read in some netcdf-files and have them displayed correctly. Some other fails miserably, as the coordinate infomation is ignored and the file is displayed south- and eastwards from 0,0 ... Is this a known problem? I have been looking at the files with ncdump, but have not been able to spot any differences that should make one of them work and the other to fail. When looking at the metadata in qgis or arcgis, it is clear that a lot is missing for those files that do not display correctly.

Edit: When converting the files from netcdf4 to netcdf3 they worked fine. is this a known issue in arcgis / qgis? (using arcgis 10, it seems to use gdal, but I cannot find which version, qgis 2.2 from osgeo4w using gdal 1.10)

Edit2: The original files were so large it was a bit difficult to share them. Now I have made an example file that behaves the same way: http://sickel.net/div/test.nc - it is created from http://sickel.net/div/test.cdl - it should, as far as I do understand, cover an area from -160 to -15 E, 20 - 60 N ie most of US main land and neary to UK in the atlantic ocean, but it ends up from 0,0 and south and eastwards in qgis, in arcmap 10.0, it ends up as a point on -160,20 ...

• underdark: Really seems like your input data is broken. Can you share it? Is there a chance to get it fixed by the provider?
• MortenSickel: It seems like qgis (at least with the gdal version in recent osgeo4w) does not read netcdf4 - I got the provider to convert the data into netcdf3 and it all worked fine...

Answer 1

Same problem with very simple netcdf files created by me. The dimensions/variables lat and lon are not interpreted correctly when in netcdf4 format. After converting with to NetCDF3 with ncks tool it works fine:

ncks -3 original.nc new_nc3.nc

Running qgis 2.12.3-Lyon on Windows with gdal 1.11.2

Dimensions are lat and lon in degrees. In CDL the file header looks like this:

netcdf 201506円_classicf {
dimensions:
 time = 1 ;
 lat = 2401 ;
 lon = 1667 ;
variables:
 float SMMean(time, lat, lon) ;
 SMMean:_FillValue = 9.96921e+36f ;
 SMMean:long_name = "non-living suspended matters mean" ;
 SMMean:standard_name = "" ;
 SMMean:units = "1e-3 kg m-3" ;
 float lat(lat) ;
 float lon(lon) ;
 int time(time) ;
// global attributes:
 :history = "Wed Mar 2 15:06:07 2016: ncap2 -s SMMean=float(SMMean) 201506_classic.nc 201506_classicf.nc\n",
 :NCO = "4.4.8" ;
 :nco_openmp_thread_number = 1 ;
data:
 SMMean =
(...)

• Simbamangu: Where is the 'ncks' tool from?
• tiagoams: ncks is part of the nco tools nco.sourceforge.net

Answer 2

There are bugs in NetCDF in GDAL - see the list at the bottom of http://trac.osgeo.org/gdal/wiki/NetCDF for those known/identified.

It is a bit difficult to say which of those bugs is affecting your particular situation since you didn't describe the data, but perhaps the "rotated pole" support problem at http://trac.osgeo.org/gdal/ticket/4285 is the particular "known problem" you were looking for.