I can create a NetCDF file (using the Python netCDF4 library) and load it in QGIS. However, as the size of these files gets sometimes quite big, I try to compress my data. This works very well using
createVariable(h, 'float32', ("y","x"), zlib=True,complevel=3,chunksizes=(1,len(x)),fill_value=np.nan)
However, playing around a bit showed that chunking both dimensions (e.g. chunksizes=(100,100) or chunksizes=(len(y),len(x))) reduced the resulting file size by more than 50%. Unfortunately, in this case I am not able any more to load this grid correctly to QGIS. The only chunksize that works is chunking only the rightmost variable (1,len(x)). I guess, this has something to do with the GDAL. About QGIS says:
QGIS version 2.14.12-Essen
Compiled against GDAL/OGR 1.10.1
Running against GDAL/OGR 1.11.2
I'm working on Ubuntu14.04 and installed QGIS via the debian-ltr PPA. Is there any way to make other chunksizes work too? Would an update to QGIS 2.18 help or is this an issue with GDAL?