Offset Longitude Values in FNL GRIB1 files.

[abrammer]

What happened?

Old FNL files return a longitude dimension of -1 to 358. This seems to be wrong and it should actually be 0 to 359.
Looking at the grib internals, the file seems to be encoded as 0 to -1 with a 1° increment and 360 values.

Dumping to netcdf and visualizing the orography field also suggests that everything is offset by 1°. Although it's kind of hard to tell as it's only 1 grid point.

What are the steps to reproduce the bug?

Grab a grib1 file from RDA
wget https://data.rda.ucar.edu/ds083.2/grib1/1999/1999.08/fnl_19990801_00_00.grib1

import cfgrib
import xarray as xr

local_file = "fnl_19990801_00_00.grib1"
ds = xr.open_dataset(local_file, 
    engine='cfgrib', backend_kwargs={'indexpath':''}, 
    filter_by_keys={'typeOfLevel': 'surface' },
    )
    
print(ds['longitude'])

<xarray.DataArray 'longitude' (longitude: 360)>
array([ -1.,   0.,   1., ..., 356., 357., 358.])

Version

0.9.10.4

Platform (OS and architecture)

x86_64 GNU/Linux

Relevant log output

The dataarray attributes reveal the grib encoding:

  :GRIB_gridType = "regular_ll";
  :GRIB_NV = 0L; // long
  :GRIB_Nx = 360L; // long
  :GRIB_Ny = 181L; // long
  :GRIB_cfName = "geopotential_height";
  :GRIB_cfVarName = "orog";
  :GRIB_gridDefinitionDescription = "Latitude/Longitude Grid";
  :GRIB_iDirectionIncrementInDegrees = 1.0; // double
  :GRIB_iScansNegatively = 0L; // long
  :GRIB_jDirectionIncrementInDegrees = 1.0; // double
  :GRIB_jPointsAreConsecutive = 0L; // long
  :GRIB_jScansPositively = 0L; // long
  :GRIB_latitudeOfFirstGridPointInDegrees = 90.0; // double
  :GRIB_latitudeOfLastGridPointInDegrees = -90.0; // double
  :GRIB_longitudeOfFirstGridPointInDegrees = 0.0; // double
  :GRIB_longitudeOfLastGridPointInDegrees = -1.0; // double

Accompanying data

Land-Sea Mask for UK and Europe highlights issue as well.

Organisation

No response