Appendix B — Recipes

Small, copyable procedures referred to in the chapters. Variable and file names are placeholders.

B.1 See what a file declares

ncdump -h file.nc                                  # attributes, grid_mapping, standard_name
gdalinfo NETCDF:"file.nc":VARIABLE                 # classic model: geotransform or GEOLOCATION
gdal mdim info file.nc                             # multidimensional model (GDAL 3.11+ CLI)

import xarray as xr
ds = xr.open_dataset("file.nc")
print(ds["VARIABLE"].attrs.get("grid_mapping"), ds["VARIABLE"].attrs.get("coordinates"))
print({k: v.dims for k, v in ds.coords.items()})   # 1D or 2D coordinates?

B.2 Assign a regular grid without resampling

GDAL, as a connection string usable anywhere GDAL opens data:

vrt://file.nc?sd_name=VARIABLE&a_srs=EPSG:3412&a_ullr=-3950000,4350000,3950000,-3950000

a_ullr is upper-left x, upper-left y, lower-right x, lower-right y. a_gt assigns a full geotransform instead. sd_name needs GDAL 3.9 or later.

R (terra):

library(terra)
r <- rast("file.nc", subds = "VARIABLE")
ext(r) <- c(-3950000, 3950000, -3950000, 4350000)   # xmin, xmax, ymin, ymax
crs(r) <- "EPSG:3412"

Python (rioxarray):

import rioxarray
from affine import Affine
da = da.rio.write_transform(Affine(25000, 0, -3950000, 0, -25000, 4350000))
da = da.rio.write_crs("EPSG:3412")

B.3 Read a rotated-pole file as a regular grid

gdal mdim info file.nc                             # works as is
gdalinfo --config GDAL_NETCDF_VERIFY_DIMS STRICT \
         --config GDAL_NETCDF_IGNORE_XY_AXIS_NAME_CHECKS YES \
         NETCDF:"file.nc":tas                      # classic model

terra::rast("file.nc", md = TRUE)

B.4 Repair a broken CF grid mapping (MAR example)

Work on a copy. With RNetCDF:

library(RNetCDF)
nc <- open.nc("copy.nc", write = TRUE)
att.put.nc(nc, "tas", "grid_mapping", "NC_CHAR", "stereographic")   # name the real variable
att.put.nc(nc, "stereographic", "grid_mapping_name", "NC_CHAR", "polar_stereographic")
att.put.nc(nc, "stereographic", "latitude_of_projection_origin", "NC_DOUBLE", -90)
att.put.nc(nc, "stereographic", "straight_vertical_longitude_from_pole", "NC_DOUBLE", 20)
att.put.nc(nc, "stereographic", "standard_parallel", "NC_DOUBLE", -71)
close.nc(nc)

Keep semi_major_axis and inverse_flattening; delete the non-CF names (central_meridian, latitude_of_origin, authority). GDAL then built +proj=stere +lat_0=-90 +lat_ts=-71 +lon_0=20 +a=6378137 +rf=298.2572.

B.5 Make a minimal reproducible example

Use NCO, which copies faithfully. CDO regenerates the grid and can erase the very bug being reported.

ncks -h -3 -O -C -v tas,stereographic,x,y,time -d time,0 -d x,0,,40 -d y,0,,40 in.nc step1.nc
ncatted -h -O -a ,global,d,, -a Conventions,global,c,c,"CF-1.6" step1.nc reprex.nc

-3 writes classic netCDF, avoiding HDF5 container overhead; -h on both commands keeps the history attribute from leaking the original file name.