Save NetCDF through Xarray with dimensions given by two coordinates - python-3.x

Question

I have geographical information in a geopandas. I have a function that saves that information to a netcdf through xarray as follows
def write_ncfile(name, new_model, variables):
## Distinct latitudes
lats=new_model.drop_duplicates(["ilat"], keep="first").geometry.y.values
## Distinct Longitudes
lons=new_model.drop_duplicates(["ilon"], keep="first").geometry.x.values
## Temporal store for DataArrays
temporal_dataset = {}
## Dimensions and coordinates
dims = ('lat', 'lon')
coords = dict(lat=lats, lon=lons)
## Variables to save
for variable in variables:
bcvar=new_model[variable].values
## Reshapes data to have sahpe of lats and lons
bcvar=np.reshape(bcvar, (-1, len(lons)))
ds = xr.DataArray(bcvar, dims=dims, coords=coords)
ds.attrs['long_name'] = descriptions[variable]
ds.attrs['_FillValue'] = 0
temporal_dataset[variable] = ds
## Create DataSet
DT=xr.Dataset(temporal_dataset)
## Save to file
makedirs(name, exist_ok = True)
filename="%s/%s.nc"%(name, name)
DT.to_netcdf(filename, format="NETCDF4_CLASSIC")
return None
This code works wonderfully if the underlaying geographical grid is squared (lat lon projection) but now if the projection is not square (ex lambert) then I would need to have the dimensions defined as a 2d array not 1d. I am stumped as to how to do that.
I am trying to achieve something like this in the header of ncdump
dimensions:
lat:dim_lat
lon:dim_lon
variables:
double lat(lat, lon)
double lon(lat, lon)
double var1(lat, lon)
double var2(lat, lon)
Currently the code is saving it as
dimensions:
lat:dim_lat
lon:dim_lon
variables:
double lat(lat)
double lon(lon)
double var1(lat, lon)
double var2(lat, lon)
How can I change this?
Example gdf:
ilat ilon geometry d_p T_P d_v T_V
22 0 0 POINT (-70.95000 -33.30000) 0.000000 0 0.000000 0
0 0 1 POINT (-70.85000 -33.30000) 383.862700 39674 120.439438 12448
1 0 2 POINT (-70.75000 -33.30000) 327.639330 33863 112.502638 11628
2 0 3 POINT (-70.65000 -33.30000) 320.808104 33157 96.602750 9984
3 0 4 POINT (-70.55000 -33.30000) 415.217240 42915 99.144774 10247
23 1 0 POINT (-70.95000 -33.40000) 0.000000 0 0.000000 0
4 1 1 POINT (-70.85000 -33.40000) 56.055971 5787 16.853605 17310
5 1 2 POINT (-70.75000 -33.40000) 6686.807845 690341 1992.373592 205691
6 1 3 POINT (-70.65000 -33.40000) 8812.040534 909749 3512.456618 362623
7 1 4 POINT (-70.55000 -33.40000) 5203.112762 537166 2015.376536 208066
24 2 0 POINT (-70.95000 -33.50000) 0.000000 0 0.000000 0
8 2 1 POINT (-70.85000 -33.50000) 133.485233 13765 40.937021 4222
9 2 2 POINT (-70.75000 -33.50000) 7358.668562 758846 2309.069300 238118
10 2 3 POINT (-70.65000 -33.50000) 10420.377036 1074578 3668.947758 378352
11 2 4 POINT (-70.55000 -33.50000) 6166.780423 635935 2047.500621 211144
12 3 0 POINT (-70.95000 -33.60000) 71.933395 74010 21.287101 2193
13 3 1 POINT (-70.85000 -33.60000) 1154.803477 118952 373.474444 38470
14 3 2 POINT (-70.75000 -33.60000) 1512.189352 155764 466.310819 48033
15 3 3 POINT (-70.65000 -33.60000) 7160.093545 737532 2095.296251 215828
16 3 4 POINT (-70.55000 -33.60000) 4870.217943 501661 1494.220152 153914
17 4 0 POINT (-70.95000 -33.70000) 767.033734 78919 241.884877 24887
18 4 1 POINT (-70.85000 -33.70000) 163.023696 16773 48.526857 4993
19 4 2 POINT (-70.75000 -33.70000) 632.011798 65027 207.326845 21332
20 4 3 POINT (-70.65000 -33.70000) 93.053338 9574 27.787137 2859
And the function usage would be
write_ncfile("Trial", gdf, ["d_p", "d_v"])
In the example above saves the information just fine with the code above, but I need to generalize it so that it works when the grid is not a square grid.

Answer 1

I had the same problem and xarray is not very intuitiv on that point. The solution to create a DataArray with the right dimension (Note this is a 3D example):
data_array = xarray.DataArray(
data,
coords={
"time": timestamps,
"latitude": (["y", "x"], latitude_grid),
"longitude": (["y", "x"], longitude_grid),
},
dims=["time", "y", "x"],
)
So the loop in your code should look like this:
coords = {
"latitude": (["lon", "lat"], latitude_grid),
"longitude": (["lon", "lat"], longitude_grid)}
dims = ['lon', 'lat']
for variable in variables:
bcvar=new_model[variable].values
## Reshapes data to have sahpe of lats and lons
bcvar=np.reshape(bcvar, (-1, len(lons)))
ds = xr.DataArray(bcvar, dims=dims, coords=coords)
ds.attrs['long_name'] = descriptions[variable]
ds.attrs['_FillValue'] = 0
temporal_dataset[variable] = ds
Feel free to adapt this to your final running solution.