Datasets concatenated along string dimension cannot write to netCDF

[digbyr]

What happened?

If you create a dataset with xr.concat along a new dimension of strings, e.g.

ds0 = xr.Dataset(data_vars={'dat':('x',np.arange(3))},
                 coords={'x': np.arange(3)})
ds1 = xr.Dataset(data_vars={'dat':('x',np.arange(3)*10)},
                 coords={'x': np.arange(3)})
ds = xr.concat([ds0,ds1],dim=pd.Index(['run0','run1'],name='run'))

the dataset is successfully created, but raises a ValueError if written to netCDF:

ds.to_netcdf('test.nc')
> ValueError: unsupported dtype for netCDF4 variable: StringDType(na_object=nan)

However, the netCDF file is actually created and can be opened without issue. This makes me think that the ValueError is spurious. The issue seems related to those in #10553 but don't seem to have been fixed by #11152.

I've tested this with fresh environments using xarray v2026.2.0 and pandas 3.0.1, or xarray 2026.1.0 and pandas 3.0.0, and get the same error in both. But this kind of concatenation used to work for me in older environments.

What did you expect to happen?

Expected result: the dataset writes to netCDF without complaining.

Minimal Complete Verifiable Example

# /// script
# requires-python = ">=3.11"
# dependencies = [
#   "xarray[complete]@git+https://github.com/pydata/xarray.git@main",
# ]
# ///
#
# This script automatically imports the development branch of xarray to check for issues.
# Please delete this header if you have _not_ tested this script with `uv run`!

import xarray as xr
import pandas as pd
import numpy as np

ds0 = xr.Dataset(data_vars={'dat':('x',np.arange(3))},
                 coords={'x': np.arange(3)})

ds1 = xr.Dataset(data_vars={'dat':('x',np.arange(3)*10)},
                 coords={'x': np.arange(3)})

runlist = np.array(['run0','run1'])
ds = xr.concat([ds0,ds1],dim=pd.Index(runlist,name='run'))

# this step raises a ValueError
ds.to_netcdf('test.nc')

# if you comment out the above lines and just open the file,
# it is present and the contents look as expected
ds = xr.open_dataset('test.nc')
print(ds)

Steps to reproduce

No response

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.
• Recent environment — the issue occurs with the latest version of xarray and its dependencies.

Relevant log output

Traceback (most recent call last):
  File "/fs/homeu3/eccc/crd/cccma/rud001/scrap.py", line 20, in <module>
    ds.to_netcdf('test.nc')
    ~~~~~~~~~~~~^^^^^^^^^^^
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/core/dataset.py", line 2123, in to_netcdf
    return to_netcdf(  # type: ignore[return-value]  # mypy cannot resolve the overloads:(
        self,
    ...<10 lines>...
        auto_complex=auto_complex,
    )
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/backends/writers.py", line 441, in to_netcdf
    dump_to_store(
    ~~~~~~~~~~~~~^
        dataset, store, writer, encoding=encoding, unlimited_dims=unlimited_dims
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/backends/writers.py", line 491, in dump_to_store
    store.store(variables, attrs, check_encoding, writer, unlimited_dims=unlimited_dims)
    ~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/backends/common.py", line 533, in store
    self.set_variables(
    ~~~~~~~~~~~~~~~~~~^
        variables, check_encoding_set, writer, unlimited_dims=unlimited_dims
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/backends/common.py", line 571, in set_variables
    target, source = self.prepare_variable(
                     ~~~~~~~~~~~~~~~~~~~~~^
        name, v, check, unlimited_dims=unlimited_dims
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/backends/netCDF4_.py", line 633, in prepare_variable
    datatype = _get_datatype(
        variable, self.format, raise_on_invalid_encoding=check_encoding
    )
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/backends/netCDF4_.py", line 165, in _get_datatype
    return _nc4_dtype(var)
  File "/home/ords/crd/cccma/rud001/envs/invproc/lib/python3.14/site-packages/xarray/backends/netCDF4_.py", line 186, in _nc4_dtype
    raise ValueError(f"unsupported dtype for netCDF4 variable: {var.dtype}")
ValueError: unsupported dtype for netCDF4 variable: StringDType(na_object=nan)

Anything else we need to know?

No response

Environment

Details

INSTALLED VERSIONS

commit: None
python: 3.14.3 | packaged by conda-forge | (main, Feb 9 2026, 21:56:02) [GCC 14.3.0]
python-bits: 64
OS: Linux
OS-release: 5.14.0-427.100.1.el9_4.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: ('en_US', 'UTF-8')
libhdf5: 1.14.6
libnetcdf: 4.9.3

xarray: 2026.2.0
pandas: 3.0.1
numpy: 2.3.5
scipy: 1.17.0
netCDF4: 1.7.4
pydap: 3.5.8
h5netcdf: 1.8.1
h5py: 3.15.1
zarr: 3.1.5
cftime: 1.6.5
nc_time_axis: 1.4.1
iris: None
bottleneck: 1.6.0
dask: 2026.1.2
distributed: 2026.1.2
matplotlib: 3.10.8
cartopy: 0.25.0
seaborn: 0.13.2
numbagg: 0.9.4
fsspec: 2026.2.0
cupy: None
pint: None
sparse: 0.18.0
flox: 0.11.1
numpy_groupies: 0.11.3
setuptools: None
pip: 26.0.1
conda: None
pytest: None
mypy: None
IPython: None
sphinx: None

[welcome]

Thanks for opening your first issue here at xarray! Be sure to follow the issue template!
If you have an idea for a solution, we would really welcome a Pull Request with proposed changes.
See the Contributing Guide for more.
It may take us a while to respond here, but we really value your contribution. Contributors like you help make xarray better.
Thank you!

[digbyr]

I was too hasty in saying that the contents of the netCDF file are what we expect. It would appear that the ValueError interrupts the to_netcdf() process midway through. Printing out the contents of ds before attempting to save it:

<xarray.Dataset> Size: 104B
Dimensions:  (run: 2, x: 3)
Coordinates:
  * run      (run) StringDType(na_object=nan) 32B 'run0' 'run1'
  * x        (x) int64 24B 0 1 2
Data variables:
    dat      (run, x) int64 48B 0 1 2 0 10 20

and reading it in from file:

<xarray.Dataset> Size: 72B
Dimensions:  (run: 2, x: 3)
Coordinates:
  * x        (x) int64 24B 0 1 2
Dimensions without coordinates: run
Data variables:
    dat      (run, x) int64 48B ...

Similarly, before writing to netCDF, ds.run.values yields ['run0' 'run1'] but after reading it in from file we get [0 1].

[headtr1ck]

Looks like we don't support the new StringDType of pandas 3 yet, see

Migration guide for the new string data type (pandas 3.0) — pandas 3.0.1 documentation https://pandas.pydata.org/docs/user_guide/migration-3-strings.html#string-migration-guide

Since netcdf probably will not support that pandas specific stuff I guess we need to add a wrapper for it.

In the meantime you can try to pin pandas to < 3.

[jsignell]

Right so xarray is gradually figuring out where all StringDType support needs to be added in the meantime if you want to stay on pandas 3 you can just specify the dtype explicitly dtype=runlist.dtype in the call to pd.Index

ds0 = xr.Dataset(data_vars={'dat':('x',np.arange(3))},
                 coords={'x': np.arange(3)})

ds1 = xr.Dataset(data_vars={'dat':('x',np.arange(3)*10)},
                 coords={'x': np.arange(3)})

runlist = np.array(['run0','run1'])
ds = xr.concat([ds0,ds1],dim=pd.Index(runlist,name='run',dtype=runlist.dtype))

ds.to_netcdf('test.nc')

ds = xr.open_dataset('test.nc')
print(ds)

[digbyr]

Aha! I'd found the workaround of re-assigning the run coordinates after concatenation, eg

runlist = np.array(['run0','run1'])
ds = xr.concat([ds0,ds1],dim=pd.Index(runlist,name='run'))
ds = ds.assign_coords({'run':runlist})

but your solution is much cleaner. Thanks!

[jsignell]

Just to clarify, internally xarray is converting the pandas.dtypes to numpy.dtypes. The issue is that the netcdf backend doesn't know how to encode np.dtypes.StringDType if you try writing to zarr for instance it works fine. Similarly you can set the dtype to a fixed length string and that will write fine to netcdf.