Problems with xarray opening datasets with timedeltas with climada 6.1.0

[ValentinGebhart]

Describe the bug
Creating an environment with climada 6.1.0 installs xarray 2025.06 which has a bug when decoding timdeltas. See this issue: pydata/xarray#10468. The bug is resolved in xarray 2025.07, see https://docs.xarray.dev/en/stable/whats-new.html#whats-new

This bug leads to breaking pipelines when using forecast data which make use of timedeltas, e.g.,

• in the copernicus module (using an envirnoment with climada 6.1.0 results in the error) and
• in downloading meteoswiss OGD data by @Evelyn-M (this was the error, right?)

To Reproduce
Simple example to reproduce

import xarray
import pandas as pd

ds = xarray.Dataset({'lead_time': pd.timedelta_range('0h', '12h', freq='h', unit='s')})
ds.to_netcdf('on-disk2.nc', format='NETCDF4')
xarray.open_dataset('on-disk2.nc')

The above code works if I create an environment with

conda create -n copernicus-dev-mode python=3.11 pip geopandas xarray

which installs xarray2025.12, but results in an error with

conda create -n copernicus-dev-mode python=3.11 pip geopandas climada

which installs xarray2025.06:

ValueError: Failed to decode variable 'lead_time': failed to prevent overwriting existing key dtype in attrs on variable 'lead_time'. This is probably an encoding field used by xarray to describe how a variable is serialized. To proceed, remove this key from the variable's attributes manually.

Question
Is this something we could fix easily, e.g., with a bug release that requires xarray 2025.07? Or just wait until a new release with a climada version that requires later xarray versions?

[peanutfun]

For now, I am skipping the tests for xarray versions that have the bug. However, I don't enforce specific xarray versions for Climada: 9685a81