Time-dimension in output `*.nc`-file not read properly by `xarray`

[ghendrickx]

The issue

When opening the output *.nc-file with xarray, a ValueError is raised as the time-variable cannot be parsed to pandas. Below an MWE:

import xarray as xr

out = xr.open_dataset('/model/directory/aeolis.nc')

The above results in the following ValueError being raised:

ValueError: unable to decode time units "seconds since ['2025-01-01' '00:00:00']" with "calendar 'julian'". Try opening your dataset with decode_times=False or installing cftime if it is not installed.
Raised while decoding variable 'time' with value <xarray.Variable (time: 366)> Size: 3kB
[366 values with dtype=float64]
Attributes:
    long_name:      time
    standard_name:  time
    units:          seconds since ['2025-01-01' '00:00:00']
    calendar:       julian
    axis:           T
    bounds:         time_bounds

The possible solution?

I am not 100% sure how to correct this, but I suspect that it has something to do with the reference date being read as a list of two str-values: "seconds since ['2025-01-01' '00:00:00']". Possibly, a join of the str-values might do the trick:

old_ref_date = ['2025-01-01' '00:00:00']
new_ref_date = ' '.join(old_ref_date)
print(new_ref_date)
>>> '2025-01-01 00:00:00'

The new_ref_date should be parsed properly to the date-handler allowing the decoding of the time-variable.

The work-around

I have found a (temporary) work-around to be able to read the output *.nc-file without the ValueError being raised. As the problem arises when decoding the time-variable, disabling this auto-decoding prevents the ValueError from being raised:

import xarray as xr

out = xr.open_dataset('/model/directory/aeolis.nc', decode_times=False)

The time-variable will be formatted as seconds since instead of dates (and times). As a result, this has to be done afterwards (if this formatting is desired), e.g.:

import datetime

out['time'] = [datetime.datetime(<year>, <month>, <day>, ...) + datetime.timedelta(seconds=t) for t in out['time'].values]