html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,performed_via_github_app,issue
https://github.com/pydata/xarray/issues/6968#issuecomment-1234368670,https://api.github.com/repos/pydata/xarray/issues/6968,1234368670,IC_kwDOAMm_X85Jkvie,39069044,2022-09-01T14:34:05Z,2022-09-01T14:34:05Z,CONTRIBUTOR,"I don't think `curvefit` works well with datetime coordinates at this point, because everything gets coerced to float by `apply_ufunc`. Probably room for improvement there. At a minimum you would need to specify good guesses (`p0`) and/or bounds in terms of `datetime64[ns]` values.

An easy enough workaround is to assign a separate non-datetime coordinate. This works:
```python
ds = xr.tutorial.open_dataset('air_temperature')
ds = ds.assign_coords({'day':(ds.time - ds.time[0]) / np.timedelta64(1, 'D')}).swap_dims({'time':'day'})

def periodic_season(x, a0, a1, a2, a3):
    # periodic function with both phase amplitude and shift parameters
    return a0 + a1 * np.cos(a2 * x - a3)

dn = ds.curvefit(
    'day', 
    func=periodic_season, 
    p0={'a0':275, 'a1':15, 'a2':2*np.pi/365}
)
```","{""total_count"": 1, ""+1"": 1, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,1355733363