to_xarray(time_dim_mode="auto")

[EddyCMWF]

Is your feature request related to a problem? Please describe.

It would be very useful to have an "auto" option for time_dim_mode to make conversion to xarray more generalisable. It should choose the appropriate time_dim_mode for the grib file provided. The logic to make the decision will be tricky.

Describe the solution you'd like

When using ek_ds.to_xarry(time_dim_mode="auto"), the conversion should identify and appropriate time_dim_mode for the data being converte. I would suggest using the value of the "stream" key in the grib files to make this decision, but this is very open for discussion. stepType is not recommended as the accumulated variables in ERA5 have stepType='fc' (forecast) despite them being more suited to a time_dim_mode="valid_time" representation.

The following stream to time_dim_mode could be used as an example to start:

• oper, wave, enda, ewda -> time_dim_mode="valid_time"
• mmsm, mmsa, mmsf -> time_dim_mode="forecast"

With the functionality in place, we may be able to use better "metadata matches" to make this automatic decision

Describe alternatives you've considered

Relying on user input works, but adds technical debt to simple examples

Additional context

Providing introductary examples of earthkit to the wider community

Organisation

ECMWF

[sandorkertesz]

@EddyCMWF I agree this would be a very useful feature. As for the rules to define the default, I do not think the stream key on it is own would be enough. E.g. when we retrieve data from MARS this combination is forecast data:

stream=oper 
class=od
expver=1
type=fc

but this is analysis:

stream=oper 
class=od
expver=1
type=an

[EddyCMWF]

hmmm, yes, tricky indeed... will get back to you on this, unless you have an idea already