Please add option for 'weekly' frequency in the SPI family of models

C15H24 commented 2 weeks ago

Addressing a Problem?

No response

Potential Solution

the function standardized_precipitation_index accepts freq = 'W' as signifying the nature of the groupings is weekly - and any downstream adjustments that would need to be made to validate the output

Additional context

Many projects in government are beginning to operate and produce products on a weekly basis

Contribution

[ ] I would be willing/able to open a Pull Request to contribute this feature.

Code of Conduct

[X] I agree to follow this project's Code of Conduct

coxipi commented 2 weeks ago

Hi! We can add this indeed. Just a couple caveats:

It will only work with standard calendars, as .resample(time="W") and .groupby("time.week") is not yet compatible with cf-time (https://github.com/pydata/xarray/discussions/6375)
Since there is not an integer number of weeks in a year, using .resample(time="W").mean() the days included in week1 can vary from year to year, e.g.
```
# days included in week 1 between 2000 and 2002
2000: [Jan 3, Jan 9]
2001: [Jan 1, Jan 7]
2002: [Dec31 2001, Jan 6]
...
```


sometimes Jan 1rst is not included, sometimes Week1 includes days from december, etc. I'm curious to know if this is what you envision when you consider weekly groupings.

C15H24 commented 4 days ago

Yes, I envision strict adherence to ISO-8601 weeks :) So, looks like it would need to be implemented upstream (cftime.datetime) first?

Thank you for your response

Ouranosinc / xclim