Stream: python-questions

Topic: xarray: seasonal-mean timeseries from a monthly timeseries

view this post on Zulip Stephen Yeager (Nov 22 2020 at 21:32):

I'm struggling to find clear web documentation on how to convert a monthly DataArray (time: 720) into a yearly DataArray(time:60) that represents a seasonal average ('DJF' or 'JFM', etc). Apparently, this requires using the resample method, but I find the documentation of xarray.Dataset.resample to be unhelpful. Can anyone share a clear example of how this is done?

view this post on Zulip Elizabeth Maroon (Nov 22 2020 at 21:50):

I'm sure there's a slicker way to do this with resample and/or groupby('time.season'), but here's how I've done this in the past using rolling to calculate running 3-month-means then grabbing every 12th mean:

startmonth=11 #for DJF, if first month = January; D=11 w/0-indexing
endmonth=len(ds['time'])

djfmean=ds.rolling(time=3).mean().isel(time=slice(startmonth,endmonth,12))

view this post on Zulip Elizabeth Maroon (Nov 22 2020 at 22:04):

Oops - look like I had the startmonth off; for DJF should be 13 (for February). By default, rolling using the right side of the window as the time index.

view this post on Zulip Stephen Yeager (Nov 22 2020 at 23:19):

I came up with the following which seems to work: 

def jfm_mean(ds):
    month_length = ds.time.dt.days_in_month
    result = ((ds * month_length).resample(time='QS-JAN').sum() /
          month_length.resample(time='QS-JAN').sum())
    return result.sel(time=result['time.month']==1)

view this post on Zulip Elizabeth Maroon (Nov 22 2020 at 23:24):

Appears to work on a test dataset over here.

Last updated: Jan 30 2022 at 12:01 UTC