Stream: CESM-diagnostics

Topic: Calling multiple CESM2-LE decadal files

Holly Olivarez (Mar 29 2024 at 14:21):

Hi all! Using dask and xr.open_mfdataset, I want to call CESM2-LE monthly DIC output from 185001-201412. What is an efficient way to do this? Previously I only needed two decades so I called one decade a time and concatenated them. Thanks in advance for suggestions and/or script you can share.

Anna-Lena Deppenmeier (Mar 29 2024 at 21:40):

Hi Holly, I'm assuming you're already using parallel=True ?

Holly Olivarez (Mar 30 2024 at 03:39):

Hi! Yes. My script is this:

n_members='DIC'

ds_members_DIC=xr.open_mfdataset(DIR+n_members+'/b.e21.BHISTsmbb.f09_g17.LE2-1251.*.pop.h.DIC.185001-185912.nc',
                     chunks={'nlat': 64, 'nlon': 80},
                     concat_dim='member_id',
                     combine='nested',
                     # join='override',
                     coords='minimal',      # uses the coordinates from the first file
                     # data_vars='minimal',
                     autoclose=True,
                     decode_times=False,
                     parallel=True)

Michael Levy (Apr 01 2024 at 14:07):

I responded on the discussion in #dask > Calling multiple decadal output datasets: CESM2-LE before I saw this more specific example -- I think adding compat='override' to the argument list should speed up the read quite a bit.

Last updated: May 16 2025 at 17:14 UTC