TomNicholas

timeseries_hist_datasets = []
timeseries_ssp126_datasets = []
timeseries_ssp245_datasets = []
timeseries_ssp370_datasets = []
timeseries_ssp585_datasets = []
for k,ds in data_timeseries.items():
    # Separate experiments
    out = ds.convert_calendar('standard')
    out = out.sel(time=slice('1850', '2100'))# cut extended runs
    out = out.assign_coords(source_id=ds.source_id)

TomNicholas

from datatree import open_datatree
timeseries.to_zarr('cmip_timeseries')  # or netcdf, with any group structure
roundtrip = open_datatree('cmip_timeseries', engine="zarr")
print(roundtrip)

TomNicholas

DataTree('None', parent=None)
├── DataTree('CESM2')
│   ├── DataTree('historical')
│   │       Dimensions:      (vertex: 4, time: 1980, bnds: 2)
│   │       Coordinates:
│   │         * time         (time) object 1850-01-15 12:59:59.999997 ... 2014-12-15 12:0...
│   │       Dimensions without coordinates: vertex, bnds
│   │       Data variables:
│   │           lat          float64 ...
│   │           lat_bounds   (vertex) float32 ...

TomNicholas

anomaly = DataTree()
for model_name, model in timeseries.children.items():
    
    # model-specific base period as an xarray.Dataset
    base_period = model["historical"].ds.sel(time=slice('1950','1980')).mean('time')
    
    anomaly[model_name] = model - base_period   # subtree - Dataset

TomNicholas

import matplotlib.pyplot as plt
from datatree import map_over_subtree
fig, ax = plt.subplots()

@map_over_subtree
def plot_temp(ds, original_ds):
    
    if ds:
        label = f'{original_ds.attrs["source_id"]} - {original_ds.attrs["experiment_id"]}'
        ds['tos'].rolling(time=2*12).mean().plot(ax=ax, label=label)

TomNicholas

TomNicholas

TomNicholas

TomNicholas

TomNicholas